Google Git
Sign in
apache / geode / 6e3e43c1fa28b4d4b626f9231d677a365051d873 / . / geode-core / src / main / java / org / apache / geode / internal / cache / Oplog.java
blob: 7838107b9517376bab90d72513943f8d0b2bb839 [file] [log] [blame]
/*
* Licensed to the Apache Software Foundation (ASF) under one or more contributor license
* agreements. See the NOTICE file distributed with this work for additional information regarding
* copyright ownership. The ASF licenses this file to You under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance with the License. You may obtain a
* copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software distributed under the License
* is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express
* or implied. See the License for the specific language governing permissions and limitations under
* the License.
*/
package org.apache.geode.internal.cache;
import java.io.BufferedInputStream;
import java.io.BufferedOutputStream;
import java.io.DataInput;
import java.io.DataInputStream;
import java.io.DataOutput;
import java.io.DataOutputStream;
import java.io.EOFException;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.InterruptedIOException;
import java.nio.ByteBuffer;
import java.nio.channels.ClosedChannelException;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.Comparator;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentMap;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicLong;
import java.util.concurrent.locks.Lock;
import java.util.concurrent.locks.ReentrantLock;
import it.unimi.dsi.fastutil.ints.Int2ObjectMap;
import it.unimi.dsi.fastutil.ints.Int2ObjectOpenHashMap;
import it.unimi.dsi.fastutil.longs.Long2ObjectMap;
import it.unimi.dsi.fastutil.longs.Long2ObjectOpenHashMap;
import it.unimi.dsi.fastutil.objects.ObjectIterator;
import org.apache.logging.log4j.Logger;
import org.apache.geode.CancelException;
import org.apache.geode.DataSerializer;
import org.apache.geode.SerializationException;
import org.apache.geode.annotations.Immutable;
import org.apache.geode.cache.CacheClosedException;
import org.apache.geode.cache.CacheWriterException;
import org.apache.geode.cache.DiskAccessException;
import org.apache.geode.cache.EntryDestroyedException;
import org.apache.geode.cache.EntryEvent;
import org.apache.geode.cache.EntryNotFoundException;
import org.apache.geode.cache.RegionDestroyedException;
import org.apache.geode.cache.TimeoutException;
import org.apache.geode.distributed.OplogCancelledException;
import org.apache.geode.distributed.internal.DistributionManager;
import org.apache.geode.internal.Assert;
import org.apache.geode.internal.HeapDataOutputStream;
import org.apache.geode.internal.InternalDataSerializer;
import org.apache.geode.internal.InternalStatisticsDisabledException;
import org.apache.geode.internal.Sendable;
import org.apache.geode.internal.cache.DiskInitFile.DiskRegionFlag;
import org.apache.geode.internal.cache.DiskStoreImpl.OplogCompactor;
import org.apache.geode.internal.cache.DiskStoreImpl.OplogEntryIdSet;
import org.apache.geode.internal.cache.DistributedRegion.DiskPosition;
import org.apache.geode.internal.cache.entries.DiskEntry;
import org.apache.geode.internal.cache.entries.DiskEntry.Helper.Flushable;
import org.apache.geode.internal.cache.entries.DiskEntry.Helper.ValueWrapper;
import org.apache.geode.internal.cache.eviction.EvictionController;
import org.apache.geode.internal.cache.eviction.EvictionList;
import org.apache.geode.internal.cache.persistence.BytesAndBits;
import org.apache.geode.internal.cache.persistence.DiskRecoveryStore;
import org.apache.geode.internal.cache.persistence.DiskRegionView;
import org.apache.geode.internal.cache.persistence.DiskStoreID;
import org.apache.geode.internal.cache.persistence.UninterruptibleFileChannel;
import org.apache.geode.internal.cache.persistence.UninterruptibleRandomAccessFile;
import org.apache.geode.internal.cache.versions.CompactVersionHolder;
import org.apache.geode.internal.cache.versions.RegionVersionHolder;
import org.apache.geode.internal.cache.versions.RegionVersionVector;
import org.apache.geode.internal.cache.versions.VersionHolder;
import org.apache.geode.internal.cache.versions.VersionSource;
import org.apache.geode.internal.cache.versions.VersionStamp;
import org.apache.geode.internal.cache.versions.VersionTag;
import org.apache.geode.internal.logging.log4j.LogMarker;
import org.apache.geode.internal.offheap.OffHeapHelper;
import org.apache.geode.internal.offheap.ReferenceCountHelper;
import org.apache.geode.internal.offheap.StoredObject;
import org.apache.geode.internal.offheap.annotations.Released;
import org.apache.geode.internal.offheap.annotations.Retained;
import org.apache.geode.internal.sequencelog.EntryLogger;
import org.apache.geode.internal.serialization.ByteArrayDataInput;
import org.apache.geode.internal.serialization.KnownVersion;
import org.apache.geode.internal.serialization.Versioning;
import org.apache.geode.internal.serialization.VersioningIO;
import org.apache.geode.internal.shared.NativeCalls;
import org.apache.geode.internal.util.BlobHelper;
import org.apache.geode.logging.internal.log4j.api.LogService;
import org.apache.geode.pdx.internal.PdxWriterImpl;
import org.apache.geode.util.internal.GeodeGlossary;
/**
* Implements an operation log to write to disk. As of prPersistSprint2 this file only supports
* persistent regions. For overflow only regions see {@link OverflowOplog}.
*
* @since GemFire 5.1
*/
public class Oplog implements CompactableOplog, Flushable {
private static final Logger logger = LogService.getLogger();
/** Extension of the oplog file * */
public static final String CRF_FILE_EXT = ".crf";
public static final String DRF_FILE_EXT = ".drf";
public static final String KRF_FILE_EXT = ".krf";
/** The file which will be created on disk * */
private File diskFile;
/** boolean marked true when this oplog is closed * */
private volatile boolean closed;
private final OplogFile crf = new OplogFile();
private final OplogFile drf = new OplogFile();
private final KRFile krf = new KRFile();
/** The stats for this store */
private final DiskStoreStats stats;
/** The store that owns this Oplog* */
private final DiskStoreImpl parent;
/**
* The oplog set this oplog is part of
*/
private final PersistentOplogSet oplogSet;
/** oplog id * */
protected final long oplogId;
/** recovered gemfire version * */
protected KnownVersion gfversion;
/**
* Recovered version of the data. Usually this is same as {@link #gfversion} except for the case
* of upgrading disk store from previous version in which case the keys/values are carried forward
* as is and need to be interpreted in load by latest product code if required.
*/
protected KnownVersion dataVersion;
/** Directory in which the file is present* */
private DirectoryHolder dirHolder;
/** The max Oplog size (user configurable) * */
private final long maxOplogSize;
private long maxCrfSize;
private long maxDrfSize;
private final AtomicBoolean hasDeletes = new AtomicBoolean();
private boolean firstRecord = true;
/**
* This system property instructs that writes be synchronously written to disk and not to file
* system. (Use rwd instead of rw - RandomAccessFile property)
*/
private static final boolean SYNC_WRITES =
Boolean.getBoolean(GeodeGlossary.GEMFIRE_PREFIX + "syncWrites");
/**
* The HighWaterMark of recentValues.
*/
private final AtomicLong totalCount = new AtomicLong(0);
/**
* The number of records in this oplog that contain the most recent value of the entry.
*/
private final AtomicLong totalLiveCount = new AtomicLong(0);
private final ConcurrentMap<Long, DiskRegionInfo> regionMap =
new ConcurrentHashMap<Long, DiskRegionInfo>();
/**
* Set to true once compact is called on this oplog.
*
* @since GemFire prPersistSprint1
*/
private volatile boolean compacting = false;
/**
* Set to true after the first drf recovery.
*/
private boolean haveRecoveredDrf = true;
/**
* Set to true after the first crf recovery.
*/
private boolean haveRecoveredCrf = true;
private OpState opState;
/**
* Written to CRF, and DRF.
*/
private static final byte OPLOG_EOF_ID = 0;
private static final byte END_OF_RECORD_ID = 21;
/**
* Written to CRF and DRF. Followed by 16 bytes which is the leastSigBits and mostSigBits of a
* UUID for the disk store we belong to. 1: EndOfRecord Is written once at the beginning of every
* oplog file.
*/
private static final byte OPLOG_DISK_STORE_ID = 62;
static final int OPLOG_DISK_STORE_REC_SIZE = 1 + 16 + 1;
/**
* Written to CRF. Followed by 8 bytes which is the BASE_ID to use for any NEW_ENTRY records. 1:
* EndOfRecord Only needs to be written once per oplog and must preceed any OPLOG_NEW_ENTRY_0ID
* records.
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_NEW_ENTRY_BASE_ID = 63;
static final int OPLOG_NEW_ENTRY_BASE_REC_SIZE = 1 + 8 + 1;
/**
* Written to CRF. The OplogEntryId is +1 the previous new_entry OplogEntryId. Byte Format: 1:
* userBits RegionId 4: valueLength (optional depending on bits) valueLength: value bytes
* (optional depending on bits) 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_NEW_ENTRY_0ID = 64;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 1 byte. Byte Format: 1: userBits 1: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_1ID = 65;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 2 bytes. Byte Format: 1: userBits 2: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_2ID = 66;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 3 bytes. Byte Format: 1: userBits 3: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_3ID = 67;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 4 bytes. Byte Format: 1: userBits 4: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_4ID = 68;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 5 bytes. Byte Format: 1: userBits 5: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_5ID = 69;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 6 bytes. Byte Format: 1: userBits 6: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_6ID = 70;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 7 bytes. Byte Format: 1: userBits 7: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_7ID = 71;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 8 bytes. Byte Format: 1: userBits 8: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_8ID = 72;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 1 byte. Byte Format: 1: userBits 1: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_1ID = 73;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 2 bytes. Byte Format: 1: userBits 2: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_2ID = 74;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 3 bytes. Byte Format: 1: userBits 3: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_3ID = 75;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 4 bytes. Byte Format: 1: userBits 4: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_4ID = 76;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 5 bytes. Byte Format: 1: userBits 5: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_5ID = 77;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 6 bytes. Byte Format: 1: userBits 6: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_6ID = 78;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 7 bytes. Byte Format: 1: userBits 7: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_7ID = 79;
/**
* Written to CRF. The OplogEntryId is relative to the previous mod_entry OplogEntryId. The signed
* difference is encoded in 8 bytes. Byte Format: 1: userBits 8: OplogEntryId RegionId 4:
* valueLength (optional depending on bits) valueLength: value bytes (optional depending on bits)
* 4: keyLength keyLength: key bytes 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_MOD_ENTRY_WITH_KEY_8ID = 80;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 1 byte. Byte Format: 1: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_1ID = 81;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 2 bytes. Byte Format: 2: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_2ID = 82;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 3 bytes. Byte Format: 3: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_3ID = 83;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 4 bytes. Byte Format: 4: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_4ID = 84;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 5 bytes. Byte Format: 5: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_5ID = 85;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 6 bytes. Byte Format: 6: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_6ID = 86;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 7 bytes. Byte Format: 7: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_7ID = 87;
/**
* Written to DRF. The OplogEntryId is relative to the previous del_entry OplogEntryId. The signed
* difference is encoded in 8 bytes. Byte Format: 8: OplogEntryId 1: EndOfRecord
*
* @since GemFire prPersistSprint1
*/
private static final byte OPLOG_DEL_ENTRY_8ID = 88;
/**
* The maximum size of a DEL_ENTRY record in bytes. Currenty this is 10; 1 for opcode and 8 for
* oplogEntryId and 1 for END_OF_RECORD_ID
*/
private static final int MAX_DELETE_ENTRY_RECORD_BYTES = 1 + 8 + 1;
/**
* Written to beginning of each CRF. Contains the RVV for all regions in the CRF. Byte Format 8:
* number of regions (variable length encoded number) for each region 4: number of members
* (variable length encoded number) for each member 4: canonical member id (variable length
* encoded number) 8: version id (variable length encoded number) 4: number of exceptions
* (variable length encoded number) variable: exceptions
*/
private static final byte OPLOG_RVV = 89;
/**
* When detected conflict, besides persisting the golden copy by modify(), also persist the
* conflict operation's region version and member id. and failedWritten to beginning of each CRF.
* Contains the RVV for all regions in the CRF. Byte Format regionId versions
*/
private static final byte OPLOG_CONFLICT_VERSION = 90;
/**
* persist Gemfire version string into crf, drf, krf Byte Format variable gemfire version string,
* such as 7.0.0.beta EndOfRecord
*/
private static final byte OPLOG_GEMFIRE_VERSION = 91;
static final int OPLOG_GEMFIRE_VERSION_REC_SIZE = 1 + 3 + 1;
/**
* Persist oplog file magic number. Written once at the beginning of every oplog file; CRF, DRF,
* KRF, IF and IRF. Followed by 6 byte magic number. Each oplog type has a different magic number
* Followed by EndOfRecord Fix for bug 43824
*
* @since GemFire 8.0
*/
static final byte OPLOG_MAGIC_SEQ_ID = 92;
public static final int OPLOG_MAGIC_SEQ_REC_SIZE = 1 + OPLOG_TYPE.getLen() + 1;
/** Compact this oplogs or no. A client configurable property * */
private final boolean compactOplogs;
/**
* This object is used to correctly identify the OpLog size so as to cause a switch of oplogs
*/
final Object lock = new Object();
final ByteBuffer[] bbArray = new ByteBuffer[2];
private boolean lockedForKRFcreate = false;
/**
* Set to true when this oplog will no longer be written to. Never set to false once it becomes
* true.
*/
private boolean doneAppending = false;
/**
* Creates new {@code Oplog} for the given region.
*
* @param oplogId int identifying the new oplog
* @param dirHolder The directory in which to create new Oplog
*
* @throws DiskAccessException if the disk files can not be initialized
*/
Oplog(long oplogId, PersistentOplogSet parent, DirectoryHolder dirHolder) {
if (oplogId > DiskId.MAX_OPLOG_ID) {
throw new IllegalStateException(
"Too many oplogs. The oplog id can not exceed " + DiskId.MAX_OPLOG_ID);
}
this.oplogId = oplogId;
this.oplogSet = parent;
this.parent = parent.getParent();
this.dirHolder = dirHolder;
// Pretend we have already seen the first record.
// This will cause a large initial record to force a switch
// which allows the maxDirSize to be checked.
this.firstRecord = false;
this.opState = new OpState();
long maxOplogSizeParam = getParent().getMaxOplogSizeInBytes();
long availableSpace = this.dirHolder.getAvailableSpace();
if (availableSpace < maxOplogSizeParam) {
if (DiskStoreImpl.PREALLOCATE_OPLOGS && !DiskStoreImpl.SET_IGNORE_PREALLOCATE) {
throw new DiskAccessException(
String.format("Could not create and pre grow file in dir %s with size=%s",
this.dirHolder,
maxOplogSizeParam),
new IOException("not enough space left to create and pre grow oplog files, available="
+ availableSpace + ", required=" + maxOplogSizeParam),
getParent());
}
this.maxOplogSize = availableSpace;
if (logger.isDebugEnabled()) {
logger.debug(
"Reducing maxOplogSize to {} because that is all the room remaining in the directory.",
availableSpace);
}
} else {
this.maxOplogSize = maxOplogSizeParam;
}
setMaxCrfDrfSize();
this.stats = getParent().getStats();
this.compactOplogs = getParent().getAutoCompact();
this.closed = false;
String n = getParent().getName();
this.diskFile = new File(this.dirHolder.getDir(), oplogSet.getPrefix() + n + "_" + oplogId);
try {
createDrf(null);
createCrf(null);
// open krf for offline compaction
if (getParent().isOfflineCompacting()) {
krfFileCreate();
}
} catch (Exception ex) {
close();
getParent().getCancelCriterion().checkCancelInProgress(ex);
if (ex instanceof DiskAccessException) {
throw (DiskAccessException) ex;
}
throw new DiskAccessException(
String.format("Failed creating operation log because: %s", ex),
getParent());
}
}
/**
* A copy constructor used for creating a new oplog based on the previous Oplog. This constructor
* is invoked only from the function switchOplog
*
* @param oplogId integer identifying the new oplog
* @param dirHolder The directory in which to create new Oplog
* @param prevOplog The previous oplog
*/
private Oplog(long oplogId, DirectoryHolder dirHolder, Oplog prevOplog) {
if (oplogId > DiskId.MAX_OPLOG_ID) {
throw new IllegalStateException(
"Too many oplogs. The oplog id can not exceed " + DiskId.MAX_OPLOG_ID);
}
this.oplogId = oplogId;
this.parent = prevOplog.parent;
this.oplogSet = prevOplog.oplogSet;
this.dirHolder = dirHolder;
this.opState = new OpState();
long maxOplogSizeParam = getParent().getMaxOplogSizeInBytes();
long availableSpace = this.dirHolder.getAvailableSpace();
if (prevOplog.compactOplogs) {
this.maxOplogSize = maxOplogSizeParam;
} else {
if (availableSpace < maxOplogSizeParam) {
this.maxOplogSize = availableSpace;
if (logger.isDebugEnabled()) {
logger.debug(
"Reducing maxOplogSize to {} because that is all the room remaining in the directory.",
availableSpace);
}
} else {
this.maxOplogSize = maxOplogSizeParam;
}
}
setMaxCrfDrfSize();
this.stats = prevOplog.stats;
this.compactOplogs = prevOplog.compactOplogs;
// copy over the previous Oplog's data version since data is not being
// transformed at this point
this.dataVersion = prevOplog.getDataVersionIfOld();
this.closed = false;
String n = getParent().getName();
this.diskFile = new File(this.dirHolder.getDir(), oplogSet.getPrefix() + n + "_" + oplogId);
try {
createDrf(prevOplog.drf);
createCrf(prevOplog.crf);
// open krf for offline compaction
if (getParent().isOfflineCompacting()) {
krfFileCreate();
}
} catch (Exception ex) {
close();
getParent().getCancelCriterion().checkCancelInProgress(ex);
if (ex instanceof DiskAccessException) {
throw (DiskAccessException) ex;
}
throw new DiskAccessException(
String.format("Failed creating operation log because: %s", ex),
getParent());
}
}
public Object getLock() {
return lock;
}
public void replaceIncompatibleEntry(DiskRegionView dr, DiskEntry old, DiskEntry repl) {
boolean useNextOplog = false;
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock) {
if (getOplogSet().getChild() != this) {
// make sure to only call replaceIncompatibleEntry for child, because
// this.lock
// can only sync with compaction thread on child oplog
useNextOplog = true;
} else {
// This method is use in recovery only and will not be called by
// compaction.
// It's only called before or after compaction. It will replace
// DiskEntry
// in DiskRegion without modifying DiskId (such as to a new oplogId),
// Not to change the entry count in oplog either. While doing that,
// this.lock will lock the current child to sync with compaction thread.
// If replace thread got this.lock, DiskEntry "old" will not be removed
// from
// current oplog (maybe not child). If compaction thread got this.lock,
// DiskEntry "old" should have been moved to child oplog when replace
// thread
// processes it.
// See #48032. A new region entry has been put into the region map, but
// we
// also have to replace it in the oplog live entries that are used to
// write
// the krf. If we don't, we will recover the wrong (old) value.
getOrCreateDRI(dr).replaceLive(old, repl);
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"replacing incompatible entry key = {} old = {} new = {} oldDiskId = {} new diskId = {} tag = {} in child oplog #{}",
old.getKey(), System.identityHashCode(old), System.identityHashCode(repl),
old.getDiskId(), repl.getDiskId(), old.getVersionStamp(), this.getOplogId());
}
}
}
if (useNextOplog) {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSwitchingOplog();
}
Assert.assertTrue(getOplogSet().getChild() != this);
getOplogSet().getChild().replaceIncompatibleEntry(dr, old, repl);
}
}
private void writeDiskStoreRecord(OplogFile olf, OPLOG_TYPE type) throws IOException {
this.opState = new OpState();
this.opState.initialize(type);
writeOpLogBytes(olf, false, true);
olf.currSize += getOpStateSize();
this.dirHolder.incrementTotalOplogSize(getOpStateSize());
clearOpState();
this.opState.initialize(getParent().getDiskStoreID());
writeOpLogBytes(olf, false, true); // fix for bug 41928
olf.currSize += getOpStateSize();
this.dirHolder.incrementTotalOplogSize(getOpStateSize());
}
private void writeGemfireVersionRecord(OplogFile olf) throws IOException {
if (this.gfversion == null) {
this.gfversion = KnownVersion.CURRENT;
}
KnownVersion dataVersion = getDataVersionIfOld();
if (dataVersion == null) {
dataVersion = KnownVersion.CURRENT;
}
// if gfversion and dataVersion are not same, then write a special token
// version and then write both, else write gfversion as before
// this is for backward compatibility with 7.0
this.opState = new OpState();
if (this.gfversion == dataVersion) {
writeProductVersionRecord(this.gfversion, olf);
} else {
writeProductVersionRecord(KnownVersion.TOKEN, olf);
clearOpState();
writeProductVersionRecord(this.gfversion, olf);
clearOpState();
writeProductVersionRecord(dataVersion, olf);
}
}
private void writeProductVersionRecord(KnownVersion version, OplogFile olf) throws IOException {
this.opState.initialize(version.ordinal());
writeOpLogBytes(olf, false, true);
olf.currSize += getOpStateSize();
this.dirHolder.incrementTotalOplogSize(getOpStateSize());
}
public KnownVersion currentRecoveredGFVersion() {
return this.gfversion;
}
/**
* Write an RVV record containing all of the live disk regions.
*/
private void writeRVVRecord(OplogFile olf, boolean writeGCRVV) throws IOException {
writeRVVRecord(olf, getParent().getAllDiskRegions(), writeGCRVV);
}
/**
* Write the RVV record for the given regions.
*
* @param olf the oplog to write to
* @param diskRegions the set of disk regions we should write the RVV of
* @param writeGCRVV true to write write the GC RVV
*/
private void writeRVVRecord(OplogFile olf, Map<Long, AbstractDiskRegion> diskRegions,
boolean writeGCRVV) throws IOException {
this.opState = new OpState();
this.opState.initialize(diskRegions, writeGCRVV);
writeOpLogBytes(olf, false, true); // fix for bug 41928
olf.currSize += getOpStateSize();
this.dirHolder.incrementTotalOplogSize(getOpStateSize());
}
private boolean wroteNewEntryBase = false;
/**
* Write a OPLOG_NEW_ENTRY_BASE_ID to this oplog. Must be called before any OPLOG_NEW_ENTRY_0ID
* records are written to this oplog.
*/
private boolean writeNewEntryBaseRecord(boolean async) throws IOException {
if (this.wroteNewEntryBase)
return false;
this.wroteNewEntryBase = true;
long newEntryBase = getOplogSet().getOplogEntryId();
OpState saved = this.opState;
try {
this.opState = new OpState();
this.opState.initialize(newEntryBase);
writeOpLogBytes(this.crf, async, false/* no need to flush this record */);
this.dirHolder.incrementTotalOplogSize(getOpStateSize());
} finally {
this.opState = saved;
}
return true;
}
/**
* Return true if this oplog has a drf but does not have a crf
*/
boolean isDrfOnly() {
return this.drf.f != null && this.crf.f == null;
}
/**
* This constructor will get invoked only in case of persistent region when it is recovering an
* oplog.
*/
Oplog(long oplogId, PersistentOplogSet parent) {
// @todo have the crf and drf use different directories.
if (oplogId > DiskId.MAX_OPLOG_ID) {
throw new IllegalStateException(
"Too many oplogs. The oplog id can not exceed " + DiskId.MAX_OPLOG_ID);
}
this.isRecovering = true;
this.oplogId = oplogId;
this.parent = parent.getParent();
this.oplogSet = parent;
this.opState = new OpState();
this.maxOplogSize = getParent().getMaxOplogSizeInBytes();
setMaxCrfDrfSize();
this.stats = getParent().getStats();
this.compactOplogs = getParent().getAutoCompact();
this.closed = true;
this.crf.RAFClosed = true;
this.deleted.set(true);
this.haveRecoveredCrf = false;
this.haveRecoveredDrf = false;
}
/**
* Returns true if added file was crf; false if drf
*/
boolean addRecoveredFile(File f, DirectoryHolder dh) {
String fname = f.getName();
if (this.dirHolder != null) {
if (!dh.equals(this.dirHolder)) {
throw new DiskAccessException(
"Oplog#" + getOplogId() + " has files in two different directories: \"" + this.dirHolder
+ "\", and \"" + dh
+ "\". Both the crf and drf for this oplog should be in the same directory.",
getParent());
}
} else {
this.dirHolder = dh;
}
if (fname.endsWith(Oplog.CRF_FILE_EXT)) {
this.crf.f = f;
return true;
} else if (fname.endsWith(Oplog.DRF_FILE_EXT)) {
this.drf.f = f;
// } else if (fname.endsWith(Oplog.KRF_FILE_EXT)) {
// this.krf.f = f;
} else {
assert false : fname;
}
return false;
}
void setRecoveredDrfSize(long size) {
this.drf.currSize += size;
this.drf.bytesFlushed += size;
}
void setRecoveredCrfSize(long size) {
this.crf.currSize += size;
this.crf.bytesFlushed += size;
}
private boolean isRecovering;
boolean isRecovering() {
return this.isRecovering;
}
private DiskStoreImpl getParent() {
return this.parent;
}
private PersistentOplogSet getOplogSet() {
return oplogSet;
}
void initAfterRecovery(boolean offline) {
this.isRecovering = false;
this.closed = false;
this.deleted.set(false);
String n = getParent().getName();
// crf might not exist; but drf always will
this.diskFile =
new File(this.drf.f.getParentFile(), oplogSet.getPrefix() + n + "_" + this.oplogId);
try {
// This is a recovered oplog and we only read from its crf.
// No need to open the drf.
this.doneAppending = true;
if (this.crf.f != null && !hasNoLiveValues()) {
this.closed = false;
// truncate crf/drf if their actual size is less than their pre-blow
// size
this.crf.raf = new UninterruptibleRandomAccessFile(this.crf.f, "rw");
this.crf.RAFClosed = false;
this.crf.channel = this.crf.raf.getChannel();
unpreblow(this.crf, getMaxCrfSize());
this.crf.raf.close();
// make crf read only
this.crf.raf = new UninterruptibleRandomAccessFile(this.crf.f, "r");
this.crf.channel = this.crf.raf.getChannel();
this.stats.incOpenOplogs();
// drf.raf is null at this point. create one and close it to retain
// existing behavior
try {
this.drf.raf = new UninterruptibleRandomAccessFile(this.drf.f, "rw");
this.drf.RAFClosed = false;
this.drf.channel = this.drf.raf.getChannel();
unpreblow(this.drf, getMaxDrfSize());
} finally {
this.drf.raf.close();
this.drf.raf = null;
this.drf.RAFClosed = true;
}
// no need to seek to the end; we will not be writing to a recovered
// oplog; only reading
// this.crf.raf.seek(this.crf.currSize);
} else if (!offline) {
// drf exists but crf has been deleted (because it was empty).
// I don't think the drf needs to be opened. It is only used during recovery.
// At some point the compacter may identify that it can be deleted.
this.crf.RAFClosed = true;
deleteCRF();
this.closed = true;
this.deleted.set(true);
}
this.drf.RAFClosed = true; // since we never open it on a recovered oplog
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(
String.format("Failed creating operation log because: %s", ex),
getParent());
}
if (hasNoLiveValues() && !offline) {
getOplogSet().removeOplog(getOplogId(), true, getHasDeletes() ? this : null);
if (!getHasDeletes()) {
getOplogSet().drfDelete(this.oplogId);
deleteFile(this.drf);
}
} else if (needsCompaction()) {
// just leave it in the list it is already in
} else {
// remove it from the compactable list
getOplogSet().removeOplog(getOplogId(),
true/*
* say we are deleting so that undeletedOplogSize is not inced
*/, null);
// add it to the inactive list
getOplogSet().addInactive(this);
}
}
boolean getHasDeletes() {
return this.hasDeletes.get();
}
private void setHasDeletes(boolean v) {
this.hasDeletes.set(v);
}
private void closeAndDeleteAfterEx(IOException ex, OplogFile olf) {
if (olf == null) {
return;
}
if (olf.raf != null) {
try {
olf.raf.close();
} catch (IOException e) {
logger.warn(String.format("Failed to close file %s", olf.f.getAbsolutePath()),
e);
}
}
olf.RAFClosed = true;
if (!olf.f.delete() && olf.f.exists()) {
throw new DiskAccessException(
String.format("Could not delete %s.", olf.f.getAbsolutePath()),
ex, getParent());
}
}
private void preblow(OplogFile olf, long maxSize) throws IOException {
long availableSpace = this.dirHolder.getAvailableSpace();
if (availableSpace >= maxSize) {
try {
NativeCalls.getInstance().preBlow(olf.f.getAbsolutePath(), maxSize,
(DiskStoreImpl.PREALLOCATE_OPLOGS && !DiskStoreImpl.SET_IGNORE_PREALLOCATE));
} catch (IOException ioe) {
if (logger.isDebugEnabled()) {
logger.debug("Could not pregrow oplog to {} because: {}", maxSize, ioe.getMessage(), ioe);
}
// I don't think I need any of this. If setLength throws then
// the file is still ok.
// I need this on windows. I'm seeing this in testPreblowErrorCondition:
// Caused by: java.io.IOException: The parameter is incorrect
// at sun.nio.ch.FileDispatcher.write0(Native Method)
// at sun.nio.ch.FileDispatcher.write(FileDispatcher.java:44)
// at sun.nio.ch.IOUtil.writeFromNativeBuffer(IOUtil.java:104)
// at sun.nio.ch.IOUtil.write(IOUtil.java:60)
// at sun.nio.ch.FileChannelImpl.write(FileChannelImpl.java:206)
// at org.apache.geode.internal.cache.Oplog.flush(Oplog.java:3377)
// at
// org.apache.geode.internal.cache.Oplog.flushAll(Oplog.java:3419)
/*
* { String os = System.getProperty("os.name"); if (os != null) { if (os.indexOf("Windows")
* != -1) { olf.raf.close(); olf.RAFClosed = true; if (!olf.f.delete() && olf.f.exists()) {
* throw new DiskAccessException
* (String.format("Could not delete %s.",.toLocalizedString (olf.f.getAbsolutePath()),
* getParent()); } olf.raf = new RandomAccessFile(olf.f, SYNC_WRITES ? "rwd" : "rw");
* olf.RAFClosed = false; } } }
*/
closeAndDeleteAfterEx(ioe, olf);
throw new DiskAccessException(String.format("Could not pre-allocate file %s with size=%s",
olf.f.getAbsolutePath(), maxSize), ioe, getParent());
}
}
// TODO: Perhaps the test flag is not requierd here. Will re-visit.
else if (DiskStoreImpl.PREALLOCATE_OPLOGS && !DiskStoreImpl.SET_IGNORE_PREALLOCATE) {
throw new DiskAccessException(
String.format("Could not pre-allocate file %s with size=%s", olf.f.getAbsolutePath(),
maxSize),
new IOException("not enough space left to pre-blow, available=" + availableSpace
+ ", required=" + maxSize),
getParent());
}
}
private void unpreblow(OplogFile olf, long maxSize) {
synchronized (/* olf */this.lock) {
if (!olf.RAFClosed && !olf.unpreblown) {
olf.unpreblown = true;
if (olf.currSize < maxSize) {
try {
olf.raf.setLength(olf.currSize);
} catch (IOException ignore) {
}
}
}
}
}
/**
* Creates the crf oplog file
*/
private void createCrf(OplogFile prevOlf) throws IOException {
File f = new File(this.diskFile.getPath() + CRF_FILE_EXT);
if (logger.isDebugEnabled()) {
logger.debug("Creating operation log file {}", f);
}
this.crf.f = f;
preblow(this.crf, getMaxCrfSize());
this.crf.raf = new UninterruptibleRandomAccessFile(f, SYNC_WRITES ? "rwd" : "rw");
this.crf.RAFClosed = false;
oplogSet.crfCreate(this.oplogId);
this.crf.writeBuf = allocateWriteBuf(prevOlf);
logger.info("Created {} {} for disk store {}.",
new Object[] {toString(), getFileType(this.crf), getParent().getName()});
this.crf.channel = this.crf.raf.getChannel();
this.stats.incOpenOplogs();
writeDiskStoreRecord(this.crf, OPLOG_TYPE.CRF);
writeGemfireVersionRecord(this.crf);
writeRVVRecord(this.crf, false);
// Fix for bug 41654 - don't count the header
// size against the size of the oplog. This ensures that
// even if we have a large RVV, we can still write up to
// max-oplog-size bytes to this oplog.
this.maxCrfSize += this.crf.currSize;
}
private static ByteBuffer allocateWriteBuf(OplogFile prevOlf) {
if (prevOlf != null && prevOlf.writeBuf != null) {
ByteBuffer result = prevOlf.writeBuf;
prevOlf.writeBuf = null;
return result;
} else {
return ByteBuffer.allocateDirect(Integer.getInteger("WRITE_BUF_SIZE", 32768));
}
}
/**
* Creates the drf oplog file
*/
private void createDrf(OplogFile prevOlf) throws IOException {
File f = new File(this.diskFile.getPath() + DRF_FILE_EXT);
this.drf.f = f;
if (logger.isDebugEnabled()) {
logger.debug("Creating operation log file {}", f);
}
preblow(this.drf, getMaxDrfSize());
this.drf.raf = new UninterruptibleRandomAccessFile(f, SYNC_WRITES ? "rwd" : "rw");
this.drf.RAFClosed = false;
this.oplogSet.drfCreate(this.oplogId);
this.drf.writeBuf = allocateWriteBuf(prevOlf);
logger.info("Created {} {} for disk store {}.",
new Object[] {toString(), getFileType(this.drf), getParent().getName()});
this.drf.channel = this.drf.raf.getChannel();
writeDiskStoreRecord(this.drf, OPLOG_TYPE.DRF);
writeGemfireVersionRecord(this.drf);
writeRVVRecord(this.drf, true);
}
/**
* Returns the {@code DiskStoreStats} for this oplog
*/
public DiskStoreStats getStats() {
return this.stats;
}
/**
* Test Method to be used only for testing purposes. Gets the underlying File object for the Oplog
* . Oplog class uses this File object to obtain the RandomAccessFile object. Before returning the
* File object , the dat present in the buffers of the RandomAccessFile object is flushed.
* Otherwise, for windows the actual file length does not match with the File size obtained from
* the File object
*/
File getOplogFileForTest() throws IOException {
// @todo check callers for drf
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* crf */) {
if (!this.crf.RAFClosed) {
this.crf.raf.getFD().sync();
}
return this.crf.f;
}
}
public File getCrfFile() {
return this.crf.f;
}
public File getDrfFile() {
return this.drf.f;
}
/**
* Given a set of Oplog file names return a Set of the oplog files that match those names that are
* managed by this Oplog.
*
* @param oplogFileNames a Set of operation log file names.
*/
public Set<String> gatherMatchingOplogFiles(Set<String> oplogFileNames) {
Set<String> matchingFiles = new HashSet<>();
// CRF match?
if ((null != this.crf.f) && this.crf.f.exists()
&& oplogFileNames.contains(this.crf.f.getName())) {
matchingFiles.add(this.crf.f.getName());
}
// DRF match?
if ((null != this.drf.f) && this.drf.f.exists()
&& oplogFileNames.contains(this.drf.f.getName())) {
matchingFiles.add(this.drf.f.getName());
}
// KRF match?
if (getParent().getDiskInitFile().hasKrf(this.oplogId)) {
File krfFile = getKrfFile();
if (krfFile.exists() && oplogFileNames.contains(krfFile.getName())) {
matchingFiles.add(krfFile.getName());
}
}
return matchingFiles;
}
public boolean hasKrf() {
return getParent().getDiskInitFile().hasKrf(this.oplogId);
}
/** the oplog identifier * */
public long getOplogId() {
return this.oplogId;
}
/**
* Returns the unserialized bytes and bits for the given Entry. If Oplog is destroyed while
* querying, then the DiskRegion is queried again to obatin the value This method should never get
* invoked for an entry which has been destroyed
*
* @since GemFire 3.2.1
* @param id The DiskId for the entry @param offset The offset in this OpLog where the entry is
* present. @param faultingIn @param bitOnly boolean indicating whether to extract just the
* UserBit or UserBit with value @return BytesAndBits object wrapping the value & user bit
*/
@Override
public BytesAndBits getBytesAndBits(DiskRegionView dr, DiskId id, boolean faultIn,
boolean bitOnly) {
Oplog retryOplog = null;
long offset = 0;
synchronized (id) {
long opId = id.getOplogId();
if (opId != getOplogId()) {
// the oplog changed on us so we need to do a recursive
// call after unsyncing
retryOplog = getOplogSet().getChild(opId);
} else {
// fetch this while synced so it will be consistent with oplogId
offset = id.getOffsetInOplog();
}
}
if (retryOplog != null) {
return retryOplog.getBytesAndBits(dr, id, faultIn, bitOnly);
}
BytesAndBits bb = null;
long start = this.stats.startRead();
// If the offset happens to be -1, still it is possible that
// the data is present in the current oplog file.
if (offset == -1) {
// Since it is given that a get operation has alreadty
// taken a
// lock on an entry , no put operation could have modified the
// oplog ID
// there fore synchronization is not needed
offset = id.getOffsetInOplog();
}
// If the current OpLog is not destroyed ( its opLogRaf file
// is still open) we can retrieve the value from this oplog.
try {
bb = basicGet(dr, offset, bitOnly, id.getValueLength(), id.getUserBits());
} catch (DiskAccessException dae) {
logger.error("Oplog::basicGet: Error in reading the data from disk for Disk ID " +
id,
dae);
throw dae;
}
if (bb == null) {
throw new EntryDestroyedException(
String.format(
"No value was found for entry with disk Id %s on a region with synchronous writing set to %s",
new Object[] {id, dr.isSync()}));
}
if (bitOnly) {
dr.endRead(start, this.stats.endRead(start, 1), 1);
} else {
dr.endRead(start, this.stats.endRead(start, bb.getBytes().length), bb.getBytes().length);
}
return bb;
}
/**
* Returns the object stored on disk with the given id. This method is used for testing purposes
* only. As such, it bypasses the buffer and goes directly to the disk. This is not a thread safe
* function , in the sense, it is possible that by the time the OpLog is queried , data might move
* HTree with the oplog being destroyed
*
* @param id A DiskId object for which the value on disk will be fetched
*/
@Override
public BytesAndBits getNoBuffer(DiskRegion dr, DiskId id) {
if (logger.isDebugEnabled()) {
logger.debug("Oplog::getNoBuffer:Before invoking Oplog.basicGet for DiskID ={}", id);
}
try {
return basicGet(dr, id.getOffsetInOplog(), false, id.getValueLength(), id.getUserBits());
} catch (DiskAccessException dae) {
logger.error("Oplog::getNoBuffer:Exception in retrieving value from disk for diskId=" +
id,
dae);
throw dae;
} catch (IllegalStateException ise) {
logger.error("Oplog::getNoBuffer:Exception in retrieving value from disk for diskId=" +
id,
ise);
throw ise;
}
}
void close(DiskRegion dr) {
// while a krf is being created can not close a region
lockCompactor();
try {
addUnrecoveredRegion(dr.getId());
DiskRegionInfo dri = getDRI(dr);
if (dri != null) {
long clearCount = dri.clear(null);
if (clearCount != 0) {
this.totalLiveCount.addAndGet(-clearCount);
// no need to call handleNoLiveValues because we now have an
// unrecovered region.
}
this.regionMap.remove(dr.getId(), dri);
}
} finally {
unlockCompactor();
}
}
void clear(DiskRegion dr, RegionVersionVector rvv) {
DiskRegionInfo dri = getDRI(dr);
if (dri != null) {
long clearCount = dri.clear(rvv);
if (clearCount != 0) {
this.totalLiveCount.addAndGet(-clearCount);
if (!isCompacting() || calledByCompactorThread()) {
handleNoLiveValues();
}
}
}
}
void destroy(DiskRegion dr) {
DiskRegionInfo dri = getDRI(dr);
if (dri != null) {
long clearCount = dri.clear(null);
if (clearCount != 0) {
this.totalLiveCount.addAndGet(-clearCount);
if (!isCompacting() || calledByCompactorThread()) {
handleNoLiveValues();
}
}
this.regionMap.remove(dr.getId(), dri);
}
}
long getMaxRecoveredOplogEntryId() {
long result = this.recoverNewEntryId;
if (this.recoverModEntryIdHWM > result) {
result = this.recoverModEntryIdHWM;
}
if (this.recoverDelEntryIdHWM > result) {
result = this.recoverDelEntryIdHWM;
}
return result;
}
/**
* Used during recovery to calculate the OplogEntryId of the next NEW_ENTRY record.
*
* @since GemFire prPersistSprint1
*/
private long recoverNewEntryId = DiskStoreImpl.INVALID_ID;
/**
* Used during writing to remember the last MOD_ENTRY OplogEntryId written to this oplog.
*
* @since GemFire prPersistSprint1
*/
private long writeModEntryId = DiskStoreImpl.INVALID_ID;
/**
* Used during recovery to calculate the OplogEntryId of the next MOD_ENTRY record.
*
* @since GemFire prPersistSprint1
*/
private long recoverModEntryId = DiskStoreImpl.INVALID_ID;
/**
* Added to fix bug 41301. High water mark of modified entries.
*/
private long recoverModEntryIdHWM = DiskStoreImpl.INVALID_ID;
/**
* Added to fix bug 41340. High water mark of deleted entries.
*/
private long recoverDelEntryIdHWM = DiskStoreImpl.INVALID_ID;
/**
* Used during writing to remember the last DEL_ENTRY OplogEntryId written to this oplog.
*
* @since GemFire prPersistSprint1
*/
private long writeDelEntryId = DiskStoreImpl.INVALID_ID;
/**
* Used during recovery to calculate the OplogEntryId of the next DEL_ENTRY record.
*
* @since GemFire prPersistSprint1
*/
private long recoverDelEntryId = DiskStoreImpl.INVALID_ID;
private void setRecoverNewEntryId(long v) {
this.recoverNewEntryId = v;
}
private long incRecoverNewEntryId() {
this.recoverNewEntryId += 1;
return this.recoverNewEntryId;
}
/**
* Given a delta calculate the OplogEntryId for a MOD_ENTRY.
*/
public long calcModEntryId(long delta) {
long oplogKeyId = this.recoverModEntryId + delta;
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"calcModEntryId delta={} recoverModEntryId={} oplogKeyId={}", delta,
this.recoverModEntryId, oplogKeyId);
}
this.recoverModEntryId = oplogKeyId;
if (oplogKeyId > this.recoverModEntryIdHWM) {
this.recoverModEntryIdHWM = oplogKeyId; // fixes bug 41301
}
return oplogKeyId;
}
/**
* Given a delta calculate the OplogEntryId for a DEL_ENTRY.
*/
public long calcDelEntryId(long delta) {
long oplogKeyId = this.recoverDelEntryId + delta;
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"calcDelEntryId delta={} recoverModEntryId={} oplogKeyId={}", delta,
this.recoverModEntryId, oplogKeyId);
}
this.recoverDelEntryId = oplogKeyId;
if (oplogKeyId > this.recoverDelEntryIdHWM) {
this.recoverDelEntryIdHWM = oplogKeyId; // fixes bug 41340
}
return oplogKeyId;
}
private boolean crashed;
boolean isCrashed() {
return this.crashed;
}
/**
* Return bytes read.
*/
long recoverDrf(OplogEntryIdSet deletedIds, boolean alreadyRecoveredOnce, boolean latestOplog) {
File drfFile = this.drf.f;
if (drfFile == null) {
this.haveRecoveredDrf = true;
return 0L;
}
lockCompactor();
try {
if (this.haveRecoveredDrf && !getHasDeletes())
return 0L; // do this while holding lock
if (!this.haveRecoveredDrf) {
this.haveRecoveredDrf = true;
}
logger.info("Recovering {} {} for disk store {}.",
new Object[] {toString(), drfFile.getAbsolutePath(), getParent().getName()});
this.recoverDelEntryId = DiskStoreImpl.INVALID_ID;
boolean readLastRecord = true;
CountingDataInputStream dis = null;
try {
int recordCount = 0;
boolean foundDiskStoreRecord = false;
FileInputStream fis = null;
try {
fis = new FileInputStream(drfFile);
dis = new CountingDataInputStream(new BufferedInputStream(fis, 32 * 1024),
drfFile.length());
boolean endOfLog = false;
while (!endOfLog) {
if (dis.atEndOfFile()) {
endOfLog = true;
break;
}
readLastRecord = false;
byte opCode = dis.readByte();
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "drf byte={} location={}", opCode,
Long.toHexString(dis.getCount()));
}
switch (opCode) {
case OPLOG_EOF_ID:
// we are at the end of the oplog. So we need to back up one byte
dis.decrementCount();
endOfLog = true;
break;
case OPLOG_DEL_ENTRY_1ID:
case OPLOG_DEL_ENTRY_2ID:
case OPLOG_DEL_ENTRY_3ID:
case OPLOG_DEL_ENTRY_4ID:
case OPLOG_DEL_ENTRY_5ID:
case OPLOG_DEL_ENTRY_6ID:
case OPLOG_DEL_ENTRY_7ID:
case OPLOG_DEL_ENTRY_8ID:
readDelEntry(dis, opCode, deletedIds, parent);
recordCount++;
break;
case OPLOG_DISK_STORE_ID:
readDiskStoreRecord(dis, this.drf.f);
foundDiskStoreRecord = true;
recordCount++;
break;
case OPLOG_MAGIC_SEQ_ID:
readOplogMagicSeqRecord(dis, this.drf.f, OPLOG_TYPE.DRF);
break;
case OPLOG_GEMFIRE_VERSION:
readGemfireVersionRecord(dis, this.drf.f);
recordCount++;
break;
case OPLOG_RVV:
long idx = dis.getCount();
readRVVRecord(dis, this.drf.f, true, latestOplog);
recordCount++;
break;
default:
throw new DiskAccessException(
String.format("Unknown opCode %s found in disk operation log.",
opCode),
getParent());
}
readLastRecord = true;
// @todo
// if (rgn.isDestroyed()) {
// break;
// }
} // while
} finally {
if (dis != null) {
dis.close();
}
if (fis != null) {
fis.close();
}
}
if (!foundDiskStoreRecord && recordCount > 0) {
throw new DiskAccessException(
"The oplog file \"" + this.drf.f + "\" does not belong to the init file \""
+ getParent().getInitFile() + "\". Drf did not contain a disk store id.",
getParent());
}
} catch (EOFException ignore) {
// ignore since a partial record write can be caused by a crash
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(
String.format("Failed to read file during recovery from %s",
drfFile.getPath()),
ex, getParent());
} catch (CancelException e) {
if (logger.isDebugEnabled()) {
logger.debug("Oplog::readOplog:Error in recovery as Cache was closed", e);
}
} catch (RegionDestroyedException e) {
if (logger.isDebugEnabled()) {
logger.debug("Oplog::readOplog:Error in recovery as Region was destroyed", e);
}
}
// Add the Oplog size to the Directory Holder which owns this oplog,
// so that available space is correctly calculated & stats updated.
long byteCount = 0;
if (!readLastRecord) {
// this means that there was a crash
// and hence we should not continue to read
// the next oplog
this.crashed = true;
if (dis != null) {
byteCount = dis.getFileLength();
}
} else {
if (dis != null) {
byteCount = dis.getCount();
}
}
if (!alreadyRecoveredOnce) {
setRecoveredDrfSize(byteCount);
this.dirHolder.incrementTotalOplogSize(byteCount);
}
return byteCount;
} finally {
unlockCompactor();
}
}
/**
* This map is used during recovery to keep track of what entries were recovered. Its keys are the
* oplogEntryId; its values are the actual logical keys that end up in the Region's keys. It used
* to be a local variable in basicInitializeOwner but now that it needs to live longer than that
* method I made it an instance variable It is now only alive during recoverRegionsThatAreReady so
* it could once again be passed down into each oplog.
* <p>
* If offlineCompaction the value in this map will have the key bytes, values bytes, user bits,
* etc (any info we need to copy forward).
*/
private OplogEntryIdMap kvMap;
private OplogEntryIdMap getRecoveryMap() {
return this.kvMap;
}
/**
* This map is used during recover to keep track of keys that are skipped. Later modify records in
* the same oplog may use this map to retrieve the correct key.
*/
private OplogEntryIdMap skippedKeyBytes;
private boolean readKrf(OplogEntryIdSet deletedIds, boolean recoverValues,
boolean recoverValuesSync, Set<Oplog> oplogsNeedingValueRecovery, boolean latestOplog) {
File f = new File(this.diskFile.getPath() + KRF_FILE_EXT);
if (!f.exists()) {
return false;
}
if (!getParent().getDiskInitFile().hasKrf(this.oplogId)) {
logger.info("Removing incomplete krf {} for oplog {}, disk store {}",
new Object[] {f.getName(), this.oplogId, getParent().getName()});
f.delete();
}
// Set krfCreated to true since we have a krf.
this.krfCreated.set(true);
// Fix for 42741 - we do this after creating setting the krfCreated flag
// so that we don't try to recreate the krf.
if (recoverValuesSync) {
return false;
}
FileInputStream fis;
try {
fis = new FileInputStream(f);
} catch (FileNotFoundException ignore) {
return false;
}
try {
if (getParent().isOffline() && !getParent().FORCE_KRF_RECOVERY) {
return false;
}
logger.info("Recovering {} {} for disk store {}.",
new Object[] {toString(), f.getAbsolutePath(), getParent().getName()});
this.recoverNewEntryId = DiskStoreImpl.INVALID_ID;
this.recoverModEntryId = DiskStoreImpl.INVALID_ID;
this.recoverModEntryIdHWM = DiskStoreImpl.INVALID_ID;
long oplogKeyIdHWM = DiskStoreImpl.INVALID_ID;
int krfEntryCount = 0;
DataInputStream dis = new DataInputStream(new BufferedInputStream(fis, 1024 * 1024));
final KnownVersion version = getProductVersionIfOld();
final ByteArrayDataInput in = new ByteArrayDataInput();
try {
try {
validateOpcode(dis, OPLOG_MAGIC_SEQ_ID);
readOplogMagicSeqRecord(dis, f, OPLOG_TYPE.KRF);
validateOpcode(dis, OPLOG_DISK_STORE_ID);
readDiskStoreRecord(dis, f);
} catch (DiskAccessException ignore) {
// Failed to read the file. There are two possibilities. Either this
// file is in old format which does not have a magic seq in the
// beginning or this is not a valid file at all. Try reading it as a
// file in old format
fis.close();
fis = new FileInputStream(f);
dis = new DataInputStream(new BufferedInputStream(fis, 1024 * 1024));
readDiskStoreRecord(dis, f);
} catch (IllegalStateException ignore) {
// Failed to read the file. There are two possibilities. Either this
// is in new format which has a magic seq in the beginning or this is
// not a valid file at all
fis.close();
fis = new FileInputStream(f);
dis = new DataInputStream(new BufferedInputStream(fis, 1024 * 1024));
readDiskStoreRecord(dis, f);
}
readGemfireVersionRecord(dis, f);
readTotalCountRecord(dis, f);
readRVVRecord(dis, f, false, latestOplog);
long lastOffset = 0;
byte[] keyBytes = DataSerializer.readByteArray(dis);
while (keyBytes != null) {
byte userBits = dis.readByte();
int valueLength = InternalDataSerializer.readArrayLength(dis);
byte[] valueBytes = null;
long drId = DiskInitFile.readDiskRegionID(dis);
DiskRecoveryStore drs = getOplogSet().getCurrentlyRecovering(drId);
// read version
VersionTag tag = null;
if (EntryBits.isWithVersions(userBits)) {
tag = readVersionsFromOplog(dis);
if (drs != null && !drs.getDiskRegionView().getFlags()
.contains(DiskRegionFlag.IS_WITH_VERSIONING)) {
// 50044 Remove version tag from entry if we don't want versioning
// for this region
tag = null;
userBits = EntryBits.setWithVersions(userBits, false);
} else {
// Update the RVV with the new entry
if (drs != null) {
drs.recordRecoveredVersionTag(tag);
}
}
}
long oplogKeyId = InternalDataSerializer.readVLOld(dis);
long oplogOffset;
if (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits)) {
oplogOffset = -1;
} else {
oplogOffset = lastOffset + InternalDataSerializer.readVLOld(dis);
lastOffset = oplogOffset;
}
if (oplogKeyId > oplogKeyIdHWM) {
oplogKeyIdHWM = oplogKeyId;
}
if (okToSkipModifyRecord(deletedIds, drId, drs, oplogKeyId, true, tag).skip()) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readNewEntry skipping oplogKeyId=<{}> drId={} userBits={} oplogOffset={} valueLen={}",
oplogKeyId, drId, userBits, oplogOffset, valueLength);
}
this.stats.incRecoveryRecordsSkipped();
incSkipped();
} else {
if (EntryBits.isAnyInvalid(userBits)) {
if (EntryBits.isInvalid(userBits)) {
valueBytes = DiskEntry.INVALID_BYTES;
} else {
valueBytes = DiskEntry.LOCAL_INVALID_BYTES;
}
} else if (EntryBits.isTombstone(userBits)) {
valueBytes = DiskEntry.TOMBSTONE_BYTES;
}
Object key = deserializeKey(keyBytes, version, in);
{
Object oldValue = getRecoveryMap().put(oplogKeyId, key);
if (oldValue != null) {
throw new AssertionError(
String.format(
"Oplog::readNewEntry: Create is present in more than one Oplog. This should not be possible. The Oplog Key ID for this entry is %s.",
oplogKeyId));
}
}
if (drs != null) {
DiskEntry de = drs.getDiskEntry(key);
if (de == null) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readNewEntry oplogKeyId=<{}> drId={} userBits={} oplogOffset={} valueLen={}",
oplogKeyId, drId, userBits, oplogOffset, valueLength);
}
DiskEntry.RecoveredEntry re =
createRecoveredEntry(valueBytes, valueLength, userBits,
getOplogId(), oplogOffset, oplogKeyId, false, version, in);
if (tag != null) {
re.setVersionTag(tag);
}
initRecoveredEntry(drs.getDiskRegionView(), drs.initializeRecoveredEntry(key, re));
drs.getDiskRegionView().incRecoveredEntryCount();
this.stats.incRecoveredEntryCreates();
krfEntryCount++;
} else {
DiskId curdid = de.getDiskId();
// assert curdid.getOplogId() != getOplogId();
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"ignore readNewEntry because getOplogId()={} != curdid.getOplogId()={} for drId={} key={}",
getOplogId(), curdid.getOplogId(), drId, key);
}
}
}
}
keyBytes = DataSerializer.readByteArray(dis);
} // while
setRecoverNewEntryId(oplogKeyIdHWM);
} catch (IOException ex) {
try {
fis.close();
fis = null;
} catch (IOException ignore) {
}
throw new DiskAccessException("Unable to recover from krf file for oplogId=" + oplogId
+ ", file=" + f.getName() + ". This file is corrupt, but may be safely deleted.", ex,
getParent());
}
if (recoverValues && krfEntryCount > 0) {
oplogsNeedingValueRecovery.add(this);
// TODO optimize this code and make it async
// It should also honor the lru limit
// The fault in logic might not work until
// the region is actually created.
// Instead of reading the crf it might be better to iterate the live
// entry
// list that was built during KRF recovery. Just fault values in until
// we
// hit the LRU limit (if we have one). Only fault in values for entries
// recovered from disk that are still in this oplog.
// Defer faulting in values until all oplogs for the ds have been
// recovered.
}
} finally {
// fix for bug 42776
if (fis != null) {
try {
fis.close();
fis = null;
} catch (IOException ignore) {
}
}
}
return true;
}
private void validateOpcode(DataInputStream dis, byte expect) throws IOException {
byte opCode = dis.readByte();
if (opCode != expect) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "expected opcode id absent: {}", expect);
}
throw new IllegalStateException();
}
}
/**
* Return number of bytes read
*/
private long readCrf(OplogEntryIdSet deletedIds, boolean recoverValues, boolean latestOplog) {
this.recoverNewEntryId = DiskStoreImpl.INVALID_ID;
this.recoverModEntryId = DiskStoreImpl.INVALID_ID;
this.recoverModEntryIdHWM = DiskStoreImpl.INVALID_ID;
boolean readLastRecord = true;
CountingDataInputStream dis = null;
try {
final LocalRegion currentRegion = LocalRegion.getInitializingRegion();
final KnownVersion version = getProductVersionIfOld();
final ByteArrayDataInput in = new ByteArrayDataInput();
final HeapDataOutputStream hdos = new HeapDataOutputStream(KnownVersion.CURRENT);
int recordCount = 0;
boolean foundDiskStoreRecord = false;
FileInputStream fis = null;
try {
fis = new FileInputStream(this.crf.f);
dis = new CountingDataInputStream(new BufferedInputStream(fis, 1024 * 1024),
this.crf.f.length());
boolean endOfLog = false;
while (!endOfLog) {
if (dis.atEndOfFile()) {
endOfLog = true;
break;
}
readLastRecord = false;
byte opCode = dis.readByte();
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "Oplog opCode={}", opCode);
}
switch (opCode) {
case OPLOG_EOF_ID:
// we are at the end of the oplog. So we need to back up one byte
dis.decrementCount();
endOfLog = true;
break;
case OPLOG_CONFLICT_VERSION:
this.readVersionTagOnlyEntry(dis, opCode);
break;
case OPLOG_NEW_ENTRY_BASE_ID: {
long newEntryBase = dis.readLong();
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "newEntryBase={}", newEntryBase);
}
readEndOfRecord(dis);
setRecoverNewEntryId(newEntryBase);
recordCount++;
}
break;
case OPLOG_NEW_ENTRY_0ID:
readNewEntry(dis, opCode, deletedIds, recoverValues, currentRegion, version, in,
hdos);
recordCount++;
break;
case OPLOG_MOD_ENTRY_1ID:
case OPLOG_MOD_ENTRY_2ID:
case OPLOG_MOD_ENTRY_3ID:
case OPLOG_MOD_ENTRY_4ID:
case OPLOG_MOD_ENTRY_5ID:
case OPLOG_MOD_ENTRY_6ID:
case OPLOG_MOD_ENTRY_7ID:
case OPLOG_MOD_ENTRY_8ID:
readModifyEntry(dis, opCode, deletedIds, recoverValues, currentRegion, version, in,
hdos);
recordCount++;
break;
case OPLOG_MOD_ENTRY_WITH_KEY_1ID:
case OPLOG_MOD_ENTRY_WITH_KEY_2ID:
case OPLOG_MOD_ENTRY_WITH_KEY_3ID:
case OPLOG_MOD_ENTRY_WITH_KEY_4ID:
case OPLOG_MOD_ENTRY_WITH_KEY_5ID:
case OPLOG_MOD_ENTRY_WITH_KEY_6ID:
case OPLOG_MOD_ENTRY_WITH_KEY_7ID:
case OPLOG_MOD_ENTRY_WITH_KEY_8ID:
readModifyEntryWithKey(dis, opCode, deletedIds, recoverValues, currentRegion, version,
in, hdos);
recordCount++;
break;
case OPLOG_DISK_STORE_ID:
readDiskStoreRecord(dis, this.crf.f);
foundDiskStoreRecord = true;
recordCount++;
break;
case OPLOG_MAGIC_SEQ_ID:
readOplogMagicSeqRecord(dis, this.crf.f, OPLOG_TYPE.CRF);
break;
case OPLOG_GEMFIRE_VERSION:
readGemfireVersionRecord(dis, this.crf.f);
recordCount++;
break;
case OPLOG_RVV:
readRVVRecord(dis, this.drf.f, false, latestOplog);
recordCount++;
break;
default:
throw new DiskAccessException(
String.format("Unknown opCode %s found in disk operation log.",
opCode),
getParent());
}
readLastRecord = true;
// @todo
// if (rgn.isDestroyed()) {
// break;
// }
} // while
} finally {
if (dis != null) {
dis.close();
}
if (fis != null) {
fis.close();
}
hdos.close();
}
if (!foundDiskStoreRecord && recordCount > 0) {
throw new DiskAccessException(
"The oplog file \"" + this.crf.f + "\" does not belong to the init file \""
+ getParent().getInitFile() + "\". Crf did not contain a disk store id.",
getParent());
}
} catch (EOFException ignore) {
// ignore since a partial record write can be caused by a crash
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(
String.format("Failed to read file during recovery from %s",
this.crf.f.getPath()),
ex, getParent());
} catch (CancelException e) {
if (logger.isDebugEnabled()) {
logger.debug("Oplog::readOplog:Error in recovery as Cache was closed", e);
}
} catch (RegionDestroyedException e) {
if (logger.isDebugEnabled()) {
logger.debug("Oplog::readOplog:Error in recovery as Region was destroyed", e);
}
}
// Add the Oplog size to the Directory Holder which owns this oplog,
// so that available space is correctly calculated & stats updated.
long byteCount = 0;
if (!readLastRecord) {
// this means that there was a crash
// and hence we should not continue to read
// the next oplog
this.crashed = true;
if (dis != null) {
byteCount = dis.getFileLength();
}
} else {
if (dis != null) {
byteCount = dis.getCount();
}
}
return byteCount;
}
/**
* @throws DiskAccessException if this file does not belong to our parent
*/
private void readDiskStoreRecord(DataInput dis, File f) throws IOException {
long leastSigBits = dis.readLong();
long mostSigBits = dis.readLong();
DiskStoreID readDSID = new DiskStoreID(mostSigBits, leastSigBits);
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "diskStoreId={}", readDSID);
}
readEndOfRecord(dis);
DiskStoreID dsid = getParent().getDiskStoreID();
if (!readDSID.equals(dsid)) {
throw new DiskAccessException("The oplog file \"" + f
+ "\" does not belong to the init file \"" + getParent().getInitFile() + "\".",
getParent());
}
}
/*
* Reads and validates magic sequence in oplog header. For existing files this will not exist.
* This method will throw a DiskAccessException in that case too.
*/
private void readOplogMagicSeqRecord(DataInput dis, File f, OPLOG_TYPE type) throws IOException {
byte[] seq = new byte[OPLOG_TYPE.getLen()];
dis.readFully(seq);
for (int i = 0; i < OPLOG_TYPE.getLen(); i++) {
if (seq[i] != type.getBytes()[i]) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"oplog magic code mismatched at byte:{}, value:{}", (i + 1), seq[i]);
}
throw new DiskAccessException("Invalid oplog (" + type.name() + ") file provided: " + f,
getParent());
}
}
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
StringBuffer sb = new StringBuffer();
for (int i = 0; i < OPLOG_TYPE.getLen(); i++) {
sb.append(" ").append(seq[i]);
}
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "oplog magic code: {}", sb);
}
readEndOfRecord(dis);
}
/**
* @throws DiskAccessException if this file does not belong to our parent
*/
private void readGemfireVersionRecord(DataInput dis, File f) throws IOException {
KnownVersion recoveredGFVersion = readProductVersionRecord(dis, f);
final boolean hasDataVersion;
if ((hasDataVersion = (recoveredGFVersion == KnownVersion.TOKEN))) {
// actual GFE version will be the next record in this case
byte opCode = dis.readByte();
if (opCode != OPLOG_GEMFIRE_VERSION) {
throw new DiskAccessException(
String.format("Unknown opCode %s found in disk operation log.",
opCode),
getParent());
}
recoveredGFVersion = readProductVersionRecord(dis, f);
}
if (this.gfversion == null) {
this.gfversion = recoveredGFVersion;
} else {
assert this.gfversion == recoveredGFVersion;
}
if (hasDataVersion) {
byte opCode = dis.readByte();
if (opCode != OPLOG_GEMFIRE_VERSION) {
throw new DiskAccessException(
String.format("Unknown opCode %s found in disk operation log.",
opCode),
getParent());
}
recoveredGFVersion = readProductVersionRecord(dis, f);
if (this.dataVersion == null) {
this.dataVersion = recoveredGFVersion;
} else {
assert this.dataVersion == recoveredGFVersion;
}
}
}
private KnownVersion readProductVersionRecord(DataInput dis, File f) throws IOException {
short ver = VersioningIO.readOrdinal(dis);
final KnownVersion recoveredGFVersion =
Versioning.getKnownVersionOrDefault(
Versioning.getVersion(ver), null);
if (recoveredGFVersion == null) {
throw new DiskAccessException(
String.format("Unknown version ordinal %s found when recovering Oplogs", ver),
getParent());
}
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "version={}", recoveredGFVersion);
}
readEndOfRecord(dis);
return recoveredGFVersion;
}
private void readTotalCountRecord(DataInput dis, File f) throws IOException {
long recoveredCount = InternalDataSerializer.readUnsignedVL(dis);
this.totalCount.set(recoveredCount);
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "totalCount={}", totalCount);
}
readEndOfRecord(dis);
}
private void readRVVRecord(DataInput dis, File f, boolean gcRVV, boolean latestOplog)
throws IOException {
final boolean isPersistRecoveryDebugEnabled =
logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE);
long numRegions = InternalDataSerializer.readUnsignedVL(dis);
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "readRVV entry numRegions={}", numRegions);
}
for (int region = 0; region < numRegions; region++) {
long drId = InternalDataSerializer.readUnsignedVL(dis);
// Get the drs. This may be null if this region is not currently
// recovering
DiskRecoveryStore drs = getOplogSet().getCurrentlyRecovering(drId);
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "readRVV drId={} region={}", drId, drs);
}
if (gcRVV) {
// Read the GCC RV
long rvvSize = InternalDataSerializer.readUnsignedVL(dis);
for (int memberNum = 0; memberNum < rvvSize; memberNum++) {
// for each member, read the member id and version
long memberId = InternalDataSerializer.readUnsignedVL(dis);
long gcVersion = InternalDataSerializer.readUnsignedVL(dis);
// if we have a recovery store, add the recovered regions
if (drs != null) {
Object member = getParent().getDiskInitFile().getCanonicalObject((int) memberId);
drs.recordRecoveredGCVersion((VersionSource) member, gcVersion);
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"adding gcRVV entry drId={}, member={}, version={}", drId, memberId, gcVersion);
}
} else {
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"skipping gcRVV entry drId={}, member={}, version={}", drId, memberId, gcVersion);
}
}
}
} else {
boolean rvvTrusted = DataSerializer.readBoolean(dis);
if (drs != null) {
if (latestOplog) {
// only set rvvtrust based on the newest oplog recovered
drs.setRVVTrusted(rvvTrusted);
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"marking RVV trusted drId={},tvvTrusted={}", drId, rvvTrusted);
}
}
}
// Read a regular RVV
long rvvSize = InternalDataSerializer.readUnsignedVL(dis);
for (int memberNum = 0; memberNum < rvvSize; memberNum++) {
// for each member, read the member id and version
long memberId = InternalDataSerializer.readUnsignedVL(dis);
RegionVersionHolder versionHolder = new RegionVersionHolder(dis);
if (drs != null) {
Object member = getParent().getDiskInitFile().getCanonicalObject((int) memberId);
drs.recordRecoveredVersionHolder((VersionSource) member, versionHolder, latestOplog);
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"adding RVV entry drId={}, member={}, versionHolder={}, latestOplog={}, oplogId={}",
drId, memberId, versionHolder, latestOplog, getOplogId());
}
} else {
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"skipping RVV entry drId={}, member={}, versionHolder={}", drId, memberId,
versionHolder);
}
}
}
}
}
readEndOfRecord(dis);
}
/**
* Recovers one oplog
*
* @param latestOplog - true if this oplog is the latest oplog in the disk store.
*/
long recoverCrf(OplogEntryIdSet deletedIds, boolean recoverValues, boolean recoverValuesSync,
boolean alreadyRecoveredOnce, Set<Oplog> oplogsNeedingValueRecovery, boolean latestOplog) {
// crf might not exist; but drf always will
this.diskFile = new File(this.drf.f.getParentFile(),
oplogSet.getPrefix() + getParent().getName() + "_" + this.oplogId);
File crfFile = this.crf.f;
if (crfFile == null) {
this.haveRecoveredCrf = true;
return 0L;
}
lockCompactor();
this.kvMap = new OplogEntryIdMap();
this.skippedKeyBytes = new OplogEntryIdMap();
try {
if (this.haveRecoveredCrf && isDeleted())
return 0; // do this check while holding lock
if (!this.haveRecoveredCrf) {
this.haveRecoveredCrf = true;
}
long byteCount;
// if we have a KRF then read it and delay reading the CRF.
// Unless we are in synchronous recovery mode
if (!readKrf(deletedIds, recoverValues, recoverValuesSync, oplogsNeedingValueRecovery,
latestOplog)) {
logger.info("Recovering {} {} for disk store {}.",
new Object[] {toString(), crfFile.getAbsolutePath(), getParent().getName()});
byteCount = readCrf(deletedIds, recoverValues, latestOplog);
} else {
byteCount = this.crf.f.length();
}
if (!isPhase2()) {
if (getParent().isOfflineCompacting()) {
getParent().incLiveEntryCount(getRecoveryMap().size());
}
long tc = totalCount.get();
long tlc = totalLiveCount.get();
if (getParent().isValidating() && tlc >= 0
&& tc > tlc) {
getParent().incDeadRecordCount(tc - tlc);
}
getParent().incDeadRecordCount(getRecordsSkipped());
}
if (getParent().isOfflineCompacting()) {
offlineCompact(deletedIds, latestOplog);
}
if (!alreadyRecoveredOnce) {
setRecoveredCrfSize(byteCount);
this.dirHolder.incrementTotalOplogSize(byteCount);
}
if (getParent().isOfflineCompacting()) {
if (isOplogEmpty()) {
this.deleted.set(false);
destroy();
}
}
return byteCount;
} finally {
this.kvMap = null;
this.skippedKeyBytes = null;
unlockCompactor();
}
}
private boolean offlineCompactPhase2 = false;
private boolean isPhase1() {
return !this.offlineCompactPhase2;
}
private boolean isPhase2() {
return this.offlineCompactPhase2;
}
private void offlineCompact(OplogEntryIdSet deletedIds, boolean latestOplog) {
// If we only do this if "(getRecordsSkipped() > 0)" then it will only
// compact
// an oplog that has some garbage in it.
// Instead if we do every oplog in case they set maxOplogSize
// then all oplogs will be converted to obey maxOplogSize.
// 45777: for normal offline compaction, we only do it when
// getRecordsSkipped() > 0
// but for upgrade disk store, we have to do it for pure creates oplog
if (getRecordsSkipped() > 0 || getHasDeletes() || getParent().isUpgradeVersionOnly()) {
this.offlineCompactPhase2 = true;
if (getOplogSet().getChild() == null) {
getOplogSet().initChild();
}
readCrf(deletedIds, true, latestOplog);
this.deleted.set(false);
destroyCrfOnly();
} else {
// For every live entry in this oplog add it to the deleted set
// so that we will skip it when we recovery the next oplogs.
for (OplogEntryIdMap.Iterator it = getRecoveryMap().iterator(); it.hasNext();) {
it.advance();
deletedIds.add(it.key());
}
close();
}
}
/**
* TODO soplog - This method is public just to test soplog recovery
*/
public DiskEntry.RecoveredEntry createRecoveredEntry(byte[] valueBytes, int valueLength,
byte userBits, long oplogId, long offsetInOplog, long oplogKeyId, boolean recoverValue,
KnownVersion version, ByteArrayDataInput in) {
DiskEntry.RecoveredEntry re = null;
if (recoverValue || EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits)) {
Object value;
if (EntryBits.isLocalInvalid(userBits)) {
value = Token.LOCAL_INVALID;
valueLength = 0;
} else if (EntryBits.isInvalid(userBits)) {
value = Token.INVALID;
valueLength = 0;
} else if (EntryBits.isSerialized(userBits)) {
value = DiskEntry.Helper.readSerializedValue(valueBytes, version, in, false,
getParent().getCache());
} else if (EntryBits.isTombstone(userBits)) {
value = Token.TOMBSTONE;
} else {
value = valueBytes;
}
re = new DiskEntry.RecoveredEntry(oplogKeyId, oplogId, offsetInOplog, userBits, valueLength,
value);
} else {
re = new DiskEntry.RecoveredEntry(oplogKeyId, oplogId, offsetInOplog, userBits, valueLength);
}
return re;
}
private void readEndOfRecord(DataInput di) throws IOException {
int b = di.readByte();
if (b != END_OF_RECORD_ID) {
if (b == 0) {
logger.warn(
"Detected a partial record in oplog file. Partial records can be caused by an abnormal shutdown in which case this warning can be safely ignored. They can also be caused by the oplog file being corrupted.");
// this is expected if this is the last record and we died while writing
// it.
throw new EOFException("found partial last record");
} else {
// Our implementation currently relies on all unwritten bytes having
// a value of 0. So throw this exception if we find one we didn't
// expect.
throw new IllegalStateException(
"expected end of record (byte==" + END_OF_RECORD_ID + ") or zero but found " + b);
}
}
}
private static void forceSkipBytes(CountingDataInputStream dis, int len) throws IOException {
int skipped = dis.skipBytes(len);
while (skipped < len) {
dis.readByte();
skipped++;
}
}
private int recordsSkippedDuringRecovery = 0;
private void incSkipped() {
this.recordsSkippedDuringRecovery++;
}
int getRecordsSkipped() {
return this.recordsSkippedDuringRecovery;
}
private VersionTag readVersionsFromOplog(DataInput dis) throws IOException {
if (KnownVersion.GFE_70.compareTo(currentRecoveredGFVersion()) <= 0) {
// this version format is for gemfire 7.0
// if we have different version format in 7.1, it will be handled in
// "else if"
int entryVersion = (int) InternalDataSerializer.readSignedVL(dis);
long regionVersion = InternalDataSerializer.readUnsignedVL(dis);
int memberId = (int) InternalDataSerializer.readUnsignedVL(dis);
Object member = getParent().getDiskInitFile().getCanonicalObject(memberId);
long timestamp = InternalDataSerializer.readUnsignedVL(dis);
int dsId = (int) InternalDataSerializer.readSignedVL(dis);
VersionTag vt = VersionTag.create((VersionSource) member);
vt.setEntryVersion(entryVersion);
vt.setRegionVersion(regionVersion);
vt.setMemberID((VersionSource) member);
vt.setVersionTimeStamp(timestamp);
vt.setDistributedSystemId(dsId);
return vt;
} else {
// pre-7.0
return null;
}
}
private synchronized VersionTag createDummyTag(DiskRecoveryStore drs) {
DiskStoreID member = getParent().getDiskStoreID();
int memberid = getParent().getDiskInitFile().getOrCreateCanonicalId(member);
long regionVersion = drs.getVersionForMember(member);
VersionTag vt = VersionTag.create(member);
vt.setEntryVersion(1);
vt.setRegionVersion(regionVersion + 1);
vt.setMemberID(member);
vt.setVersionTimeStamp(getParent().getCache().cacheTimeMillis());
vt.setDistributedSystemId(-1);
return vt;
}
/**
* Returns true if the values for the given disk recovery store should be recovered.
*/
private boolean recoverLruValue(DiskRecoveryStore drs) {
if (isLruValueRecoveryDisabled(drs)) {
return false;
} else if (drs.lruLimitExceeded()) {
this.stats.incRecoveredValuesSkippedDueToLRU();
return false;
}
return true;
}
private boolean isLruValueRecoveryDisabled(DiskRecoveryStore store) {
return !store.getDiskStore().isOffline() && !getParent().RECOVER_LRU_VALUES
&& !store.getEvictionAttributes().getAlgorithm().isNone();
}
/**
* Reads an oplog entry of type Create
*
* @param dis DataInputStream from which the oplog is being read
* @param opcode byte whether the id is short/int/long
*/
private void readNewEntry(CountingDataInputStream dis, byte opcode, OplogEntryIdSet deletedIds,
boolean recoverValue, final LocalRegion currentRegion, KnownVersion version,
ByteArrayDataInput in,
HeapDataOutputStream hdos) throws IOException {
final boolean isPersistRecoveryDebugEnabled =
logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE);
long oplogOffset = -1;
byte userBits = dis.readByte();
byte[] objValue = null;
int valueLength = 0;
long oplogKeyId = incRecoverNewEntryId();
long drId = DiskInitFile.readDiskRegionID(dis);
DiskRecoveryStore drs = getOplogSet().getCurrentlyRecovering(drId);
// read versions
VersionTag tag = null;
boolean isDummy = false;
if (EntryBits.isWithVersions(userBits)) {
tag = readVersionsFromOplog(dis);
isDummy = false;
} else if (getParent().isUpgradeVersionOnly() && drs != null) {
tag = this.createDummyTag(drs);
userBits = EntryBits.setWithVersions(userBits, true);
isDummy = true;
}
if (drs != null
&& !drs.getDiskRegionView().getFlags().contains(DiskRegionFlag.IS_WITH_VERSIONING)) {
// 50044 Remove version tag from entry if we don't want versioning for
// this region
tag = null;
userBits = EntryBits.setWithVersions(userBits, false);
}
OkToSkipResult skipResult = okToSkipModifyRecord(deletedIds, drId, drs, oplogKeyId, true, tag);
if (skipResult.skip()) {
if (!isPhase2()) {
this.stats.incRecoveryRecordsSkipped();
incSkipped();
}
} else if (recoverValue && !getParent().isOfflineCompacting()) {
recoverValue = recoverLruValue(drs);
}
CompactionRecord p2cr = null;
long crOffset;
if (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits)) {
if (EntryBits.isInvalid(userBits)) {
objValue = DiskEntry.INVALID_BYTES;
} else if (EntryBits.isTombstone(userBits)) {
objValue = DiskEntry.TOMBSTONE_BYTES;
} else {
objValue = DiskEntry.LOCAL_INVALID_BYTES;
}
crOffset = dis.getCount();
if (!skipResult.skip()) {
if (isPhase2()) {
p2cr = (CompactionRecord) getRecoveryMap().get(oplogKeyId);
if (p2cr != null && p2cr.getOffset() != crOffset) {
skipResult = OkToSkipResult.SKIP_RECORD;
}
}
}
} else {
int len = dis.readInt();
oplogOffset = dis.getCount();
crOffset = oplogOffset;
valueLength = len;
if (!skipResult.skip()) {
if (isPhase2()) {
p2cr = (CompactionRecord) getRecoveryMap().get(oplogKeyId);
if (p2cr != null && p2cr.getOffset() != crOffset) {
skipResult = OkToSkipResult.SKIP_RECORD;
}
}
}
if (recoverValue && !skipResult.skip()) {
byte[] valueBytes = new byte[len];
dis.readFully(valueBytes);
objValue = valueBytes;
validateValue(valueBytes, userBits, version, in);
} else {
forceSkipBytes(dis, len);
}
}
{
int len = dis.readInt();
incTotalCount();
if (skipResult.skip()) {
if (skipResult.skipKey()) {
forceSkipBytes(dis, len);
} else {
byte[] keyBytes = new byte[len];
dis.readFully(keyBytes);
skippedKeyBytes.put(oplogKeyId, keyBytes);
}
readEndOfRecord(dis);
if (drs != null && tag != null) {
// Update the RVV with the new entry
// This must be done after reading the end of record to make sure
// we don't have a corrupt record. See bug #45538
drs.recordRecoveredVersionTag(tag);
}
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readNewEntry SKIPPING oplogKeyId=<{}> drId={} userBits={} keyLen={} valueLen={} tag={}",
oplogKeyId, drId, userBits, len, valueLength, tag);
}
} else {
byte[] keyBytes = null;
if (isPhase2()) {
forceSkipBytes(dis, len);
} else {
keyBytes = new byte[len];
dis.readFully(keyBytes);
}
readEndOfRecord(dis);
if (drs != null && tag != null) {
// Update the RVV with the new entry
// This must be done after reading the end of record to make sure
// we don't have a corrupt record. See bug #45538
drs.recordRecoveredVersionTag(tag);
}
if (getParent().isOfflineCompacting()) {
if (isPhase1()) {
CompactionRecord cr = new CompactionRecord(keyBytes, crOffset);
getRecoveryMap().put(oplogKeyId, cr);
if (drs != null) {
drs.getDiskRegionView().incRecoveredEntryCount();
}
this.stats.incRecoveredEntryCreates();
} else { // phase2
Assert.assertNotNull(p2cr, "First pass did not find create a compaction record");
getOplogSet().getChild().copyForwardForOfflineCompact(oplogKeyId, p2cr.getKeyBytes(),
objValue, userBits, drId, tag);
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readNewEntry copyForward oplogKeyId=<{}>", oplogKeyId);
}
// add it to the deletedIds set so we will ignore it in earlier
// oplogs
deletedIds.add(oplogKeyId);
}
} else {
Object key = deserializeKey(keyBytes, version, in);
{
Object oldValue = getRecoveryMap().put(oplogKeyId, key);
if (oldValue != null) {
throw new AssertionError(
String.format(
"Oplog::readNewEntry: Create is present in more than one Oplog. This should not be possible. The Oplog Key ID for this entry is %s.",
oplogKeyId));
}
}
if (drs != null) {
DiskEntry de = drs.getDiskEntry(key);
if (de == null) {
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readNewEntry oplogKeyId=<{}> drId={} key={} userBits={} oplogOffset={} valueLen={} tag={}",
oplogKeyId, drId, key, userBits, oplogOffset, valueLength, tag);
}
DiskEntry.RecoveredEntry re = createRecoveredEntry(objValue, valueLength, userBits,
getOplogId(), oplogOffset, oplogKeyId, recoverValue, version, in);
if (tag != null) {
re.setVersionTag(tag);
}
initRecoveredEntry(drs.getDiskRegionView(), drs.initializeRecoveredEntry(key, re));
drs.getDiskRegionView().incRecoveredEntryCount();
this.stats.incRecoveredEntryCreates();
} else {
DiskId curdid = de.getDiskId();
assert curdid.getOplogId() != getOplogId();
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"ignore readNewEntry because getOplogId()={} != curdid.getOplogId()={} for drId={} key={}",
getOplogId(), curdid.getOplogId(), drId, key);
}
}
}
}
}
}
}
/**
* Reads an oplog entry of type Modify
*
* @param dis DataInputStream from which the oplog is being read
* @param opcode byte whether the id is short/int/long
*/
private void readModifyEntry(CountingDataInputStream dis, byte opcode, OplogEntryIdSet deletedIds,
boolean recoverValue, LocalRegion currentRegion, KnownVersion version, ByteArrayDataInput in,
HeapDataOutputStream hdos) throws IOException {
final boolean isPersistRecoveryDebugEnabled =
logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE);
long oplogOffset = -1;
byte userBits = dis.readByte();
int idByteCount = (opcode - OPLOG_MOD_ENTRY_1ID) + 1;
long oplogKeyId = getModEntryId(dis, idByteCount);
long drId = DiskInitFile.readDiskRegionID(dis);
DiskRecoveryStore drs = getOplogSet().getCurrentlyRecovering(drId);
// read versions
VersionTag tag = null;
if (EntryBits.isWithVersions(userBits)) {
tag = readVersionsFromOplog(dis);
} else if (getParent().isUpgradeVersionOnly() && drs != null) {
tag = this.createDummyTag(drs);
userBits = EntryBits.setWithVersions(userBits, true);
}
if (drs != null
&& !drs.getDiskRegionView().getFlags().contains(DiskRegionFlag.IS_WITH_VERSIONING)) {
// 50044 Remove version tag from entry if we don't want versioning for
// this region
tag = null;
userBits = EntryBits.setWithVersions(userBits, false);
}
OkToSkipResult skipResult = okToSkipModifyRecord(deletedIds, drId, drs, oplogKeyId, false, tag);
if (skipResult.skip()) {
if (!isPhase2()) {
incSkipped();
this.stats.incRecoveryRecordsSkipped();
}
} else if (recoverValue && !getParent().isOfflineCompacting()) {
recoverValue = recoverLruValue(drs);
}
byte[] objValue = null;
int valueLength = 0;
CompactionRecord p2cr = null;
long crOffset;
if (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits)) {
if (EntryBits.isInvalid(userBits)) {
objValue = DiskEntry.INVALID_BYTES;
} else if (EntryBits.isTombstone(userBits)) {
objValue = DiskEntry.TOMBSTONE_BYTES;
} else {
objValue = DiskEntry.LOCAL_INVALID_BYTES;
}
crOffset = dis.getCount();
if (!skipResult.skip()) {
if (isPhase2()) {
p2cr = (CompactionRecord) getRecoveryMap().get(oplogKeyId);
if (p2cr != null && p2cr.getOffset() != crOffset) {
skipResult = OkToSkipResult.SKIP_RECORD;
}
}
}
} else {
int len = dis.readInt();
oplogOffset = dis.getCount();
crOffset = oplogOffset;
valueLength = len;
if (!skipResult.skip()) {
if (isPhase2()) {
p2cr = (CompactionRecord) getRecoveryMap().get(oplogKeyId);
if (p2cr != null && p2cr.getOffset() != crOffset) {
skipResult = OkToSkipResult.SKIP_RECORD;
}
}
}
if (!skipResult.skip() && recoverValue) {
byte[] valueBytes = new byte[len];
dis.readFully(valueBytes);
objValue = valueBytes;
validateValue(valueBytes, userBits, version, in);
} else {
forceSkipBytes(dis, len);
}
}
readEndOfRecord(dis);
if (drs != null && tag != null) {
// Update the RVV with the new entry
// This must be done after reading the end of record to make sure
// we don't have a corrupt record. See bug #45538
drs.recordRecoveredVersionTag(tag);
}
incTotalCount();
if (!skipResult.skip()) {
Object key = getRecoveryMap().get(oplogKeyId);
// if the key is not in the recover map, it's possible it
// was previously skipped. Check the skipped bytes map for the key.
if (key == null) {
byte[] keyBytes = (byte[]) skippedKeyBytes.get(oplogKeyId);
if (keyBytes != null) {
key = deserializeKey(keyBytes, version, in);
}
}
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readModifyEntry oplogKeyId=<{}> drId={} key=<{}> userBits={} oplogOffset={} tag={} valueLen={}",
oplogKeyId, drId, key, userBits, oplogOffset, tag, valueLength);
}
// Will no longer be null since 1st modify record in any oplog
// will now be a MOD_ENTRY_WITH_KEY record.
assert key != null;
if (getParent().isOfflineCompacting()) {
if (isPhase1()) {
CompactionRecord cr = (CompactionRecord) key;
incSkipped(); // we are going to compact the previous record away
cr.update(crOffset);
} else { // phase2
Assert.assertNotNull(p2cr, "First pass did not find create a compaction record");
getOplogSet().getChild().copyForwardForOfflineCompact(oplogKeyId, p2cr.getKeyBytes(),
objValue, userBits, drId, tag);
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readModifyEntry copyForward oplogKeyId=<{}>", oplogKeyId);
}
// add it to the deletedIds set so we will ignore it in earlier oplogs
deletedIds.add(oplogKeyId);
}
} else {
// Check the actual region to see if it has this key from
// a previous recovered oplog.
if (drs != null) {
DiskEntry de = drs.getDiskEntry(key);
// This may actually be create, if the previous create or modify
// of this entry was cleared through the RVV clear.
if (de == null) {
DiskRegionView drv = drs.getDiskRegionView();
// and create an entry
DiskEntry.RecoveredEntry re = createRecoveredEntry(objValue, valueLength, userBits,
getOplogId(), oplogOffset, oplogKeyId, recoverValue, version, in);
if (tag != null) {
re.setVersionTag(tag);
}
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readModEntryWK init oplogKeyId=<{}> drId={} key=<{}> oplogOffset={} userBits={} valueLen={} tag={}",
oplogKeyId, drId, key, oplogOffset, userBits, valueLength, tag);
}
initRecoveredEntry(drv, drs.initializeRecoveredEntry(key, re));
drs.getDiskRegionView().incRecoveredEntryCount();
this.stats.incRecoveredEntryCreates();
} else {
DiskEntry.RecoveredEntry re = createRecoveredEntry(objValue, valueLength, userBits,
getOplogId(), oplogOffset, oplogKeyId, recoverValue, version, in);
if (tag != null) {
re.setVersionTag(tag);
}
de = drs.updateRecoveredEntry(key, re);
updateRecoveredEntry(drs.getDiskRegionView(), de, re);
this.stats.incRecoveredEntryUpdates();
}
}
}
} else {
if (isPersistRecoveryDebugEnabled) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"skipping readModifyEntry oplogKeyId=<{}> drId={}", oplogKeyId, drId);
}
}
}
private void readVersionTagOnlyEntry(CountingDataInputStream dis, byte opcode)
throws IOException {
long oplogOffset = -1;
long drId = DiskInitFile.readDiskRegionID(dis);
DiskRecoveryStore drs = getOplogSet().getCurrentlyRecovering(drId);
// read versions
VersionTag tag = readVersionsFromOplog(dis);
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "readVersionTagOnlyEntry drId={} tag={}",
drId, tag);
}
readEndOfRecord(dis);
// Update the RVV with the new entry
if (drs != null) {
drs.recordRecoveredVersionTag(tag);
}
}
private void validateValue(byte[] valueBytes, byte userBits, KnownVersion version,
ByteArrayDataInput in) {
if (getParent().isValidating()) {
if (EntryBits.isSerialized(userBits)) {
// make sure values are deserializable
if (!PdxWriterImpl.isPdx(valueBytes)) { // fix bug 43011
try {
DiskEntry.Helper.readSerializedValue(valueBytes, version, in, true,
getParent().getCache());
} catch (SerializationException ex) {
if (logger.isDebugEnabled()) {
logger.debug("Could not deserialize recovered value: {}", ex.getCause(), ex);
}
}
}
}
}
}
/**
* Reads an oplog entry of type ModifyWithKey
*
* @param dis DataInputStream from which the oplog is being read
* @param opcode byte whether the id is short/int/long
*/
private void readModifyEntryWithKey(CountingDataInputStream dis, byte opcode,
OplogEntryIdSet deletedIds, boolean recoverValue, final LocalRegion currentRegion,
KnownVersion version, ByteArrayDataInput in, HeapDataOutputStream hdos) throws IOException {
long oplogOffset = -1;
byte userBits = dis.readByte();
int idByteCount = (opcode - OPLOG_MOD_ENTRY_WITH_KEY_1ID) + 1;
long oplogKeyId = getModEntryId(dis, idByteCount);
long drId = DiskInitFile.readDiskRegionID(dis);
DiskRecoveryStore drs = getOplogSet().getCurrentlyRecovering(drId);
// read version
VersionTag tag = null;
if (EntryBits.isWithVersions(userBits)) {
tag = readVersionsFromOplog(dis);
} else if (getParent().isUpgradeVersionOnly() && drs != null) {
tag = this.createDummyTag(drs);
userBits = EntryBits.setWithVersions(userBits, true);
}
if (drs != null
&& !drs.getDiskRegionView().getFlags().contains(DiskRegionFlag.IS_WITH_VERSIONING)) {
// 50044 Remove version tag from entry if we don't want versioning for
// this region
tag = null;
userBits = EntryBits.setWithVersions(userBits, false);
}
OkToSkipResult skipResult = okToSkipModifyRecord(deletedIds, drId, drs, oplogKeyId, true, tag);
if (skipResult.skip()) {
if (!isPhase2()) {
incSkipped();
this.stats.incRecoveryRecordsSkipped();
}
} else if (recoverValue && !getParent().isOfflineCompacting()) {
recoverValue = recoverLruValue(drs);
}
byte[] objValue = null;
int valueLength = 0;
CompactionRecord p2cr = null;
long crOffset;
if (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits)) {
if (EntryBits.isInvalid(userBits)) {
objValue = DiskEntry.INVALID_BYTES;
} else if (EntryBits.isTombstone(userBits)) {
objValue = DiskEntry.TOMBSTONE_BYTES;
} else {
objValue = DiskEntry.LOCAL_INVALID_BYTES;
}
crOffset = dis.getCount();
if (!skipResult.skip()) {
if (isPhase2()) {
p2cr = (CompactionRecord) getRecoveryMap().get(oplogKeyId);
if (p2cr != null && p2cr.getOffset() != crOffset) {
skipResult = OkToSkipResult.SKIP_RECORD;
}
}
}
} else {
int len = dis.readInt();
oplogOffset = dis.getCount();
crOffset = oplogOffset;
valueLength = len;
if (!skipResult.skip()) {
if (isPhase2()) {
p2cr = (CompactionRecord) getRecoveryMap().get(oplogKeyId);
if (p2cr != null && p2cr.getOffset() != crOffset) {
skipResult = OkToSkipResult.SKIP_RECORD;
}
}
}
if (!skipResult.skip() && recoverValue) {
byte[] valueBytes = new byte[len];
dis.readFully(valueBytes);
objValue = valueBytes;
validateValue(valueBytes, userBits, version, in);
} else {
forceSkipBytes(dis, len);
}
}
int keyLen = dis.readInt();
incTotalCount();
if (skipResult.skip()) {
if (skipResult.skipKey()) {
forceSkipBytes(dis, keyLen);
} else {
byte[] keyBytes = new byte[keyLen];
dis.readFully(keyBytes);
skippedKeyBytes.put(oplogKeyId, keyBytes);
}
readEndOfRecord(dis);
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"skipping readModEntryWK init oplogKeyId=<{}> drId={}", oplogKeyId, drId);
}
} else {
// read the key
byte[] keyBytes = null;
if (isPhase2()) {
forceSkipBytes(dis, keyLen);
} else {
keyBytes = new byte[keyLen];
dis.readFully(keyBytes);
}
readEndOfRecord(dis);
if (drs != null && tag != null) {
// Update the RVV with the new entry
// This must be done after reading the end of record to make sure
// we don't have a corrupt record. See bug #45538
drs.recordRecoveredVersionTag(tag);
}
assert oplogKeyId >= 0;
if (getParent().isOfflineCompacting()) {
if (isPhase1()) {
CompactionRecord cr = new CompactionRecord(keyBytes, crOffset);
getRecoveryMap().put(oplogKeyId, cr);
if (drs != null) {
drs.getDiskRegionView().incRecoveredEntryCount();
}
this.stats.incRecoveredEntryCreates();
} else { // phase2
Assert.assertNotNull(p2cr, "First pass did not find create a compaction record");
getOplogSet().getChild().copyForwardForOfflineCompact(oplogKeyId, p2cr.getKeyBytes(),
objValue, userBits, drId, tag);
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readModifyEntryWithKey copyForward oplogKeyId=<{}>", oplogKeyId);
}
// add it to the deletedIds set so we will ignore it in earlier oplogs
deletedIds.add(oplogKeyId);
}
} else {
Object key = deserializeKey(keyBytes, version, in);
Object oldValue = getRecoveryMap().put(oplogKeyId, key);
if (oldValue != null) {
throw new AssertionError(
String.format(
"Oplog::readNewEntry: Create is present in more than one Oplog. This should not be possible. The Oplog Key ID for this entry is %s.",
oplogKeyId));
}
// Check the actual region to see if it has this key from
// a previous recovered oplog.
if (drs != null) {
DiskEntry de = drs.getDiskEntry(key);
if (de == null) {
DiskRegionView drv = drs.getDiskRegionView();
// and create an entry
DiskEntry.RecoveredEntry re = createRecoveredEntry(objValue, valueLength, userBits,
getOplogId(), oplogOffset, oplogKeyId, recoverValue, version, in);
if (tag != null) {
re.setVersionTag(tag);
}
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"readModEntryWK init oplogKeyId=<{}> drId={} key={} oplogOffset={} userBits={} valueLen={} tag={}",
oplogKeyId, drId, key, oplogOffset, userBits, valueLength, tag);
}
initRecoveredEntry(drv, drs.initializeRecoveredEntry(key, re));
drs.getDiskRegionView().incRecoveredEntryCount();
this.stats.incRecoveredEntryCreates();
} else {
DiskId curdid = de.getDiskId();
assert curdid
.getOplogId() != getOplogId() : "Mutiple ModEntryWK in the same oplog for getOplogId()="
+ getOplogId() + " , curdid.getOplogId()=" + curdid.getOplogId()
+ " , for drId="
+ drId + " , key=" + key;
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"ignore readModEntryWK because getOplogId()={} != curdid.getOplogId()={} for drId={} key={}",
getOplogId(), curdid.getOplogId(), drId, key);
}
}
}
}
}
}
/**
* Reads an oplog entry of type Delete
*
* @param dis DataInputStream from which the oplog is being read
* @param opcode byte whether the id is short/int/long
* @param parent instance of disk region
*/
private void readDelEntry(CountingDataInputStream dis, byte opcode, OplogEntryIdSet deletedIds,
DiskStoreImpl parent) throws IOException {
int idByteCount = (opcode - OPLOG_DEL_ENTRY_1ID) + 1;
long oplogKeyId = getDelEntryId(dis, idByteCount);
readEndOfRecord(dis);
deletedIds.add(oplogKeyId);
setHasDeletes(true);
this.stats.incRecoveredEntryDestroys();
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "readDelEntry oplogKeyId=<{}>", oplogKeyId);
}
}
/**
* Keeps track of the drId of Regions that have records in this oplog that have not yet been
* recovered. If this count is > 0 then this oplog can not be compacted.
*/
private final AtomicInteger unrecoveredRegionCount = new AtomicInteger();
private void addUnrecoveredRegion(long drId) {
DiskRegionInfo dri = getOrCreateDRI(drId);
if (dri.testAndSetUnrecovered()) {
this.unrecoveredRegionCount.incrementAndGet();
}
}
/**
* For each dri that this oplog has that is currently unrecoverable check to see if a DiskRegion
* that is recoverable now exists.
*/
void checkForRecoverableRegion(DiskRegionView dr) {
if (this.unrecoveredRegionCount.get() > 0) {
DiskRegionInfo dri = getDRI(dr);
if (dri != null) {
if (dri.testAndSetRecovered(dr)) {
this.unrecoveredRegionCount.decrementAndGet();
}
}
}
}
void updateDiskRegion(DiskRegionView dr) {
DiskRegionInfo dri = getDRI(dr);
if (dri != null) {
dri.setDiskRegion(dr);
}
}
/**
* Returns true if it is ok the skip the current modify record which had the given oplogEntryId.
* It is ok to skip if any of the following are true: 1. deletedIds contains the id 2. the last
* modification of the entry was done by a record read from an oplog other than this oplog
*/
private OkToSkipResult okToSkipModifyRecord(OplogEntryIdSet deletedIds, long drId,
DiskRecoveryStore drs, long oplogEntryId, boolean checkRecoveryMap, VersionTag tag) {
if (deletedIds.contains(oplogEntryId)) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"okToSkip because oplogEntryId={} was deleted for drId={}", oplogEntryId, drId);
}
return OkToSkipResult.SKIP_RECORD;
}
if (drs == null) { // we are not currently recovering this region
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "okToSkip because drs is null for drId={}",
drId);
}
// Now when the diskStore is created we recover all the regions
// immediately.
// After that they can close and reopen a region but the close code calls
// addUnrecoveredRegion. So I think at this point we don't need to do
// anything.
// // Prevent compactor from removing this oplog.
// // It needs to be in this state until all the regions stored it in
// // are recovered.
// addUnrecoveredRegion(drId);
return OkToSkipResult.SKIP_RECORD;
}
if (!checkRecoveryMap) {
Object key = getRecoveryMap().get(oplogEntryId);
if (key != null) {
DiskEntry de = drs.getDiskEntry(key);
if (de != null) {
DiskId curdid = de.getDiskId();
if (curdid != null) {
if (curdid.getOplogId() != getOplogId()) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"okToSkip because getOplogId()={} != curdid.getOplogId()={} for drId={} key={}",
getOplogId(), curdid.getOplogId(), drId, key);
}
return OkToSkipResult.SKIP_RECORD;
}
}
}
}
}
return okToSkipRegion(drs.getDiskRegionView(), oplogEntryId, tag);
}
/**
* Returns true if the drId region has been destroyed or if oplogKeyId preceeds the last clear
* done on the drId region
*/
private OkToSkipResult okToSkipRegion(DiskRegionView drv, long oplogKeyId, VersionTag tag) {
long lastClearKeyId = drv.getClearOplogEntryId();
if (lastClearKeyId != DiskStoreImpl.INVALID_ID) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "lastClearKeyId={} oplogKeyId={}",
lastClearKeyId, oplogKeyId);
}
if (lastClearKeyId >= 0) {
if (oplogKeyId <= lastClearKeyId) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"okToSkip because oplogKeyId={} <= lastClearKeyId={} for drId={}", oplogKeyId,
lastClearKeyId, drv.getId());
}
// @todo add some wraparound logic
return OkToSkipResult.SKIP_RECORD;
}
} else {
// lastClearKeyId is < 0 which means it wrapped around
// treat it like an unsigned value (-1 == MAX_UNSIGNED)
if (oplogKeyId > 0 || oplogKeyId <= lastClearKeyId) {
// If oplogKeyId > 0 then it happened before the clear
// (assume clear happened after we wrapped around to negative).
// If oplogKeyId < 0 then it happened before the clear
// if it is < lastClearKeyId
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"okToSkip because oplogKeyId={} <= lastClearKeyId={} for drId={}", oplogKeyId,
lastClearKeyId, drv.getId());
}
return OkToSkipResult.SKIP_RECORD;
}
}
}
RegionVersionVector clearRVV = drv.getClearRVV();
if (clearRVV != null) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE, "clearRVV={} tag={}", clearRVV, tag);
}
if (clearRVV.contains(tag.getMemberID(), tag.getRegionVersion())) {
if (logger.isTraceEnabled(LogMarker.PERSIST_RECOVERY_VERBOSE)) {
logger.trace(LogMarker.PERSIST_RECOVERY_VERBOSE,
"okToSkip because tag={} <= clearRVV={} for drId={}", tag, clearRVV, drv.getId());
}
// For an RVV clear, we can only skip the value during recovery
// because later modifies may use the oplog key id.
return OkToSkipResult.SKIP_VALUE;
}
}
return OkToSkipResult.DONT_SKIP;
}
private long getModEntryId(CountingDataInputStream dis, int idByteCount) throws IOException {
return calcModEntryId(getEntryIdDelta(dis, idByteCount));
}
private long getDelEntryId(CountingDataInputStream dis, int idByteCount) throws IOException {
return calcDelEntryId(getEntryIdDelta(dis, idByteCount));
}
private static long getEntryIdDelta(CountingDataInputStream dis, int idByteCount)
throws IOException {
assert idByteCount >= 1 && idByteCount <= 8 : idByteCount;
long delta;
delta = dis.readByte();
idByteCount--;
while (idByteCount > 0) {
delta <<= 8;
delta |= (0x00FF & dis.readByte());
idByteCount--;
}
return delta;
}
/**
* Call this when the cache is closed or region is destroyed. Deletes the lock files.
*/
public void close() {
if (this.closed) {
return;
}
if (logger.isDebugEnabled()) {
logger.debug("Oplog::close: Store name ={} Oplog ID = {}", parent.getName(), oplogId);
}
basicClose(false);
}
/**
* Close the files of a oplog but don't set any state. Used by unit tests
*/
public void testClose() {
try {
this.crf.channel.close();
} catch (IOException ignore) {
}
try {
this.crf.raf.close();
} catch (IOException ignore) {
}
this.crf.RAFClosed = true;
try {
this.drf.channel.close();
} catch (IOException ignore) {
}
try {
this.drf.raf.close();
} catch (IOException ignore) {
}
this.drf.RAFClosed = true;
}
private void basicClose(boolean forceDelete) {
flushAll();
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* crf */) {
unpreblow(this.crf, getMaxCrfSize());
if (!this.crf.RAFClosed) {
try {
this.crf.channel.close();
} catch (IOException ignore) {
}
try {
this.crf.raf.close();
} catch (IOException ignore) {
}
this.crf.RAFClosed = true;
this.stats.decOpenOplogs();
}
this.closed = true;
}
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* drf */) {
unpreblow(this.drf, getMaxDrfSize());
if (!this.drf.RAFClosed) {
try {
this.drf.channel.close();
} catch (IOException ignore) {
}
try {
this.drf.raf.close();
} catch (IOException ignore) {
}
this.drf.RAFClosed = true;
}
}
if (forceDelete) {
this.deleteFiles(false);
}
}
/**
* Used by tests to confirm that an oplog was compacted
*/
boolean testConfirmCompacted() {
return this.closed && this.deleted.get() && getOplogSize() == 0;
}
// @todo add state to determine when both crf and drf and been deleted.
/**
* Note that this can return true even when we still need to keep the oplog around because its drf
* is still needed.
*/
boolean isDeleted() {
return this.deleted.get();
}
/**
* Destroys this oplog. First it will call close which will cleanly close all Async threads and
* then the oplog file will be deleted. The deletion of lock files will be taken care of by the
* close.
*
*/
public void destroy() {
lockCompactor();
try {
if (!this.closed) {
this.basicClose(true /* force delete */);
} else {
// do the following even if we were already closed
deleteFiles(false);
}
} finally {
unlockCompactor();
}
}
/*
* In offline compaction, after compacted each oplog, only the crf will be deleted. Oplog with drf
* only will be housekepted later.
*/
public void destroyCrfOnly() {
lockCompactor();
try {
if (!this.closed) {
this.basicClose(true /* force delete */);
} else {
// do the following even if we were already closed
deleteFiles(true);
}
} finally {
unlockCompactor();
}
}
/**
* A check to confirm that the oplog has been closed because of the cache being closed
*
*/
private void checkClosed() {
getParent().getCancelCriterion().checkCancelInProgress(null);
if (!this.closed) {
return;
}
throw new OplogCancelledException("This Oplog has been closed.");
}
/**
* Return the number of bytes needed to encode the given long. Value returned will be >= 1 and <=
* 8.
*/
static byte bytesNeeded(long v) {
if (v < 0) {
v = ~v;
}
return (byte) (((64 - Long.numberOfLeadingZeros(v)) / 8) + 1);
}
/**
* Return absolute value of v.
*/
static long abs(long v) {
if (v < 0) {
return -v;
} else {
return v;
}
}
private long calcDelta(long opLogKeyID, byte opCode) {
long delta;
if (opCode == OPLOG_DEL_ENTRY_1ID) {
delta = opLogKeyID - this.writeDelEntryId;
this.writeDelEntryId = opLogKeyID;
} else {
delta = opLogKeyID - this.writeModEntryId;
this.writeModEntryId = opLogKeyID;
}
return delta;
}
/**
* This function records all the data for the current op into this.opState.
*
* @param opCode The int value identifying whether it is create/modify or delete operation
* @param entry The DiskEntry object being operated upon
* @param value The byte array representing the value
*/
private void initOpState(byte opCode, DiskRegionView dr, DiskEntry entry, ValueWrapper value,
byte userBits, boolean notToUseUserBits) throws IOException {
this.opState.initialize(opCode, dr, entry, value, userBits, notToUseUserBits);
}
private void clearOpState() {
this.opState.clear();
}
/**
* Returns the number of bytes it will take to serialize this.opState.
*/
private int getOpStateSize() {
return this.opState.getSize();
}
private int getOpStateValueOffset() {
return this.opState.getValueOffset();
}
private byte calcUserBits(ValueWrapper vw) {
return vw.getUserBits();
}
/**
* Returns true if the given entry has not yet been written to this oplog.
*/
private boolean modNeedsKey(DiskEntry entry) {
DiskId did = entry.getDiskId();
synchronized (did) {
if (did.getOplogId() != getOplogId()) {
// the last record for it was written in a different oplog
// so we need the key.
return true;
} else {
return false;
}
}
}
/**
* Modified the code so as to reuse the already created ByteBuffer during transition. Creates a
* key/value pair from a region entry on disk. Updates all of the necessary
* {@linkplain DiskStoreStats statistics} and invokes basicCreate
*
* @param entry The DiskEntry object for this key/value pair.
* @param value byte array representing the value
*/
public void create(InternalRegion region, DiskEntry entry, ValueWrapper value, boolean async) {
if (this != getOplogSet().getChild()) {
getOplogSet().getChild().create(region, entry, value, async);
} else {
DiskId did = entry.getDiskId();
boolean exceptionOccurred = false;
byte prevUsrBit = did.getUserBits();
int len = did.getValueLength();
try {
// It is ok to do this outside of "lock" because
// create records do not need to change.
byte userBits = calcUserBits(value);
// save versions for creates and updates even if value is bytearrary in
// 7.0
if (entry.getVersionStamp() != null) {
if (entry.getVersionStamp().getMemberID() == null) {
throw new AssertionError(
"Version stamp should have a member at this point for entry " + entry);
}
// pdx and tx will not use version
userBits = EntryBits.setWithVersions(userBits, true);
}
basicCreate(region.getDiskRegion(), entry, value, userBits, async);
} catch (IOException ex) {
exceptionOccurred = true;
region.getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format("Failed writing key to %s",
this.diskFile.getPath()), ex, region.getFullPath());
} catch (InterruptedException ie) {
Thread.currentThread().interrupt();
exceptionOccurred = true;
region.getCancelCriterion().checkCancelInProgress(ie);
throw new DiskAccessException(
String.format(
"Failed writing key to %s due to failure in acquiring read lock for asynch writing",
this.diskFile.getPath()),
ie, region.getFullPath());
} finally {
if (exceptionOccurred) {
did.setValueLength(len);
did.setUserBits(prevUsrBit);
}
}
}
}
/**
* Return true if no records have been written to the oplog yet.
*/
private boolean isFirstRecord() {
return this.firstRecord;
}
/**
* A helper function which identifies whether to create the entry in the current oplog or to make
* the switch to the next oplog. This function enables us to reuse the byte buffer which got
* created for an oplog which no longer permits us to use itself
*
* @param entry DiskEntry object representing the current Entry
*/
private void basicCreate(DiskRegion dr, DiskEntry entry, ValueWrapper value, byte userBits,
boolean async) throws IOException, InterruptedException {
DiskId id = entry.getDiskId();
boolean useNextOplog = false;
long startPosForSynchOp = -1;
if (DiskStoreImpl.KRF_DEBUG) {
// wait for cache close to create krf
System.out.println("basicCreate KRF_DEBUG");
Thread.sleep(1000);
}
getParent().getBackupLock().lock();
try {
synchronized (this.lock) { // TODO soplog perf analysis shows this as a
// contention point
// synchronized (this.crf) {
initOpState(OPLOG_NEW_ENTRY_0ID, dr, entry, value, userBits, false);
// Check if the current data in ByteBuffer will cause a
// potential increase in the size greater than the max allowed
long temp = (getOpStateSize() + this.crf.currSize);
if (!this.wroteNewEntryBase) {
temp += OPLOG_NEW_ENTRY_BASE_REC_SIZE;
}
if (this != getOplogSet().getChild()) {
useNextOplog = true;
} else if (temp > getMaxCrfSize() && !isFirstRecord()) {
switchOpLog(dr, getOpStateSize(), entry);
useNextOplog = true;
} else {
if (this.lockedForKRFcreate) {
CacheClosedException cce =
getParent().getCache().getCacheClosedException("The disk store is closed.");
dr.getCancelCriterion().checkCancelInProgress(cce);
throw cce;
}
this.firstRecord = false;
writeNewEntryBaseRecord(async);
// Now we can finally call newOplogEntryId.
// We need to make sure the create records
// are written in the same order as they are created.
// This allows us to not encode the oplogEntryId explicitly in the
// record
long createOplogEntryId = getOplogSet().newOplogEntryId();
id.setKeyId(createOplogEntryId);
// startPosForSynchOp = this.crf.currSize;
// Allow it to be added to the OpLOg so increase the
// size of currenstartPosForSynchOpt oplog
int dataLength = getOpStateSize();
// It is necessary that we set the
// Oplog ID here without releasing the lock on object as we are
// writing to the file after releasing the lock. This can cause
// a situation where the
// switching thread has added Oplog for compaction while the previous
// thread has still not started writing. Thus compactor can
// miss an entry as the oplog Id was not set till then.
// This is because a compactor thread will iterate over the entries &
// use only those which have OplogID equal to that of Oplog being
// compacted without taking any lock. A lock is taken only if the
// entry is a potential candidate.
// Further the compactor may delete the file as a compactor thread does
// not require to take any shared/exclusive lock at DiskStoreImpl
// or Oplog level.
// It is also assumed that compactor thread will take a lock on both
// entry as well as DiskID while compacting. In case of synch
// mode we can
// safely set OplogID without taking lock on DiskId. But
// for asynch mode
// we have to take additional precaution as the asynch
// writer of previous
// oplog can interfere with the current oplog.
id.setOplogId(getOplogId());
// do the io while holding lock so that switch can set doneAppending
// Write the data to the opLog for the synch mode
startPosForSynchOp = writeOpLogBytes(this.crf, async, true);
// if (this.crf.currSize != startPosForSynchOp) {
// assert false;
// }
this.crf.currSize = temp;
if (EntryBits.isNeedsValue(userBits)) {
id.setValueLength(value.getLength());
} else {
id.setValueLength(0);
}
id.setUserBits(userBits);
if (logger.isTraceEnabled()) {
logger.trace("Oplog::basicCreate:Release dByteBuffer with data for Disk ID = {}", id);
}
// As such for any put or get operation , a synch is taken
// on the Entry object in the DiskEntry's Helper functions.
// Compactor thread will also take a lock on entry object. Therefore
// we do not require a lock on DiskID, as concurrent access for
// value will not occur.
startPosForSynchOp += getOpStateValueOffset();
if (logger.isTraceEnabled(LogMarker.PERSIST_WRITES_VERBOSE)) {
VersionTag tag = null;
if (entry.getVersionStamp() != null) {
tag = entry.getVersionStamp().asVersionTag();
}
logger.trace(LogMarker.PERSIST_WRITES_VERBOSE,
"basicCreate: id=<{}> key=<{}> valueOffset={} userBits={} valueLen={} valueBytes={} drId={} versionTag={} oplog#{}",
abs(id.getKeyId()), entry.getKey(), startPosForSynchOp, userBits,
value.getLength(), value.getBytesAsString(), dr.getId(), tag,
getOplogId());
}
id.setOffsetInOplog(startPosForSynchOp);
addLive(dr, entry);
// Size of the current oplog being increased
// due to 'create' operation. Set the change in stats.
this.dirHolder.incrementTotalOplogSize(dataLength);
incTotalCount();
// Update the region version vector for the disk store.
// This needs to be done under lock so that we don't switch oplogs
// unit the version vector accurately represents what is in this oplog
RegionVersionVector rvv = dr.getRegionVersionVector();
if (rvv != null && entry.getVersionStamp() != null) {
rvv.recordVersion(entry.getVersionStamp().getMemberID(),
entry.getVersionStamp().getRegionVersion());
}
EntryLogger.logPersistPut(dr.getName(), entry.getKey(), dr.getDiskStoreID());
}
clearOpState();
// }
}
} finally {
getParent().getBackupLock().unlock();
}
if (useNextOplog) {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSwitchingOplog();
}
Assert.assertTrue(this != getOplogSet().getChild());
getOplogSet().getChild().basicCreate(dr, entry, value, userBits, async);
} else {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSettingOplogOffSet(startPosForSynchOp);
}
}
}
@Immutable
private static final byte[] TOMBSTONE_BYTES;
static {
try {
TOMBSTONE_BYTES = BlobHelper.serializeToBlob(Token.TOMBSTONE);
} catch (Exception e) {
throw new RuntimeException(e);
}
}
/**
* This oplog will be forced to switch to a new oplog
*/
void forceRolling(DiskRegion dr) {
if (getOplogSet().getChild() == this) {
getParent().getBackupLock().lock();
try {
synchronized (this.lock) {
if (getOplogSet().getChild() == this) {
switchOpLog(dr, 0, null);
}
}
} finally {
getParent().getBackupLock().unlock();
}
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSwitchingOplog();
}
}
}
/**
* Return true if it is possible that compaction of this oplog will be done.
*/
private boolean isCompactionPossible() {
return getOplogSet().isCompactionPossible();
}
/**
* This function is used to switch from one op Log to another , when the size of the current oplog
* has reached the maximum permissible. It is always called from synch block with lock object
* being the OpLog File object We will reuse the ByteBuffer Pool. We should add the current Oplog
* for compaction first & then try to get next directory holder as in case there is only a single
* directory with space being full, compaction has to happen before it can be given a new
* directory. If the operation causing the switching is on an Entry which already is referencing
* the oplog to be compacted, then the compactor thread will skip compaction that entry & the
* switching thread will roll the entry explicitly.
*
* @param lengthOfOperationCausingSwitch length of the operation causing the switch
* @param entryCausingSwitch DiskEntry object operation on which caused the switching of Oplog.
* This can be null if the switching has been invoked by the forceRolling which does not
* need an operation on entry to cause the switch
*/
private void switchOpLog(DiskRegionView dr, int lengthOfOperationCausingSwitch,
DiskEntry entryCausingSwitch) {
String drName;
if (dr != null) {
drName = dr.getName();
} else {
drName = getParent().getName();
}
flushAll(); // needed in case of async
lengthOfOperationCausingSwitch += 20; // for worstcase overhead of writing
// first record
// if length of operation is greater than max Dir Size than an exception is
// thrown
if (logger.isDebugEnabled()) {
logger.debug("Oplog::switchOpLog: Entry causing Oplog switch has diskID={}",
(entryCausingSwitch != null ? entryCausingSwitch.getDiskId() : "Entry is null"));
}
if (lengthOfOperationCausingSwitch > getParent().getMaxDirSize()) {
throw new DiskAccessException(
String.format(
"Operation size cannot exceed the maximum directory size. Switching problem for entry having DiskID=%s",
(entryCausingSwitch != null
? entryCausingSwitch.getDiskId().toString() : "\"null Entry\"")),
drName);
}
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().beforeSwitchingOplog();
}
if (logger.isDebugEnabled()) {
logger.debug(
"Oplog::switchOpLog: About to add the Oplog = {} for compaction. Entry causing the switch is having DiskID = {}",
this.oplogId,
(entryCausingSwitch != null ? entryCausingSwitch.getDiskId() : "null Entry"));
}
if (needsCompaction()) {
addToBeCompacted();
} else {
getOplogSet().addInactive(this);
}
try {
DirectoryHolder nextDirHolder =
getOplogSet().getNextDir(lengthOfOperationCausingSwitch, true);
Oplog newOplog = new Oplog(this.oplogId + 1, nextDirHolder, this);
newOplog.firstRecord = true;
getOplogSet().setChild(newOplog);
finishedAppending();
// Defer the unpreblow and close of the RAF. We saw pauses in testing from
// unpreblow of the drf, maybe because it is freeing pages that were
// preallocated. Close can pause if another thread is calling force on the
// file channel - see 50254. These operations should be safe to defer,
// a concurrent read will synchronize on the oplog and use or reopen the
// RAF
// as needed.
getParent().executeDelayedExpensiveWrite(new Runnable() {
@Override
public void run() {
// need to truncate crf and drf if their actual size is less than
// their pregrow size
unpreblow(Oplog.this.crf, getMaxCrfSize());
unpreblow(Oplog.this.drf, getMaxDrfSize());
// Close the crf using closeRAF. We will reopen the crf if we
// need it to fault in or to read values during compaction.
closeRAF();
// I think at this point the drf no longer needs to be open
synchronized (Oplog.this.lock/* drf */) {
if (!Oplog.this.drf.RAFClosed) {
try {
Oplog.this.drf.channel.close();
} catch (IOException ignore) {
}
try {
Oplog.this.drf.raf.close();
} catch (IOException ignore) {
}
Oplog.this.drf.RAFClosed = true;
}
}
}
});
// offline compaction will not execute create Krf in the task, becasue
// this.totalLiveCount.get() == 0
if (getParent().isOfflineCompacting()) {
krfClose();
} else {
createKrfAsync();
}
} catch (DiskAccessException dae) {
// Remove the Oplog which was added in the DiskStoreImpl
// for compaction as compaction cannot be done.
// However, it is also possible that compactor
// may have done the compaction of the Oplog but the switching thread
// encountered timeout exception.
// So the code below may or may not actually
// ensure that the Oplog has been compacted or not.
getOplogSet().removeOplog(this.oplogId);
throw dae;
}
}
/**
* Schedule a task to create a krf asynchronously
*/
protected void createKrfAsync() {
getParent().executeDiskStoreTask(new Runnable() {
@Override
public void run() {
createKrf(false);
}
});
}
private void writeOneKeyEntryForKRF(KRFEntry ke) throws IOException {
DiskEntry de = ke.getDiskEntry();
long diskRegionId = ke.getDiskRegionView().getId();
long oplogKeyId;
byte userBits;
long valueOffset;
int valueLength;
Object deKey;
VersionHolder tag = ke.versionTag;
synchronized (de) {
DiskId di = de.getDiskId();
if (di == null) {
return;
}
if (de.isRemovedFromDisk()) {
// the entry was concurrently removed
return;
}
synchronized (di) {
// Make sure each one is still in this oplog.
if (di.getOplogId() != getOplogId()) {
return;
}
userBits = di.getUserBits();
oplogKeyId = di.getKeyId();
valueOffset = di.getOffsetInOplog();
valueLength = di.getValueLength();
deKey = de.getKey();
if (valueOffset < 0) {
assert (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits));
}
}
if (tag == null) {
if (EntryBits.isWithVersions(userBits) && de.getVersionStamp() != null) {
tag = de.getVersionStamp().asVersionTag();
} else if (de.getVersionStamp() != null) {
throw new AssertionError("No version bits on entry we're writing to the krf " + de);
}
}
}
if (logger.isTraceEnabled(LogMarker.PERSIST_WRITES_VERBOSE)) {
logger.trace(LogMarker.PERSIST_WRITES_VERBOSE,
"krf oplogId={} key={} oplogKeyId={} de={} vo={} vl={} diskRegionId={} version tag={}",
oplogId, deKey, oplogKeyId, System.identityHashCode(de), valueOffset, valueLength,
diskRegionId, tag);
}
byte[] keyBytes = EntryEventImpl.serialize(deKey);
// skip the invalid entries, theire valueOffset is -1
writeOneKeyEntryForKRF(keyBytes, userBits, valueLength, diskRegionId, oplogKeyId, valueOffset,
tag);
}
private void writeOneKeyEntryForKRF(byte[] keyBytes, byte userBits, int valueLength,
long diskRegionId, long oplogKeyId, long valueOffset, VersionHolder tag) throws IOException {
if (getParent().isValidating()) {
return;
}
if (!getParent().isOfflineCompacting()) {
assert (this.krf.dos != null);
} else {
if (this.krf.dos == null) {
// krf already exist, thus not re-opened for write
return;
}
}
DataSerializer.writeByteArray(keyBytes, this.krf.dos);
this.krf.dos.writeByte(EntryBits.getPersistentBits(userBits));
InternalDataSerializer.writeArrayLength(valueLength, this.krf.dos);
DiskInitFile.writeDiskRegionID(this.krf.dos, diskRegionId);
if (EntryBits.isWithVersions(userBits) && tag != null) {
serializeVersionTag(tag, this.krf.dos);
}
InternalDataSerializer.writeVLOld(oplogKeyId, this.krf.dos);
// skip the invalid entries, theire valueOffset is -1
if (!EntryBits.isAnyInvalid(userBits) && !EntryBits.isTombstone(userBits)) {
InternalDataSerializer.writeVLOld((valueOffset - this.krf.lastOffset), this.krf.dos);
// save the lastOffset in krf object
this.krf.lastOffset = valueOffset;
}
this.krf.keyNum++;
}
private final AtomicBoolean krfCreated = new AtomicBoolean();
public void krfFileCreate() throws IOException {
// this method is only used by offline compaction. validating will not
// create krf
assert (getParent().isValidating() == false);
this.krf.f = new File(this.diskFile.getPath() + KRF_FILE_EXT);
if (this.krf.f.exists()) {
throw new IllegalStateException("krf file " + this.krf.f + " already exists.");
}
this.krf.fos = new FileOutputStream(this.krf.f);
this.krf.bos = new BufferedOutputStream(this.krf.fos, 32768);
this.krf.dos = new DataOutputStream(this.krf.bos);
// write oplog magic seq
this.krf.dos.writeByte(OPLOG_MAGIC_SEQ_ID);
this.krf.dos.write(Oplog.OPLOG_TYPE.KRF.getBytes(), 0, Oplog.OPLOG_TYPE.getLen());
this.krf.dos.writeByte(END_OF_RECORD_ID);
// write the disk store id to the krf
this.krf.dos.writeByte(OPLOG_DISK_STORE_ID);
this.krf.dos.writeLong(getParent().getDiskStoreID().getLeastSignificantBits());
this.krf.dos.writeLong(getParent().getDiskStoreID().getMostSignificantBits());
this.krf.dos.writeByte(END_OF_RECORD_ID);
// write product versions
assert this.gfversion != null;
// write both gemfire and data versions if the two are different else write
// only gemfire version; a token distinguishes the two cases while reading
// like in writeGemFireVersionRecord
KnownVersion dataVersion = getDataVersionIfOld();
if (dataVersion == null) {
dataVersion = KnownVersion.CURRENT;
}
if (this.gfversion == dataVersion) {
VersioningIO.writeOrdinal(this.krf.dos, this.gfversion.ordinal(), false);
} else {
VersioningIO.writeOrdinal(this.krf.dos, KnownVersion.TOKEN.ordinal(), false);
this.krf.dos.writeByte(END_OF_RECORD_ID);
this.krf.dos.writeByte(OPLOG_GEMFIRE_VERSION);
VersioningIO.writeOrdinal(this.krf.dos, this.gfversion.ordinal(), false);
this.krf.dos.writeByte(END_OF_RECORD_ID);
this.krf.dos.writeByte(OPLOG_GEMFIRE_VERSION);
VersioningIO.writeOrdinal(this.krf.dos, dataVersion.ordinal(), false);
}
this.krf.dos.writeByte(END_OF_RECORD_ID);
// Write the total entry count to the krf so that when we recover,
// our compaction statistics will be accurate
InternalDataSerializer.writeUnsignedVL(this.totalCount.get(), this.krf.dos);
this.krf.dos.writeByte(END_OF_RECORD_ID);
// Write the RVV to the krf.
Map<Long, AbstractDiskRegion> drMap = getParent().getAllDiskRegions();
byte[] rvvBytes = serializeRVVs(drMap, false);
this.krf.dos.write(rvvBytes);
this.krf.dos.writeByte(END_OF_RECORD_ID);
}
// if IOException happened during krf creation, close and delete it
private void closeAndDeleteKrf() {
try {
if (this.krf.dos != null) {
this.krf.dos.close();
this.krf.dos = null;
}
} catch (IOException ignore) {
}
try {
if (this.krf.bos != null) {
this.krf.bos.close();
this.krf.bos = null;
}
} catch (IOException ignore) {
}
try {
if (this.krf.fos != null) {
this.krf.fos.close();
this.krf.fos = null;
}
} catch (IOException ignore) {
}
if (this.krf.f.exists()) {
this.krf.f.delete();
}
}
public void krfClose() {
boolean allClosed = false;
try {
if (this.krf.fos != null) {
DataSerializer.writeByteArray(null, this.krf.dos);
} else {
return;
}
this.krf.dos.flush();
this.krf.fos.getChannel().force(true);
this.krf.dos.close();
this.krf.dos = null;
this.krf.bos.close();
this.krf.bos = null;
this.krf.fos.close();
this.krf.fos = null;
if (this.krf.keyNum == 0) {
// this is an empty krf file
this.krf.f.delete();
assert this.krf.f.exists() == false;
} else {
// Mark that this krf is complete.
getParent().getDiskInitFile().krfCreate(this.oplogId);
logger.info("Created {} {} for disk store {}.",
new Object[] {toString(), "krf", getParent().getName()});
}
allClosed = true;
} catch (IOException e) {
if (getParent().getDiskAccessException() == null) {
throw new DiskAccessException("Fail to close krf file " + this.krf.f, e, getParent());
} else {
logger.info(
"Fail to close krf file " + this.krf.f + ", but a DiskAccessException happened ealier",
getParent().getDiskAccessException());
}
} finally {
if (!allClosed) {
// IOException happened during close, delete this krf
closeAndDeleteKrf();
}
}
}
/**
* Create the KRF file for this oplog. It is ok for this method to be async. finishKRF will be
* called and it must block until KRF generation is complete.
*
* @param cancel if true then prevent the krf from being created if possible
*/
void createKrf(boolean cancel) {
if (cancel) {
this.krfCreated.compareAndSet(false, true);
return;
}
if (!couldHaveKrf()) {
return;
}
// Make sure regions can not become unrecovered while creating the KRF.
getParent().acquireCompactorReadLock();
try {
if (!getParent().allowKrfCreation()) {
return;
}
lockCompactor();
try {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock) {
// 42733: after set it to true, we will not reset it, since this oplog
// will be
// inactive forever
this.lockedForKRFcreate = true;
}
synchronized (this.krfCreated) {
if (this.krfCreated.get()) {
return;
}
this.krfCreated.set(true);
if (this.unrecoveredRegionCount.get() > 0) {
// if we have unrecovered regions then we can't create
// a KRF because we don't have the list of live entries.
return;
}
int tlc = (int) this.totalLiveCount.get();
if (tlc <= 0) {
// no need to create a KRF since this oplog will be deleted.
// TODO should we create an empty KRF anyway?
return;
}
Collection<DiskRegionInfo> regions = this.regionMap.values();
List<KRFEntry> sortedLiveEntries = getSortedLiveEntries(regions);
if (sortedLiveEntries == null) {
// no need to create a krf if there are no live entries.
return;
}
boolean krfCreateSuccess = false;
try {
krfFileCreate();
// sortedLiveEntries are now sorted
// so we can start writing them to disk.
for (KRFEntry ke : sortedLiveEntries) {
writeOneKeyEntryForKRF(ke);
}
krfClose();
krfCreateSuccess = true;
for (DiskRegionInfo dri : regions) {
dri.afterKrfCreated();
}
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterKrfCreated();
}
} catch (FileNotFoundException ex) {
// we couldn't open the krf file
throw new IllegalStateException("could not create krf " + this.krf.f, ex);
} catch (IOException ex) {
// we failed to write to the file
throw new IllegalStateException("failed writing krf " + this.krf.f, ex);
} finally {
// if IOException happened in writeOneKeyEntryForKRF(), delete krf here
if (!krfCreateSuccess) {
closeAndDeleteKrf();
}
}
}
} finally {
unlockCompactor();
}
} finally {
getParent().releaseCompactorReadLock();
}
}
public File getKrfFile() {
return new File(this.diskFile.getPath() + KRF_FILE_EXT);
}
public List<KRFEntry> getSortedLiveEntries(Collection<DiskRegionInfo> targetRegions) {
int tlc = (int) this.totalLiveCount.get();
if (tlc <= 0) {
// no need to create a KRF since this oplog will be deleted.
// TODO should we create an empty KRF anyway?
return null;
}
KRFEntry[] sortedLiveEntries = new KRFEntry[tlc];
int idx = 0;
for (DiskRegionInfo dri : targetRegions) {
if (dri.getDiskRegion() != null) {
idx = dri.addLiveEntriesToList(sortedLiveEntries, idx);
}
}
// idx is now the length of sortedLiveEntries
Arrays.sort(sortedLiveEntries, 0, idx, new Comparator<KRFEntry>() {
@Override
public int compare(KRFEntry o1, KRFEntry o2) {
long val1 = o1.getOffsetInOplogForSorting();
long val2 = o2.getOffsetInOplogForSorting();
return Long.signum(val1 - val2);
}
});
return Arrays.asList(sortedLiveEntries).subList(0, idx);
}
/**
* This function retrieves the value for an entry being compacted subject to entry referencing the
* oplog being compacted. Attempt is made to retrieve the value from in memory , if available,
* else from asynch buffers ( if asynch mode is enabled), else from the Oplog being compacted. It
* is invoked from switchOplog as well as OplogCompactor's compact function.
*
* @param entry DiskEntry being compacted referencing the Oplog being compacted
* @param wrapper Object of type BytesAndBitsForCompactor. The data if found is set in the wrapper
* Object. The wrapper Object also contains the user bit associated with the entry
* @return boolean false indicating that entry need not be compacted. If true it means that
* wrapper has been appropriately filled with data
*/
private boolean getBytesAndBitsForCompaction(DiskRegionView dr, DiskEntry entry,
BytesAndBitsForCompactor wrapper) {
// caller is synced on did
DiskId did = entry.getDiskId();
byte userBits = 0;
long oplogOffset = did.getOffsetInOplog();
ReferenceCountHelper.skipRefCountTracking();
@Retained
@Released
Object value = entry.getValueRetain(dr, true);
ReferenceCountHelper.unskipRefCountTracking();
boolean foundData = false;
if (value == null) {
// If the mode is synch it is guaranteed to be present in the disk
foundData = basicGetForCompactor(dr, oplogOffset, false, did.getValueLength(),
did.getUserBits(), wrapper);
// after we have done the get do one more check to see if the
// disk id of interest is still stored in the current oplog.
// Do this to fix bug 40648
// Since we now call this with the diskId synced I think
// it is impossible for this oplogId to change.
if (did.getOplogId() != getOplogId()) {
// if it is not then no need to compact it
return false;
} else {
// if the disk id indicates its most recent value is in oplogInFocus
// then we should have found data
assert foundData : "compactor get failed on oplog#" + getOplogId();
}
userBits = wrapper.getBits();
if (EntryBits.isAnyInvalid(userBits)) {
if (EntryBits.isInvalid(userBits)) {
wrapper.setData(DiskEntry.INVALID_BYTES, userBits, DiskEntry.INVALID_BYTES.length,
false/*
* Can not be reused
*/);
} else {
wrapper.setData(DiskEntry.LOCAL_INVALID_BYTES, userBits,
DiskEntry.LOCAL_INVALID_BYTES.length, false/*
* Can not be reused
*/);
}
} else if (EntryBits.isTombstone(userBits)) {
wrapper.setData(DiskEntry.TOMBSTONE_BYTES, userBits, DiskEntry.TOMBSTONE_BYTES.length,
false/*
* Can not be reused
*/);
}
if (EntryBits.isWithVersions(did.getUserBits())) {
userBits = EntryBits.setWithVersions(userBits, true);
}
} else {
foundData = true;
userBits = 0;
if (EntryBits.isRecoveredFromDisk(did.getUserBits())) {
userBits = EntryBits.setRecoveredFromDisk(userBits, true);
}
if (EntryBits.isWithVersions(did.getUserBits())) {
userBits = EntryBits.setWithVersions(userBits, true);
}
// no need to preserve pendingAsync bit since we will clear it anyway
// since we
// (the compactor) are writing the value out to disk.
if (value == Token.INVALID) {
userBits = EntryBits.setInvalid(userBits, true);
wrapper.setData(DiskEntry.INVALID_BYTES, userBits, DiskEntry.INVALID_BYTES.length,
false /*
* Cannot be reused
*/);
} else if (value == Token.LOCAL_INVALID) {
userBits = EntryBits.setLocalInvalid(userBits, true);
wrapper.setData(DiskEntry.LOCAL_INVALID_BYTES, userBits,
DiskEntry.LOCAL_INVALID_BYTES.length, false /*
* Cannot be reused
*/);
} else if (value == Token.TOMBSTONE) {
userBits = EntryBits.setTombstone(userBits, true);
wrapper.setData(DiskEntry.TOMBSTONE_BYTES, userBits, DiskEntry.TOMBSTONE_BYTES.length,
false /*
* Cannot be reused
*/);
} else if (value instanceof CachedDeserializable) {
CachedDeserializable proxy = (CachedDeserializable) value;
if (proxy instanceof StoredObject) {
@Released
StoredObject ohproxy = (StoredObject) proxy;
try {
ohproxy.fillSerializedValue(wrapper, userBits);
} finally {
OffHeapHelper.releaseWithNoTracking(ohproxy);
}
} else {
userBits = EntryBits.setSerialized(userBits, true);
proxy.fillSerializedValue(wrapper, userBits);
}
} else if (value instanceof byte[]) {
byte[] valueBytes = (byte[]) value;
// If the value is already a byte array then the user bit
// is 0, which is the default value of the userBits variable,
// indicating that it is non serialized data. Thus it is
// to be used as it is & not to be deserialized to
// convert into Object
wrapper.setData(valueBytes, userBits, valueBytes.length,
false /*
* the wrapper is not reusable
*/);
} else if (Token.isRemoved(value) && value != Token.TOMBSTONE) {
// TODO - RVV - We need to handle tombstones differently here!
if (entry.getDiskId().isPendingAsync()) {
entry.getDiskId().setPendingAsync(false);
try {
getOplogSet().getChild().basicRemove(dr, entry, false, false);
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format("Failed writing key to %s",
this.diskFile.getPath()), ex, dr.getName());
} catch (InterruptedException ie) {
Thread.currentThread().interrupt();
getParent().getCache().getCancelCriterion().checkCancelInProgress(ie);
throw new DiskAccessException(
String.format(
"Failed writing key to %s due to failure in acquiring read lock for asynch writing",
this.diskFile.getPath()),
ie, dr.getName());
}
} else {
rmLive(dr, entry);
}
foundData = false;
} else {
userBits = EntryBits.setSerialized(userBits, true);
EntryEventImpl.fillSerializedValue(wrapper, value, userBits);
}
}
if (foundData) {
// since the compactor is writing it out clear the async flag
entry.getDiskId().setPendingAsync(false);
}
return foundData;
}
/**
* Modifies a key/value pair from a region entry on disk. Updates all of the necessary
* {@linkplain DiskStoreStats statistics} and invokes basicModify
* <p>
* Modified the code so as to reuse the already created ByteBuffer during transition. Minimizing
* the synchronization allowing multiple put operations for different entries to proceed
* concurrently for asynch mode
*
* @param entry DiskEntry object representing the current Entry
*
* @param value byte array representing the value
*/
public void modify(InternalRegion region, DiskEntry entry, ValueWrapper value, boolean async) {
if (getOplogSet().getChild() != this) {
getOplogSet().getChild().modify(region, entry, value, async);
} else {
DiskId did = entry.getDiskId();
boolean exceptionOccurred = false;
byte prevUsrBit = did.getUserBits();
int len = did.getValueLength();
try {
byte userBits = calcUserBits(value);
// save versions for creates and updates even if value is bytearrary in
// 7.0
if (entry.getVersionStamp() != null) {
if (entry.getVersionStamp().getMemberID() == null) {
throw new AssertionError(
"Version stamp should have a member at this point for entry " + entry);
}
// pdx and tx will not use version
userBits = EntryBits.setWithVersions(userBits, true);
}
basicModify(region.getDiskRegion(), entry, value, userBits, async, false);
} catch (IOException ex) {
exceptionOccurred = true;
region.getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format("Failed writing key to %s",
this.diskFile.getPath()), ex, region.getFullPath());
} catch (InterruptedException ie) {
Thread.currentThread().interrupt();
exceptionOccurred = true;
region.getCancelCriterion().checkCancelInProgress(ie);
throw new DiskAccessException(
String.format(
"Failed writing key to %s due to failure in acquiring read lock for asynch writing",
this.diskFile.getPath()),
ie, region.getFullPath());
} finally {
if (exceptionOccurred) {
did.setValueLength(len);
did.setUserBits(prevUsrBit);
}
}
}
}
public void offlineModify(DiskRegionView drv, DiskEntry entry, byte[] value,
boolean isSerializedObject) {
try {
ValueWrapper vw = new DiskEntry.Helper.ByteArrayValueWrapper(isSerializedObject, value);
byte userBits = calcUserBits(vw);
// save versions for creates and updates even if value is bytearrary in 7.0
VersionStamp vs = entry.getVersionStamp();
if (vs != null) {
if (vs.getMemberID() == null) {
throw new AssertionError(
"Version stamp should have a member at this point for entry " + entry);
}
// Since we are modifying this entry's value while offline make sure its version stamp
// has this disk store as its member id and bump the version
vs.setMemberID(getParent().getDiskStoreID());
VersionTag vt = vs.asVersionTag();
vt.setRegionVersion(drv.getRegionVersionVector().getNextVersion());
vt.setEntryVersion(vt.getEntryVersion() + 1);
vt.setVersionTimeStamp(System.currentTimeMillis());
vs.setVersions(vt);
userBits = EntryBits.setWithVersions(userBits, true);
}
basicModify(drv, entry, vw, userBits, false, false);
} catch (IOException ex) {
throw new DiskAccessException(
String.format("Failed writing key to %s", this.diskFile.getPath()),
ex, drv.getName());
} catch (InterruptedException ie) {
Thread.currentThread().interrupt();
throw new DiskAccessException(
String.format(
"Failed writing key to %s due to failure in acquiring read lock for asynch writing",
this.diskFile.getPath()),
ie, drv.getName());
}
}
public void saveConflictVersionTag(InternalRegion region, VersionTag tag, boolean async) {
if (getOplogSet().getChild() != this) {
getOplogSet().getChild().saveConflictVersionTag(region, tag, async);
} else {
try {
basicSaveConflictVersionTag(region.getDiskRegion(), tag, async);
} catch (IOException ex) {
region.getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format("Failed writing conflict version tag to %s",
this.diskFile.getPath()), ex, region.getFullPath());
} catch (InterruptedException ie) {
Thread.currentThread().interrupt();
region.getCancelCriterion().checkCancelInProgress(ie);
throw new DiskAccessException(String.format("Failed writing conflict version tag to %s",
this.diskFile.getPath()), ie, region.getFullPath());
}
}
}
private void copyForwardForOfflineCompact(long oplogKeyId, byte[] keyBytes, byte[] valueBytes,
byte userBits, long drId, VersionTag tag) {
try {
basicCopyForwardForOfflineCompact(oplogKeyId, keyBytes, valueBytes, userBits, drId, tag);
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(
String.format("Failed writing key to %s", this.diskFile.getPath()),
ex, getParent());
} catch (InterruptedException ie) {
Thread.currentThread().interrupt();
getParent().getCancelCriterion().checkCancelInProgress(ie);
throw new DiskAccessException(
String.format(
"Failed writing key to %s due to failure in acquiring read lock for asynch writing",
this.diskFile.getPath()),
ie, getParent());
}
}
private void copyForwardModifyForCompact(DiskRegionView dr, DiskEntry entry,
BytesAndBitsForCompactor wrapper) {
if (getOplogSet().getChild() != this) {
getOplogSet().getChild().copyForwardModifyForCompact(dr, entry, wrapper);
} else {
DiskId did = entry.getDiskId();
boolean exceptionOccurred = false;
int len = did.getValueLength();
try {
// TODO: compaction needs to get version?
byte userBits = wrapper.getBits();
ValueWrapper vw;
if (wrapper.getOffHeapData() != null) {
vw = new DiskEntry.Helper.OffHeapValueWrapper(wrapper.getOffHeapData());
} else {
vw = new DiskEntry.Helper.CompactorValueWrapper(wrapper.getBytes(),
wrapper.getValidLength());
}
// Compactor always says to do an async basicModify so that its writes
// will be grouped. This is not a true async write; just a grouped one.
basicModify(dr, entry, vw, userBits, true, true);
} catch (IOException ex) {
exceptionOccurred = true;
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format("Failed writing key to %s",
this.diskFile.getPath()), ex, getParent());
} catch (InterruptedException ie) {
exceptionOccurred = true;
Thread.currentThread().interrupt();
getParent().getCancelCriterion().checkCancelInProgress(ie);
throw new DiskAccessException(
String.format(
"Failed writing key to %s due to failure in acquiring read lock for asynch writing",
this.diskFile.getPath()),
ie, getParent());
} finally {
if (wrapper.getOffHeapData() != null) {
wrapper.setOffHeapData(null, (byte) 0);
}
if (exceptionOccurred) {
did.setValueLength(len);
}
}
}
}
/**
* A helper function which identifies whether to modify the entry in the current oplog or to make
* the switch to the next oplog. This function enables us to reuse the byte buffer which got
* created for an oplog which no longer permits us to use itself. It will also take acre of
* compaction if required
*
* @param entry DiskEntry object representing the current Entry
*/
private void basicModify(DiskRegionView dr, DiskEntry entry, ValueWrapper value, byte userBits,
boolean async, boolean calledByCompactor) throws IOException, InterruptedException {
DiskId id = entry.getDiskId();
boolean useNextOplog = false;
long startPosForSynchOp = -1L;
int adjustment = 0;
Oplog emptyOplog = null;
if (DiskStoreImpl.KRF_DEBUG) {
// wait for cache close to create krf
System.out.println("basicModify KRF_DEBUG");
Thread.sleep(1000);
}
getParent().getBackupLock().lock();
try {
synchronized (this.lock) {
// synchronized (this.crf) {
if (getOplogSet().getChild() != this) {
useNextOplog = true;
} else {
initOpState(OPLOG_MOD_ENTRY_1ID, dr, entry, value, userBits, false);
adjustment = getOpStateSize();
assert adjustment > 0;
long temp = (this.crf.currSize + adjustment);
if (temp > getMaxCrfSize() && !isFirstRecord()) {
switchOpLog(dr, adjustment, entry);
// we can't reuse it since it contains variable length data
useNextOplog = true;
} else {
if (this.lockedForKRFcreate) {
CacheClosedException cce =
getParent().getCache().getCacheClosedException("The disk store is closed.");
dr.getCancelCriterion().checkCancelInProgress(cce);
throw cce;
}
this.firstRecord = false;
long oldOplogId;
// do the io while holding lock so that switch can set doneAppending
// Write the data to the opLog for the synch mode
startPosForSynchOp = writeOpLogBytes(this.crf, async, true);
this.crf.currSize = temp;
startPosForSynchOp += getOpStateValueOffset();
if (logger.isTraceEnabled(LogMarker.PERSIST_WRITES_VERBOSE)) {
VersionTag tag = null;
if (entry.getVersionStamp() != null) {
tag = entry.getVersionStamp().asVersionTag();
}
logger.trace(LogMarker.PERSIST_WRITES_VERBOSE,
"basicModify: id=<{}> key=<{}> valueOffset={} userBits={} valueLen={} valueBytes=<{}> drId={} versionStamp={} oplog#{}",
abs(id.getKeyId()), entry.getKey(), startPosForSynchOp, userBits,
value.getLength(), value.getBytesAsString(), dr.getId(), tag, getOplogId());
}
if (EntryBits.isNeedsValue(userBits)) {
id.setValueLength(value.getLength());
} else {
id.setValueLength(0);
}
id.setUserBits(userBits);
if (logger.isTraceEnabled()) {
logger.trace("Oplog::basicModify:Released ByteBuffer with data for Disk ID = {}", id);
}
synchronized (id) {
// Need to do this while synced on id
// now that we compact forward to most recent oplog.
// @todo darrel: The sync logic in the disk code is so complex
// a really doubt is is correct.
// I think we need to do a fresh rewrite of it.
oldOplogId = id.setOplogId(getOplogId());
if (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits)) {
id.setOffsetInOplog(-1);
} else {
id.setOffsetInOplog(startPosForSynchOp);
}
}
// Set the oplog size change for stats
this.dirHolder.incrementTotalOplogSize(adjustment);
this.incTotalCount();
EntryLogger.logPersistPut(dr.getName(), entry.getKey(), dr.getDiskStoreID());
if (oldOplogId != getOplogId()) {
Oplog oldOplog = getOplogSet().getChild(oldOplogId);
if (oldOplog != null) {
oldOplog.rmLive(dr, entry);
emptyOplog = oldOplog;
}
addLive(dr, entry);
// Note if this mod was done to oldOplog then this entry is already
// in
// the linked list. All we needed to do in this case is call
// incTotalCount
} else {
getOrCreateDRI(dr).update(entry);
}
// Update the region version vector for the disk store.
// This needs to be done under lock so that we don't switch oplogs
// unit the version vector accurately represents what is in this oplog
RegionVersionVector rvv = dr.getRegionVersionVector();
if (rvv != null && entry.getVersionStamp() != null) {
rvv.recordVersion(entry.getVersionStamp().getMemberID(),
entry.getVersionStamp().getRegionVersion());
}
}
clearOpState();
}
// }
}
} finally {
getParent().getBackupLock().unlock();
}
if (useNextOplog) {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSwitchingOplog();
}
Assert.assertTrue(getOplogSet().getChild() != this);
getOplogSet().getChild().basicModify(dr, entry, value, userBits, async, calledByCompactor);
} else {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSettingOplogOffSet(startPosForSynchOp);
}
if (emptyOplog != null
&& (!emptyOplog.isCompacting() || emptyOplog.calledByCompactorThread())) {
if (calledByCompactor && emptyOplog.hasNoLiveValues()) {
// Since compactor will only append to crf no need to flush drf.
// Before we have the compactor delete an oplog it has emptied out
// we want to have it flush anything it has written to the current
// oplog.
// Note that since sync writes may be done to the same oplog we are
// doing
// async writes to any sync writes will cause a flush to be done
// immediately.
flushAll(true);
}
emptyOplog.handleNoLiveValues();
}
}
}
private void basicSaveConflictVersionTag(DiskRegionView dr, VersionTag tag, boolean async)
throws IOException, InterruptedException {
boolean useNextOplog = false;
int adjustment = 0;
getParent().getBackupLock().lock();
try {
synchronized (this.lock) {
if (getOplogSet().getChild() != this) {
useNextOplog = true;
} else {
this.opState.initialize(OPLOG_CONFLICT_VERSION, dr.getId(), tag);
adjustment = getOpStateSize();
assert adjustment > 0;
long temp = (this.crf.currSize + adjustment);
if (temp > getMaxCrfSize() && !isFirstRecord()) {
switchOpLog(dr, adjustment, null);
// we can't reuse it since it contains variable length data
useNextOplog = true;
} else {
if (this.lockedForKRFcreate) {
CacheClosedException cce =
getParent().getCache().getCacheClosedException("The disk store is closed.");
dr.getCancelCriterion().checkCancelInProgress(cce);
throw cce;
}
this.firstRecord = false;
writeOpLogBytes(this.crf, async, true);
this.crf.currSize = temp;
if (logger.isTraceEnabled(LogMarker.PERSIST_WRITES_VERBOSE)) {
logger.trace(LogMarker.PERSIST_WRITES_VERBOSE,
"basicSaveConflictVersionTag: drId={} versionStamp={} oplog#{}", dr.getId(), tag,
getOplogId());
}
this.dirHolder.incrementTotalOplogSize(adjustment);
// Update the region version vector for the disk store.
// This needs to be done under lock so that we don't switch oplogs
// unit the version vector accurately represents what is in this oplog
RegionVersionVector rvv = dr.getRegionVersionVector();
if (rvv != null && dr.getFlags().contains(DiskRegionFlag.IS_WITH_VERSIONING)) {
rvv.recordVersion(tag.getMemberID(), tag.getRegionVersion());
}
}
clearOpState();
}
}
} finally {
getParent().getBackupLock().unlock();
}
if (useNextOplog) {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSwitchingOplog();
}
Assert.assertTrue(getOplogSet().getChild() != this);
getOplogSet().getChild().basicSaveConflictVersionTag(dr, tag, async);
}
}
private void basicCopyForwardForOfflineCompact(long oplogKeyId, byte[] keyBytes,
byte[] valueBytes, byte userBits, long drId, VersionTag tag)
throws IOException, InterruptedException {
boolean useNextOplog = false;
long startPosForSynchOp = -1L;
int adjustment = 0;
// No need to get the backup lock since this is only for offline compaction
synchronized (this.lock) {
// synchronized (this.crf) {
if (getOplogSet().getChild() != this) {
useNextOplog = true;
} else {
this.opState.initialize(oplogKeyId, keyBytes, valueBytes, userBits, drId, tag, false);
adjustment = getOpStateSize();
assert adjustment > 0;
long temp = (this.crf.currSize + adjustment);
if (temp > getMaxCrfSize() && !isFirstRecord()) {
switchOpLog(null, adjustment, null);
// we can't reuse it since it contains variable length data
useNextOplog = true;
} else {
this.firstRecord = false;
// do the io while holding lock so that switch can set doneAppending
// Write the data to the opLog async since we are offline compacting
startPosForSynchOp = writeOpLogBytes(this.crf, true, true);
this.crf.currSize = temp;
startPosForSynchOp += getOpStateValueOffset();
getOplogSet().getChild().writeOneKeyEntryForKRF(keyBytes, userBits, valueBytes.length,
drId, oplogKeyId, startPosForSynchOp, tag);
if (logger.isTraceEnabled(LogMarker.PERSIST_WRITES_VERBOSE)) {
logger.trace(LogMarker.PERSIST_WRITES_VERBOSE,
"basicCopyForwardForOfflineCompact: id=<{}> keyBytes=<{}> valueOffset={} userBits={} valueLen={} valueBytes=<{}> drId={} oplog#{}",
oplogKeyId, baToString(keyBytes), startPosForSynchOp, userBits, valueBytes.length,
baToString(valueBytes), drId, getOplogId());
}
this.dirHolder.incrementTotalOplogSize(adjustment);
this.incTotalCount();
}
clearOpState();
}
}
if (useNextOplog) {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSwitchingOplog();
}
Assert.assertTrue(getOplogSet().getChild() != this);
getOplogSet().getChild().basicCopyForwardForOfflineCompact(oplogKeyId, keyBytes, valueBytes,
userBits, drId, tag);
}
}
private boolean isCompacting() {
return this.compacting;
}
private void addLive(DiskRegionView dr, DiskEntry de) {
getOrCreateDRI(dr).addLive(de);
incLiveCount();
}
private void rmLive(DiskRegionView dr, DiskEntry de) {
DiskRegionInfo dri = getOrCreateDRI(dr);
synchronized (dri) {
dri.rmLive(de, this);
}
}
private DiskRegionInfo getDRI(long drId) {
return this.regionMap.get(drId);
}
private DiskRegionInfo getDRI(DiskRegionView dr) {
return getDRI(dr.getId());
}
private DiskRegionInfo getOrCreateDRI(DiskRegionView dr) {
DiskRegionInfo dri = getDRI(dr);
if (dri == null) {
dri = (isCompactionPossible() || couldHaveKrf())
? new DiskRegionInfoWithList(dr, couldHaveKrf(), this.krfCreated.get())
: new DiskRegionInfoNoList(dr);
DiskRegionInfo oldDri = this.regionMap.putIfAbsent(dr.getId(), dri);
if (oldDri != null) {
dri = oldDri;
}
}
return dri;
}
public boolean needsKrf() {
return couldHaveKrf() && !krfCreated.get();
}
/**
* @return true if this Oplog could end up having a KRF file.
*/
private boolean couldHaveKrf() {
return getOplogSet().couldHaveKrf();
}
private DiskRegionInfo getOrCreateDRI(long drId) {
DiskRegionInfo dri = getDRI(drId);
if (dri == null) {
dri = (isCompactionPossible() || couldHaveKrf())
? new DiskRegionInfoWithList(null, couldHaveKrf(), this.krfCreated.get())
: new DiskRegionInfoNoList(null);
DiskRegionInfo oldDri = this.regionMap.putIfAbsent(drId, dri);
if (oldDri != null) {
dri = oldDri;
}
}
return dri;
}
/**
* Removes the key/value pair with the given id on disk.
*
* @param entry DiskEntry object on which remove operation is called
*/
public void remove(InternalRegion region, DiskEntry entry, boolean async, boolean isClear) {
DiskRegion dr = region.getDiskRegion();
if (getOplogSet().getChild() != this) {
getOplogSet().getChild().remove(region, entry, async, isClear);
} else {
DiskId did = entry.getDiskId();
boolean exceptionOccurred = false;
byte prevUsrBit = did.getUserBits();
int len = did.getValueLength();
try {
basicRemove(dr, entry, async, isClear);
} catch (IOException ex) {
exceptionOccurred = true;
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format("Failed writing key to %s",
this.diskFile.getPath()), ex, dr.getName());
} catch (InterruptedException ie) {
Thread.currentThread().interrupt();
region.getCancelCriterion().checkCancelInProgress(ie);
exceptionOccurred = true;
throw new DiskAccessException(
String.format(
"Failed writing key to %s due to failure in acquiring read lock for asynch writing",
this.diskFile.getPath()),
ie, dr.getName());
} finally {
if (exceptionOccurred) {
did.setValueLength(len);
did.setUserBits(prevUsrBit);
}
}
}
}
/**
* Write the GC RVV for a single region to disk
*/
public void writeGCRVV(DiskRegion dr) {
boolean useNextOplog = false;
getParent().getBackupLock().lock();
try {
synchronized (this.lock) {
if (getOplogSet().getChild() != this) {
useNextOplog = true;
} else {
try {
writeRVVRecord(this.drf,
Collections.<Long, AbstractDiskRegion>singletonMap(dr.getId(), dr), true);
} catch (IOException ex) {
dr.getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format(
"Failed in persisting the garbage collection of entries because of: %s",
this.diskFile.getPath()), ex, dr.getName());
}
}
}
} finally {
getParent().getBackupLock().unlock();
}
if (useNextOplog) {
getOplogSet().getChild().writeGCRVV(dr);
} else {
DiskStoreObserver.endWriteGCRVV(dr);
}
}
/**
* There're 3 cases to use writeRVV: 1) endGII: DiskRegion.writeRVV(region=null, true),
* Oplog.writeRVV(true,null) 2) beginGII: DiskRegion.writeRVV(region=this, false),
* Oplog.writeRVV(false,sourceRVV!=null) 3) clear: DiskRegion.writeRVV(region=this, null),
* Oplog.writeRVV(null,sourceRVV!=null)
*/
public void writeRVV(DiskRegion dr, RegionVersionVector sourceRVV, Boolean isRVVTrusted) {
boolean useNextOplog = false;
getParent().getBackupLock().lock();
try {
synchronized (this.lock) {
if (getOplogSet().getChild() != this) {
useNextOplog = true;
} else {
try {
// We'll update the RVV of the disk region while holding the lock on
// the oplog,
// to make sure we don't switch oplogs while we're in the middle of
// this.
if (sourceRVV != null) {
dr.getRegionVersionVector().recordVersions(sourceRVV);
} else {
// it's original EndGII, not to write duplicate rvv if its trusted
if (dr.getRVVTrusted()) {
return;
}
}
if (isRVVTrusted != null) {
// isRVVTrusted == null means "as is"
dr.setRVVTrusted(isRVVTrusted);
}
writeRVVRecord(this.crf,
Collections.<Long, AbstractDiskRegion>singletonMap(dr.getId(), dr), false);
} catch (IOException ex) {
dr.getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(String.format(
"Failed in persisting the garbage collection of entries because of: %s",
this.diskFile.getPath()), ex, dr.getName());
}
}
}
} finally {
getParent().getBackupLock().unlock();
}
if (useNextOplog) {
getOplogSet().getChild().writeRVV(dr, sourceRVV, isRVVTrusted);
}
}
private long getMaxCrfSize() {
return this.maxCrfSize;
}
private long getMaxDrfSize() {
return this.maxDrfSize;
}
private void setMaxCrfDrfSize() {
int crfPct = Integer.getInteger(GeodeGlossary.GEMFIRE_PREFIX + "CRF_MAX_PERCENTAGE", 90);
if (crfPct > 100 || crfPct < 0) {
crfPct = 90;
}
this.maxCrfSize = (long) (this.maxOplogSize * (crfPct / 100.0));
this.maxDrfSize = this.maxOplogSize - this.maxCrfSize;
}
/**
* A helper function which identifies whether to record a removal of entry in the current oplog or
* to make the switch to the next oplog. This function enables us to reuse the byte buffer which
* got created for an oplog which no longer permits us to use itself. It will also take acre of
* compaction if required
*
* @param entry DiskEntry object representing the current Entry
*/
private void basicRemove(DiskRegionView dr, DiskEntry entry, boolean async, boolean isClear)
throws IOException, InterruptedException {
DiskId id = entry.getDiskId();
boolean useNextOplog = false;
long startPosForSynchOp = -1;
Oplog emptyOplog = null;
if (DiskStoreImpl.KRF_DEBUG) {
// wait for cache close to create krf
System.out.println("basicRemove KRF_DEBUG");
Thread.sleep(1000);
}
getParent().getBackupLock().lock();
try {
synchronized (this.lock) {
if (getOplogSet().getChild() != this) {
useNextOplog = true;
} else if ((this.drf.currSize + MAX_DELETE_ENTRY_RECORD_BYTES) > getMaxDrfSize()
&& !isFirstRecord()) {
switchOpLog(dr, MAX_DELETE_ENTRY_RECORD_BYTES, entry);
useNextOplog = true;
} else {
if (this.lockedForKRFcreate) {
CacheClosedException cce =
parent.getCache().getCacheClosedException("The disk store is closed.");
dr.getCancelCriterion().checkCancelInProgress(cce);
throw cce;
}
long oldOplogId = id.setOplogId(getOplogId());
if (!isClear) {
this.firstRecord = false;
// Ok now we can go ahead and find out its actual size
// This is the only place to set notToUseUserBits=true
initOpState(OPLOG_DEL_ENTRY_1ID, dr, entry, null, (byte) 0, true);
int adjustment = getOpStateSize();
this.drf.currSize += adjustment;
// do the io while holding lock so that switch can set doneAppending
if (logger.isTraceEnabled()) {
logger.trace(
"Oplog::basicRemove: Recording the Deletion of entry in the Oplog with id = {} The Oplog Disk ID for the entry being deleted = {} Mode is Synch",
getOplogId(), id);
}
// Write the data to the opLog for the synch mode
// TODO: if we don't sync write destroys what will happen if
// we do 1. create k1 2. destroy k1 3. create k1?
// It would be possible for the crf to be flushed but not the drf.
// Then during recovery we will find identical keys with different
// entryIds.
// I think we can safely have drf writes be async as long as we flush
// the drf
// before we flush the crf.
// However we can't have removes by async if we are doing a sync write
// because we might be killed right after we do this write.
startPosForSynchOp = writeOpLogBytes(this.drf, async, true);
setHasDeletes(true);
if (logger.isDebugEnabled(LogMarker.PERSIST_WRITES_VERBOSE)) {
logger.debug("basicRemove: id=<{}> key=<{}> drId={} oplog#{}", abs(id.getKeyId()),
entry.getKey(), dr.getId(), getOplogId());
}
if (logger.isTraceEnabled()) {
logger.trace("Oplog::basicRemove:Released ByteBuffer for Disk ID = {}", id);
}
this.dirHolder.incrementTotalOplogSize(adjustment);
}
// Set the oplog size change for stats
id.setOffsetInOplog(-1);
EntryLogger.logPersistDestroy(dr.getName(), entry.getKey(), dr.getDiskStoreID());
Oplog rmOplog = null;
if (oldOplogId == getOplogId()) {
rmOplog = this;
} else {
rmOplog = getOplogSet().getChild(oldOplogId);
}
if (rmOplog != null) {
rmOplog.rmLive(dr, entry);
emptyOplog = rmOplog;
}
clearOpState();
}
}
} finally {
getParent().getBackupLock().unlock();
}
if (useNextOplog) {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSwitchingOplog();
}
Assert.assertTrue(getOplogSet().getChild() != this);
getOplogSet().getChild().basicRemove(dr, entry, async, isClear);
} else {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
CacheObserverHolder.getInstance().afterSettingOplogOffSet(startPosForSynchOp);
}
if (emptyOplog != null
&& (!emptyOplog.isCompacting() || emptyOplog.calledByCompactorThread())) {
emptyOplog.handleNoLiveValues();
}
}
}
/**
* test hook
*/
ByteBuffer getWriteBuf() {
return this.crf.writeBuf;
}
private void flushNoSync(OplogFile olf) throws IOException {
flushAllNoSync(false); // @todo
}
@Override
public void flush() throws IOException {
flushAllNoSync(false);
}
@Override
public void flush(ByteBuffer bb, ByteBuffer chunkbb) throws IOException {
if (bb == this.drf.writeBuf) {
flush(this.drf, bb, chunkbb);
flush(this.crf, false);
} else {
flush(this.drf, false);
flush(this.crf, bb, chunkbb);
}
}
private void flushAndSync(OplogFile olf) throws IOException {
flushAll(false);
}
private static final int MAX_CHANNEL_RETRIES = 5;
private void flush(OplogFile olf, boolean doSync) throws IOException {
try {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* olf */) {
if (olf.RAFClosed) {
return;
}
ByteBuffer bb = olf.writeBuf;
if (bb != null && bb.position() != 0) {
bb.flip();
int flushed = 0;
int numChannelRetries = 0;
do {
int channelBytesWritten = 0;
final int bbStartPos = bb.position();
final long channelStartPos = olf.channel.position();
// differentiate between bytes written on this channel.write() iteration and the
// total number of bytes written to the channel on this call
channelBytesWritten = olf.channel.write(bb);
// Expect channelBytesWritten and the changes in pp.position() and channel.position() to
// be the same. If they are not, then the channel.write() silently failed. The following
// retry separates spurious failures from permanent channel failures.
if (channelBytesWritten != bb.position() - bbStartPos) {
if (numChannelRetries++ < MAX_CHANNEL_RETRIES) {
// Reset the ByteBuffer position, but take into account anything that did get
// written to the channel
channelBytesWritten = (int) (olf.channel.position() - channelStartPos);
bb.position(bbStartPos + channelBytesWritten);
} else {
throw new IOException("Failed to write Oplog entry to" + olf.f.getName() + ": "
+ "channel.write() returned " + channelBytesWritten + ", "
+ "change in channel position = " + (olf.channel.position() - channelStartPos)
+ ", " + "change in source buffer position = " + (bb.position() - bbStartPos));
}
}
flushed += channelBytesWritten;
} while (bb.hasRemaining());
// update bytesFlushed after entire writeBuffer is flushed to fix bug
// 41201
olf.bytesFlushed += flushed;
bb.clear();
}
if (doSync) {
if (SYNC_WRITES) {
// Synch Meta Data as well as content
olf.channel.force(true);
}
}
}
} catch (ClosedChannelException ignore) {
// It is possible for a channel to be closed when our code does not
// explicitly call channel.close (when we will set RAFclosed).
// This can happen when a thread is doing an io op and is interrupted.
// That thread will see ClosedByInterruptException but it will also
// close the channel and then we will see ClosedChannelException.
}
}
private void flush(OplogFile olf, ByteBuffer b1, ByteBuffer b2) throws IOException {
try {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* olf */) {
if (olf.RAFClosed) {
return;
}
this.bbArray[0] = b1;
this.bbArray[1] = b2;
b1.flip();
long flushed = 0;
do {
flushed += olf.channel.write(this.bbArray);
} while (b2.hasRemaining());
this.bbArray[0] = null;
this.bbArray[1] = null;
// update bytesFlushed after entire writeBuffer is flushed to fix bug 41201
olf.bytesFlushed += flushed;
b1.clear();
}
} catch (ClosedChannelException ignore) {
// It is possible for a channel to be closed when our code does not
// explicitly call channel.close (when we will set RAFclosed).
// This can happen when a thread is doing an io op and is interrupted.
// That thread will see ClosedByInterruptException but it will also
// close the channel and then we will see ClosedChannelException.
}
}
public void flushAll() {
flushAll(false);
}
public void flushAllNoSync(boolean skipDrf) {
flushAll(skipDrf, false);
}
public void flushAll(boolean skipDrf) {
flushAll(skipDrf, true/* doSync */);
}
public void flushAll(boolean skipDrf, boolean doSync) {
try {
// TODO: if skipDrf then only need to do drf if crf has flushable data
flush(this.drf, doSync);
flush(this.crf, doSync);
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(
String.format("Failed writing key to %s", this.diskFile.getPath()),
ex, getParent());
}
}
/**
* Since the ByteBuffer being writen to can have additional bytes which are used for extending the
* size of the file, it is necessary that the ByteBuffer provided should have limit which is set
* to the position till which it contains the actual bytes. If the mode is synched write then only
* we will write up to the capacity & opLogSpace variable have any meaning. For asynch mode it
* will be zero. Also this method must be synchronized on the file , whether we use synch or
* asynch write because the fault in operations can clash with the asynch writing. Write the
* specified bytes to the oplog. Note that since extending a file is expensive this code will
* possibly write OPLOG_EXTEND_SIZE zero bytes to reduce the number of times the file is extended.
*
*
* @param olf the file to write the bytes to
* @return The long offset at which the data present in the ByteBuffer gets written to
*/
private long writeOpLogBytes(OplogFile olf, boolean async, boolean doFlushIfSync)
throws IOException {
long startPos = -1L;
getParent().getBackupLock().lock();
try {
synchronized (this.lock/* olf */) {
Assert.assertTrue(!this.doneAppending);
if (this.closed) {
Assert.assertTrue(false, "The Oplog " + this.oplogId + " for store "
+ getParent().getName()
+ " has been closed for synch mode while writing is going on. This should not happen");
}
// It is assumed that the file pointer is already at the
// appropriate position in the file so as to allow writing at the end.
// Any fault in operations will set the pointer back to the write
// location.
// Also it is only in case of synch writing, we are writing more
// than what is actually needed, we will have to reset the pointer.
// Also need to add in offset in writeBuf in case we are not flushing
// writeBuf
startPos = olf.channel.position() + olf.writeBuf.position();
// Assert.assertTrue(startPos > lastWritePos,
// "startPos=" + startPos +
// " was not > lastWritePos=" + lastWritePos);
long bytesWritten = this.opState.write(olf);
if (!async && doFlushIfSync) {
flushAndSync(olf);
}
getStats().incWrittenBytes(bytesWritten, async);
// // Moved the set of lastWritePos to after write
// // so if write throws an exception it will not be updated.
// // This fixes bug 40449.
// this.lastWritePos = startPos;
}
} finally {
getParent().getBackupLock().unlock();
}
return startPos;
}
boolean isRAFOpen() {
return !this.crf.RAFClosed; // volatile read
}
private boolean okToReopen;
boolean closeRAF() {
if (this.beingRead)
return false;
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* crf */) {
if (this.beingRead)
return false;
if (!this.doneAppending)
return false;
if (this.crf.RAFClosed) {
return false;
} else {
try {
this.crf.raf.close();
} catch (IOException ignore) {
}
this.crf.RAFClosed = true;
this.okToReopen = true;
this.stats.decOpenOplogs();
return true;
}
}
}
private volatile boolean beingRead;
/**
* If crfRAF has been closed then attempt to reopen the oplog for this read. Verify that this only
* happens when test methods are invoked.
*
* @return true if oplog file is open and can be read from; false if not
*/
private boolean reopenFileIfClosed() throws IOException {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* crf */) {
boolean result = !this.crf.RAFClosed;
if (!result && this.okToReopen) {
result = true;
this.crf.raf = new UninterruptibleRandomAccessFile(this.crf.f, "r");
this.stats.incOpenOplogs();
this.crf.RAFClosed = false;
this.okToReopen = false;
}
return result;
}
}
private BytesAndBits attemptGet(DiskRegionView dr, long offsetInOplog, boolean bitOnly,
int valueLength, byte userBits) throws IOException {
boolean didReopen = false;
boolean accessedInactive = false;
try {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock) {
this.beingRead = true;
if ((offsetInOplog + valueLength) > this.crf.bytesFlushed && !this.closed) {
flushAllNoSync(true); // fix for bug 41205
}
try {
UninterruptibleRandomAccessFile myRAF = null;
if (this.crf.RAFClosed) {
myRAF = new UninterruptibleRandomAccessFile(this.crf.f, "r");
this.stats.incOpenOplogs();
if (this.okToReopen) {
this.crf.RAFClosed = false;
this.okToReopen = false;
this.crf.raf = myRAF;
didReopen = true;
}
} else {
myRAF = this.crf.raf;
accessedInactive = true;
}
BytesAndBits bb = null;
try {
final long writePosition =
(this.doneAppending) ? this.crf.bytesFlushed : myRAF.getFilePointer();
if ((offsetInOplog + valueLength) > writePosition) {
throw new DiskAccessException(
String.format(
"Tried to seek to %s, but the file length is %s. Oplog File object used for reading=%s",
offsetInOplog + valueLength, writePosition, this.crf.raf),
dr.getName());
} else if (offsetInOplog < 0) {
throw new DiskAccessException(
String.format("Cannot find record %s when reading from %s",
offsetInOplog, this.diskFile.getPath()),
dr.getName());
}
try {
myRAF.seek(offsetInOplog);
this.stats.incOplogSeeks();
byte[] valueBytes = new byte[valueLength];
myRAF.readFully(valueBytes);
this.stats.incOplogReads();
bb = new BytesAndBits(valueBytes, userBits);
// also set the product version for an older product
final KnownVersion version = getProductVersionIfOld();
if (version != null) {
bb.setVersion(version);
}
} finally {
// if this oplog is no longer being appended to then don't waste
// disk io
if (!this.doneAppending) {
// by seeking back to writePosition
myRAF.seek(writePosition);
this.stats.incOplogSeeks();
}
}
return bb;
} finally {
if (myRAF != this.crf.raf) {
try {
myRAF.close();
} catch (IOException ignore) {
}
}
}
} finally {
this.beingRead = false;
}
} // sync
} finally {
if (accessedInactive) {
getOplogSet().inactiveAccessed(this);
} else if (didReopen) {
getOplogSet().inactiveReopened(this);
}
}
}
/**
* Extracts the Value byte array & UserBit from the OpLog
*
* @param offsetInOplog The starting position from which to read the data in the opLog
* @param bitOnly boolean indicating whether the value needs to be extracted along with the
* UserBit or not.
* @param valueLength The length of the byte array which represents the value
* @param userBits The userBits of the value.
* @return BytesAndBits object which wraps the extracted value & user bit
*/
private BytesAndBits basicGet(DiskRegionView dr, long offsetInOplog, boolean bitOnly,
int valueLength, byte userBits) {
BytesAndBits bb = null;
if (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits) || bitOnly
|| valueLength == 0) {
if (EntryBits.isInvalid(userBits)) {
bb = new BytesAndBits(DiskEntry.INVALID_BYTES, userBits);
} else if (EntryBits.isTombstone(userBits)) {
bb = new BytesAndBits(DiskEntry.TOMBSTONE_BYTES, userBits);
} else {
bb = new BytesAndBits(DiskEntry.LOCAL_INVALID_BYTES, userBits);
}
} else {
if (offsetInOplog == -1)
return null;
try {
for (;;) {
dr.getCancelCriterion().checkCancelInProgress(null);
boolean interrupted = Thread.interrupted();
try {
bb = attemptGet(dr, offsetInOplog, bitOnly, valueLength, userBits);
break;
} catch (InterruptedIOException ignore) { // bug 39756
// ignore, we'll clear and retry.
} finally {
if (interrupted) {
Thread.currentThread().interrupt();
}
}
} // for
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(
String.format(
"Failed reading from %s. oplogID, %s Offset being read= %s Current Oplog Size= %s Actual File Size, %s IS ASYNCH MODE, %s IS ASYNCH WRITER ALIVE= %s",
this.diskFile.getPath(), this.oplogId, offsetInOplog,
this.crf.currSize, this.crf.bytesFlushed, !dr.isSync(), Boolean.FALSE),
ex, dr.getName());
} catch (IllegalStateException ex) {
checkClosed();
throw ex;
}
}
return bb;
}
/**
* Extracts the Value byte array & UserBit from the OpLog and inserts it in the wrapper Object of
* type BytesAndBitsForCompactor which is passed
*
* @param offsetInOplog The starting position from which to read the data in the opLog
* @param bitOnly boolean indicating whether the value needs to be extracted along with the
* UserBit or not.
* @param valueLength The length of the byte array which represents the value
* @param userBits The userBits of the value.
* @param wrapper Object of type BytesAndBitsForCompactor. The data is set in the wrapper Object.
* The wrapper Object also contains the user bit associated with the entry
* @return true if data is found false if not
*/
private boolean basicGetForCompactor(DiskRegionView dr, long offsetInOplog, boolean bitOnly,
int valueLength, byte userBits, BytesAndBitsForCompactor wrapper) {
if (EntryBits.isAnyInvalid(userBits) || EntryBits.isTombstone(userBits) || bitOnly
|| valueLength == 0) {
if (EntryBits.isInvalid(userBits)) {
wrapper.setData(DiskEntry.INVALID_BYTES, userBits, DiskEntry.INVALID_BYTES.length,
false /*
* Cannot be reused
*/);
} else if (EntryBits.isTombstone(userBits)) {
wrapper.setData(DiskEntry.TOMBSTONE_BYTES, userBits, DiskEntry.TOMBSTONE_BYTES.length,
false /*
* Cannot be reused
*/);
} else {
wrapper.setData(DiskEntry.LOCAL_INVALID_BYTES, userBits,
DiskEntry.LOCAL_INVALID_BYTES.length, false /*
* Cannot be reused
*/);
}
} else {
try {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* crf */) {
if (/*
* !getParent().isSync() since compactor groups writes &&
*/(offsetInOplog + valueLength) > this.crf.bytesFlushed && !this.closed) {
flushAllNoSync(true); // fix for bug 41205
}
if (!reopenFileIfClosed()) {
return false; // fix for bug 40648
}
final long writePosition =
(this.doneAppending) ? this.crf.bytesFlushed : this.crf.raf.getFilePointer();
if ((offsetInOplog + valueLength) > writePosition) {
throw new DiskAccessException(
String.format(
"Tried to seek to %s, but the file length is %s. Oplog File object used for reading=%s",
offsetInOplog + valueLength, writePosition, this.crf.raf),
dr.getName());
} else if (offsetInOplog < 0) {
throw new DiskAccessException(
String.format("Cannot find record %s when reading from %s",
offsetInOplog, this.diskFile.getPath()),
dr.getName());
}
try {
this.crf.raf.seek(offsetInOplog);
this.stats.incOplogSeeks();
byte[] valueBytes = null;
if (wrapper.getBytes().length < valueLength) {
valueBytes = new byte[valueLength];
this.crf.raf.readFully(valueBytes);
} else {
valueBytes = wrapper.getBytes();
this.crf.raf.readFully(valueBytes, 0, valueLength);
}
this.stats.incOplogReads();
wrapper.setData(valueBytes, userBits, valueLength, true);
} finally {
// if this oplog is no longer being appended to then don't waste
// disk io
if (!this.doneAppending) {
this.crf.raf.seek(writePosition);
this.stats.incOplogSeeks();
}
}
}
} catch (IOException ex) {
getParent().getCancelCriterion().checkCancelInProgress(ex);
throw new DiskAccessException(
String.format(
"Failed reading from %s. oplogID, %s Offset being read=%s Current Oplog Size=%s Actual File Size,%s IS ASYNCH MODE,%s IS ASYNCH WRITER ALIVE=%s",
this.diskFile.getPath(), this.oplogId, offsetInOplog,
this.crf.currSize, this.crf.bytesFlushed, Boolean.FALSE, Boolean.FALSE),
ex, dr.getName());
} catch (IllegalStateException ex) {
checkClosed();
throw ex;
}
}
return true;
}
private final AtomicBoolean deleted = new AtomicBoolean();
/**
* deletes the oplog's file(s)
*/
void deleteFiles(boolean crfOnly) {
// try doing the removeOplog unconditionally since I'm see an infinite loop
// in destroyOldestReadyToCompact
boolean needsDestroy = this.deleted.compareAndSet(false, true);
if (needsDestroy) {
// I don't under stand why the compactor would have anything to do with
// an oplog file that we are removing from disk.
// So I'm commenting out the following if
// if (!isCompactionPossible()) {
// moved this from close to fix bug 40574
// If we get to the point that it is ok to close the file
// then we no longer need the parent to be able to find this
// oplog using its id so we can unregister it now.
// If compaction is possible then we need to leave this
// oplog registered with the parent and allow the compactor to unregister
// it.
deleteCRF();
if (!crfOnly || !getHasDeletes()) {
setHasDeletes(false);
deleteDRF();
// no need to call removeDrf since parent removeOplog did it
}
// Fix for bug 42495 - Don't remove the oplog from this list
// of oplogs until it has been removed from the init file. This guarantees
// that if the oplog is in the init file, the backup code can find it and
// try to back it up.
boolean addToDrfOnly = crfOnly && getHasDeletes();
getOplogSet().removeOplog(getOplogId(), true, addToDrfOnly ? this : null);
} else if (!crfOnly && getHasDeletes()) {
setHasDeletes(false);
deleteDRF();
getOplogSet().removeDrf(this);
}
}
public void deleteCRF() {
oplogSet.crfDelete(this.oplogId);
if (!getInternalCache().getBackupService().deferCrfDelete(getParent(), this)) {
deleteCRFFileOnly();
}
}
public void deleteCRFFileOnly() {
deleteFile(this.crf);
// replace .crf at the end with .krf
if (this.crf.f != null) {
final File krf = new File(
this.crf.f.getAbsolutePath().replaceFirst("\\" + CRF_FILE_EXT + "$", KRF_FILE_EXT));
if (!krf.exists()) {
return;
}
getParent().executeDelayedExpensiveWrite(new Runnable() {
@Override
public void run() {
if (!krf.delete()) {
if (krf.exists()) {
logger.warn("Could not delete the file {} {} for disk store {}.",
new Object[] {Oplog.this.toString(), "krf", getParent().getName()});
}
} else {
logger.info("Deleted {} {} for disk store {}.",
new Object[] {Oplog.this.toString(), "krf", getParent().getName()});
}
}
});
}
}
public void deleteDRF() {
getOplogSet().drfDelete(this.oplogId);
if (!getInternalCache().getBackupService().deferDrfDelete(getParent(), this)) {
deleteDRFFileOnly();
}
}
public void deleteDRFFileOnly() {
deleteFile(this.drf);
}
/**
* Returns "crf" or "drf".
*/
private static String getFileType(OplogFile olf) {
String name = olf.f.getName();
int index = name.lastIndexOf('.');
return name.substring(index + 1);
}
private void deleteFile(final OplogFile olf) {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock) {
if (olf.currSize != 0) {
this.dirHolder.decrementTotalOplogSize(olf.currSize);
olf.currSize = 0;
}
if (olf.f == null)
return;
if (!olf.f.exists())
return;
assert olf.RAFClosed;
if (olf.raf != null) {
try {
olf.raf.close();
olf.RAFClosed = true;
} catch (IOException ignore) {
}
}
// Delete the file asynchronously. Based on perf testing, deletes
// can block at the filesystem level. See #50254
// It's safe to do this asynchronously, because we have already
// marked this file as deleted in the init file.
// Note - could we run out of disk space because the compaction thread is
// doing this and creating files? For a real fix, you probably need a
// bounded
// queue
getParent().executeDelayedExpensiveWrite(new Runnable() {
@Override
public void run() {
if (!olf.f.delete() && olf.f.exists()) {
logger.warn("Could not delete the file {} {} for disk store {}.",
Oplog.this.toString(), getFileType(olf),
getParent().getName());
} else {
logger.info("Deleted {} {} for disk store {}.",
Oplog.this.toString(), getFileType(olf), getParent().getName());
}
}
});
}
}
/**
* Helper function for the test
*
* @return FileChannel object representing the Oplog
*/
UninterruptibleFileChannel getFileChannel() {
return this.crf.channel;
}
public DirectoryHolder getDirectoryHolder() {
return this.dirHolder;
}
/**
* The current size of Oplog. It may be less than the actual Oplog file size ( in case of asynch
* writing as it also takes into account data present in asynch buffers which will get flushed in
* course of time o
*
* @return long value indicating the current size of the oplog.
*/
long getOplogSize() {
return this.crf.currSize + this.drf.currSize;
}
boolean isOplogEmpty() {
return this.crf.currSize <= (OPLOG_DISK_STORE_REC_SIZE + OPLOG_MAGIC_SEQ_REC_SIZE)
&& this.drf.currSize <= (OPLOG_DISK_STORE_REC_SIZE + OPLOG_MAGIC_SEQ_REC_SIZE);
}
void incLiveCount() {
this.totalLiveCount.incrementAndGet();
}
private void decLiveCount() {
this.totalLiveCount.decrementAndGet();
}
/**
* Return true if a record (crf or drf) has been added to this oplog
*/
boolean hasBeenUsed() {
return this.hasDeletes.get() || this.totalCount.get() > 0;
}
void incTotalCount() {
if (!isPhase2()) {
this.totalCount.incrementAndGet();
}
}
private void finishedAppending() {
// No need to get the backup lock prior to synchronizing (correct lock order) since the
// synchronized block does not attempt to get the backup lock (incorrect lock order)
synchronized (this.lock/* crf */) {
this.doneAppending = true;
}
handleNoLiveValues();
// I'm deadcoding the following because it is not safe unless we change to
// always recover values. If we don't recover values then
// an oplog we recovered from may still need to fault values in from memory.
// if (!getParent().isOverflowEnabled()) {
// // If !overflow then we can close the file even
// // when it has recent values because
// // we will never need to fault values in from this
// // file since they are all in memory.
// close();
// }
}
boolean needsCompaction() {
if (!isCompactionPossible())
return false;
if (this.unrecoveredRegionCount.get() > 0)
return false;
if (parent.getCompactionThreshold() == 100)
return true;
if (parent.getCompactionThreshold() == 0)
return false;
// otherwise check if we have enough garbage to collect with a compact
long rvHWMtmp = this.totalCount.get();
if (rvHWMtmp > 0) {
long tlc = this.totalLiveCount.get();
if (tlc < 0) {
tlc = 0;
}
double rv = tlc;
if (((rv / (double) rvHWMtmp) * 100) <= parent.getCompactionThreshold()) {
return true;
}
} else {
return true;
}
return false;
}
public boolean hadLiveEntries() {
return this.totalCount.get() != 0;
}
public boolean hasNoLiveValues() {
return this.totalLiveCount.get() <= 0
// if we have an unrecoveredRegion then we don't know how many liveValues we
// have
&& this.unrecoveredRegionCount.get() == 0 && !getParent().isOfflineCompacting();
}
private void handleEmptyAndOldest(boolean calledByCompactor) {
if (!calledByCompactor && logger.isDebugEnabled()) {
logger.debug(
"Deleting oplog early because it is empty. It is for disk store {} and has oplog#{}",
getParent().getName(), oplogId);
}
destroy();
getOplogSet().destroyOldestReadyToCompact();
}
private void handleEmpty(boolean calledByCompactor) {
lockCompactor();
try {
if (!calledByCompactor) {
logger.info("Closing {} early since it is empty. It is for disk store {}.",
new Object[] {getParent().getName(), toString()});
}
cancelKrf();
close();
deleteFiles(getHasDeletes());
} finally {
unlockCompactor();
}
}
void cancelKrf() {
createKrf(true);
}
private static final ThreadLocal isCompactorThread = new ThreadLocal();
boolean calledByCompactorThread() {
if (!this.compacting)
return false;
Object v = isCompactorThread.get();
return v != null && v == Boolean.TRUE;
}
void handleNoLiveValues() {
if (!this.doneAppending)
return;
if (hasNoLiveValues()) {
if (LocalRegion.ISSUE_CALLBACKS_TO_CACHE_OBSERVER) {
if (calledByCompactorThread()) {
// after compaction, remove the oplog from the list & destroy it
CacheObserverHolder.getInstance().beforeDeletingCompactedOplog(this);
} else {
CacheObserverHolder.getInstance().beforeDeletingEmptyOplog(this);
}
}
if (isOldest()) {
if (calledByCompactorThread()) {
// do it in this compactor thread
handleEmptyAndOldest(true);
} else {
// schedule another thread to do it
getParent().executeDiskStoreTask(new Runnable() {
@Override
public void run() {
handleEmptyAndOldest(false);
}
});
}
} else {
if (calledByCompactorThread()) {
// do it in this compactor thread
handleEmpty(true);
} else {
// schedule another thread to do it
getParent().executeDiskStoreTask(new Runnable() {
@Override
public void run() {
handleEmpty(false);
}
});
}
}
} else if (needsCompaction()) {
addToBeCompacted();
}
}
private InternalCache getInternalCache() {
return getParent().getCache();
}
/**
* Return true if this oplog is the oldest one of those ready to compact
*/
private boolean isOldest() {
long myId = getOplogId();
return getOplogSet().isOldestExistingOplog(myId);
}
private boolean added = false;
private synchronized void addToBeCompacted() {
if (this.added)
return;
this.added = true;
getOplogSet().addToBeCompacted(this);
if (logger.isDebugEnabled()) {
logger.debug("Oplog::switchOpLog: Added the Oplog = {} for compacting.", this.oplogId);
}
}
private void initRecoveredEntry(DiskRegionView drv, DiskEntry de) {
addLive(drv, de);
}
/**
* The oplogId in re points to the oldOplogId. "this" oplog is the current oplog.
*/
private void updateRecoveredEntry(DiskRegionView drv, DiskEntry de, DiskEntry.RecoveredEntry re) {
if (getOplogId() != re.getOplogId()) {
Oplog oldOplog = getOplogSet().getChild(re.getOplogId());
oldOplog.rmLive(drv, de);
initRecoveredEntry(drv, de);
} else {
getDRI(drv).update(de);
}
}
@Override
public void prepareForCompact() {
this.compacting = true;
}
private final Lock compactorLock = new ReentrantLock();
private void lockCompactor() {
this.compactorLock.lock();
}
private void unlockCompactor() {
this.compactorLock.unlock();
}
/**
* Copy any live entries last stored in this oplog to the current oplog. No need to copy deletes
* in the drf. Backup only needs them until all the older crfs are empty.
*/
@Override
public int compact(OplogCompactor compactor) {
if (!needsCompaction()) {
return 0; // @todo check new logic that deals with not compacting oplogs
// which have unrecovered regions
}
isCompactorThread.set(Boolean.TRUE);
assert calledByCompactorThread();
getParent().acquireCompactorReadLock();
try {
if (!compactor.keepCompactorRunning()) {
return 0;
}
lockCompactor();
try {
if (hasNoLiveValues()) {
handleNoLiveValues();
return 0; // do this while holding compactorLock
}
// Start with a fresh wrapper on every compaction so that
// if previous run used some high memory byte array which was
// exceptional, it gets garbage collected.
long opStart = getStats().getStatTime();
BytesAndBitsForCompactor wrapper = new BytesAndBitsForCompactor();
DiskEntry de;
DiskEntry lastDe = null;
boolean compactFailed = /*
* getParent().getOwner().isDestroyed ||
*/!compactor.keepCompactorRunning();
int totalCount = 0;
for (DiskRegionInfo dri : this.regionMap.values()) {
final DiskRegionView dr = dri.getDiskRegion();
if (dr == null)
continue;
boolean didCompact = false;
while ((de = dri.getNextLiveEntry()) != null) {
if (/*
* getParent().getOwner().isDestroyed ||
*/!compactor.keepCompactorRunning()) {
compactFailed = true;
break;
}
if (lastDe != null) {
if (lastDe == de) {
throw new IllegalStateException("compactor would have gone into infinite loop");
}
assert lastDe != de;
}
lastDe = de;
didCompact = false;
synchronized (de) { // fix for bug 41797
DiskId did = de.getDiskId();
assert did != null;
synchronized (did) {
long oplogId = did.getOplogId();
if (oplogId != getOplogId()) {
continue;
}
boolean toCompact = getBytesAndBitsForCompaction(dr, de, wrapper);
if (toCompact) {
if (oplogId != did.getOplogId()) {
// @todo: Is this even possible? Perhaps I should just assert here
// skip this one, its oplogId changed
if (!wrapper.isReusable()) {
wrapper = new BytesAndBitsForCompactor();
} else if (wrapper.getOffHeapData() != null) {
wrapper.setOffHeapData(null, (byte) 0);
}
continue;
}
// write it to the current oplog
getOplogSet().getChild().copyForwardModifyForCompact(dr, de, wrapper);
// the did's oplogId will now be set to the current active oplog
didCompact = true;
}
} // did
} // de
if (didCompact) {
totalCount++;
getStats().endCompactionUpdate(opStart);
opStart = getStats().getStatTime();
// Check if the value byte array happens to be any of the
// constant
// static byte arrays or references the value byte array of
// underlying RegionEntry.
// If so for preventing data corruption across regions
// ( in case of static byte arrays) & for RegionEntry,
// recreate the wrapper
if (!wrapper.isReusable()) {
wrapper = new BytesAndBitsForCompactor();
}
}
}
}
cleanupAfterCompaction(compactFailed);
return totalCount;
} finally {
unlockCompactor();
}
} finally {
getParent().releaseCompactorReadLock();
assert calledByCompactorThread();
isCompactorThread.remove();
}
}
void cleanupAfterCompaction(boolean compactFailed) {
if (!compactFailed) {
// all data has been copied forward to new oplog so no live entries remain
getTotalLiveCount().set(0);
// Need to still remove the oplog even if it had nothing to compact.
handleNoLiveValues();
}
}
AtomicLong getTotalLiveCount() {
return totalLiveCount;
}
public static boolean isCRFFile(String filename) {
return filename.endsWith(Oplog.CRF_FILE_EXT);
}
public static boolean isDRFFile(String filename) {
return filename.endsWith(Oplog.DRF_FILE_EXT);
}
public static String getKRFFilenameFromCRFFilename(String crfFilename) {
return crfFilename.substring(0, crfFilename.length() - Oplog.CRF_FILE_EXT.length())
+ Oplog.KRF_FILE_EXT;
}
long testGetOplogFileLength() throws IOException {
long result = 0;
if (this.crf.raf != null) {
result += this.crf.raf.length();
}
if (this.drf.raf != null) {
result += this.drf.raf.length();
}
return result;
}
/**
* This method is called by the async value recovery task to recover the values from the crf if
* the keys were recovered from the krf.
*/
public void recoverValuesIfNeeded(Map<Long, DiskRecoveryStore> diskRecoveryStores) {
// Early out if we start closing the parent.
if (getParent().isClosing()) {
return;
}
List<KRFEntry> sortedLiveEntries;
HashMap<Long, DiskRegionInfo> targetRegions = new HashMap<Long, DiskRegionInfo>(this.regionMap);
synchronized (diskRecoveryStores) {
Iterator<DiskRecoveryStore> itr = diskRecoveryStores.values().iterator();
while (itr.hasNext()) {
DiskRecoveryStore store = itr.next();
if (isLruValueRecoveryDisabled(store) || store.lruLimitExceeded()) {
itr.remove();
}
}
// Get the a sorted list of live entries from the target regions
targetRegions.keySet().retainAll(diskRecoveryStores.keySet());
}
sortedLiveEntries = getSortedLiveEntries(targetRegions.values());
if (sortedLiveEntries == null) {
// There are no live entries in this oplog to recover.
return;
}
final ByteArrayDataInput in = new ByteArrayDataInput();
for (KRFEntry entry : sortedLiveEntries) {
// Early out if we start closing the parent.
if (getParent().isClosing()) {
return;
}
DiskEntry diskEntry = entry.getDiskEntry();
DiskRegionView diskRegionView = entry.getDiskRegionView();
long diskRegionId = diskRegionView.getId();
// TODO DAN ok, here's what we need to do
// 1) lock and obtain the correct RegionEntry that we are recovering too.
// this will likely mean obtaining the correct DiskRecoveryStore, since
// with
// that we can find the region entry I believe.
// 2) Make sure that the lru limit is not exceeded
// 3) Update the region entry with the value from disk, assuming the value
// from
// disk is still valid. That is going to be something like
synchronized (diskRecoveryStores) {
DiskRecoveryStore diskRecoveryStore = diskRecoveryStores.get(diskRegionId);
if (diskRecoveryStore == null) {
continue;
}
// Reset the disk region view because it may have changed
// due to the region being created.
diskRegionView = diskRecoveryStore.getDiskRegionView();
if (diskRegionView == null) {
continue;
}
if (diskRecoveryStore.lruLimitExceeded()) {
diskRecoveryStores.remove(diskRegionId);
continue;
}
if (diskRegionView.isEntriesMapIncompatible()) {
// Refetch the disk entry because it may have changed due to copying
// an incompatible region map
diskEntry = (DiskEntry) diskRecoveryStore.getRegionMap().getEntryInVM(diskEntry.getKey());
if (diskEntry == null) {
continue;
}
}
synchronized (diskEntry) {
// Make sure the entry hasn't been modified
if (diskEntry.getDiskId() != null && diskEntry.getDiskId().getOplogId() == oplogId) {
// dear lord, this goes through a lot of layers. Maybe we should
// skip some?
// * specifically, this could end up faulting in from a different
// oplog, causing
// us to seek.
// * Also, there may be lock ordering issues here, Really, I guess I
// want
// a flavor of faultInValue that only faults in from this oplog.
// * We could have some churn here, opening and closing this oplog
// * We also might not be buffering adjacent entries? Not sure about
// that one
// * Ideally, this would fault the thing in only if it were in this
// oplog and the lru limit wasn't hit
// and it would return a status if the lru limit was hit to make us
// remove the store.
try {
DiskEntry.Helper.recoverValue(diskEntry, getOplogId(), diskRecoveryStore, in);
} catch (RegionDestroyedException ignore) {
// This region has been destroyed, stop recovering from it.
diskRecoveryStores.remove(diskRegionId);
}
}
}
}
}
}
private byte[] serializeRVVs(Map<Long, AbstractDiskRegion> drMap, boolean gcRVV)
throws IOException {
HeapDataOutputStream out = new HeapDataOutputStream(KnownVersion.CURRENT);
// Filter out any regions that do not have versioning enabled
drMap = new HashMap<Long, AbstractDiskRegion>(drMap);
for (Iterator<Map.Entry<Long, AbstractDiskRegion>> itr = drMap.entrySet().iterator(); itr
.hasNext();) {
Map.Entry<Long, AbstractDiskRegion> regionEntry = itr.next();
AbstractDiskRegion dr = regionEntry.getValue();
if (!dr.getFlags().contains(DiskRegionFlag.IS_WITH_VERSIONING)) {
itr.remove();
}
}
// Write the size first
InternalDataSerializer.writeUnsignedVL(drMap.size(), out);
// Now write regions RVV.
for (Map.Entry<Long, AbstractDiskRegion> regionEntry : drMap.entrySet()) {
// For each region, write the RVV for the region.
Long diskRegionID = regionEntry.getKey();
AbstractDiskRegion dr = regionEntry.getValue();
RegionVersionVector rvv = dr.getRegionVersionVector();
if (logger.isTraceEnabled(LogMarker.PERSIST_WRITES_VERBOSE)) {
logger.trace(LogMarker.PERSIST_WRITES_VERBOSE,
"serializeRVVs: isGCRVV={} drId={} rvv={} oplog#{}", gcRVV, diskRegionID,
rvv.fullToString(), getOplogId());
}
// Write the disk region id
InternalDataSerializer.writeUnsignedVL(diskRegionID, out);
if (gcRVV) {
// For the GC RVV, we will just write the GC versions
Map<VersionSource, Long> memberToVersion = rvv.getMemberToGCVersion();
InternalDataSerializer.writeUnsignedVL(memberToVersion.size(), out);
for (Entry<VersionSource, Long> memberEntry : memberToVersion.entrySet()) {
// For each member, write the canonicalized member id,
// and the version number for that member
VersionSource member = memberEntry.getKey();
Long gcVersion = memberEntry.getValue();
int id = getParent().getDiskInitFile().getOrCreateCanonicalId(member);
InternalDataSerializer.writeUnsignedVL(id, out);
InternalDataSerializer.writeUnsignedVL(gcVersion, out);
}
} else {
DataSerializer.writeBoolean(dr.getRVVTrusted(), out);
// Otherwise, we will write the version and exception list for each
// member
Map<VersionSource, RegionVersionHolder> memberToVersion = rvv.getMemberToVersion();
InternalDataSerializer.writeUnsignedVL(memberToVersion.size(), out);
for (Map.Entry<VersionSource, RegionVersionHolder> memberEntry : memberToVersion
.entrySet()) {
// For each member, right the canonicalized member id,
// and the version number with exceptions for that member
VersionSource member = memberEntry.getKey();
RegionVersionHolder versionHolder = memberEntry.getValue();
int id = getParent().getDiskInitFile().getOrCreateCanonicalId(member);
InternalDataSerializer.writeUnsignedVL(id, out);
synchronized (versionHolder) {
InternalDataSerializer.invokeToData(versionHolder, out);
}
}
}
}
return out.toByteArray();
}
@Override
public String toString() {
return "oplog#" + getOplogId();
}
/**
* Method to be used only for testing
*
* @param ch Object to replace the channel in the Oplog.crf
* @return original channel object
*/
UninterruptibleFileChannel testSetCrfChannel(UninterruptibleFileChannel ch) {
UninterruptibleFileChannel chPrev = this.crf.channel;
this.crf.channel = ch;
return chPrev;
}
private static String baToString(byte[] ba) {
return baToString(ba, ba != null ? ba.length : 0);
}
private static String baToString(byte[] ba, int len) {
if (ba == null) {
return "null";
}
StringBuilder sb = new StringBuilder();
for (int i = 0; i < len; i++) {
sb.append(ba[i]).append(", ");
}
return sb.toString();
}
void serializeVersionTag(VersionHolder tag, DataOutput out) throws IOException {
int entryVersion = tag.getEntryVersion();
long regionVersion = tag.getRegionVersion();
VersionSource versionMember = tag.getMemberID();
long timestamp = tag.getVersionTimeStamp();
int dsId = tag.getDistributedSystemId();
serializeVersionTag(entryVersion, regionVersion, versionMember, timestamp, dsId, out);
}
byte[] serializeVersionTag(VersionTag tag) throws IOException {
int entryVersion = tag.getEntryVersion();
long regionVersion = tag.getRegionVersion();
VersionSource versionMember = tag.getMemberID();
long timestamp = tag.getVersionTimeStamp();
int dsId = tag.getDistributedSystemId();
return serializeVersionTag(entryVersion, regionVersion, versionMember, timestamp, dsId);
}
byte[] serializeVersionTag(VersionStamp stamp) throws IOException {
int entryVersion = stamp.getEntryVersion();
long regionVersion = stamp.getRegionVersion();
VersionSource versionMember = stamp.getMemberID();
long timestamp = stamp.getVersionTimeStamp();
int dsId = stamp.getDistributedSystemId();
return serializeVersionTag(entryVersion, regionVersion, versionMember, timestamp, dsId);
}
private byte[] serializeVersionTag(int entryVersion, long regionVersion,
VersionSource versionMember, long timestamp, int dsId) throws IOException {
HeapDataOutputStream out = new HeapDataOutputStream(4 + 8 + 4 + 8 + 4, KnownVersion.CURRENT);
serializeVersionTag(entryVersion, regionVersion, versionMember, timestamp, dsId, out);
return out.toByteArray();
}
private void serializeVersionTag(int entryVersion, long regionVersion,
VersionSource versionMember, long timestamp, int dsId, DataOutput out) throws IOException {
int memberId = getParent().getDiskInitFile().getOrCreateCanonicalId(versionMember);
InternalDataSerializer.writeSignedVL(entryVersion, out);
InternalDataSerializer.writeUnsignedVL(regionVersion, out);
InternalDataSerializer.writeUnsignedVL(memberId, out);
InternalDataSerializer.writeUnsignedVL(timestamp, out);
InternalDataSerializer.writeSignedVL(dsId, out);
}
public void finishKrf() {
createKrf(false);
}
void prepareForClose() {
try {
finishKrf();
} catch (CancelException e) {
// workaround for 50465
if (logger.isDebugEnabled()) {
logger.debug("Got a cancel exception while creating a krf during shutown", e);
}
}
}
private Object deserializeKey(byte[] keyBytes, final KnownVersion version,
final ByteArrayDataInput in) {
if (!getParent().isOffline() || !PdxWriterImpl.isPdx(keyBytes)) {
return EntryEventImpl.deserialize(keyBytes, version, in);
} else {
return new RawByteKey(keyBytes);
}
}
/**
* If this OpLog is from an older version of the product, then return that
* {@link KnownVersion} else
* return null.
*/
public KnownVersion getProductVersionIfOld() {
final KnownVersion version = this.gfversion;
if (version == null) {
// check for the case of diskstore upgrade from 6.6 to >= 7.0
if (getParent().isUpgradeVersionOnly()) {
// assume previous release version
return KnownVersion.GFE_66;
} else {
return null;
}
} else if (version == KnownVersion.CURRENT) {
return null;
} else {
// version changed so return that for VersionedDataStream
return version;
}
}
/**
* If this OpLog has data that was written by an older version of the product, then return that
* {@link KnownVersion} else return null.
*/
public KnownVersion getDataVersionIfOld() {
final KnownVersion version = this.dataVersion;
if (version == null) {
// check for the case of diskstore upgrade from 6.6 to >= 7.0
if (getParent().isUpgradeVersionOnly()) {
// assume previous release version
return KnownVersion.GFE_66;
} else {
return null;
}
} else if (version == KnownVersion.CURRENT) {
return null;
} else {
// version changed so return that for VersionedDataStream
return version;
}
}
public enum OPLOG_TYPE {
CRF(new byte[] {0x47, 0x46, 0x43, 0x52, 0x46, 0x31}), // GFCRF1
DRF(new byte[] {0x47, 0x46, 0x44, 0x52, 0x46, 0x31}), // GFDRF1
IRF(new byte[] {0x47, 0x46, 0x49, 0x52, 0x46, 0x31}), // GFIRF1
KRF(new byte[] {0x47, 0x46, 0x4b, 0x52, 0x46, 0x31}), // GFKRF1
IF(new byte[] {0x47, 0x46, 0x49, 0x46, 0x30, 0x31}); // GFIF01
private final byte[] bytes;
OPLOG_TYPE(byte[] byteSeq) {
this.bytes = byteSeq;
}
public byte[] getBytes() {
return bytes;
}
public static int getLen() {
return 6;
}
}
/**
* Enumeration of the possible results of the okToSkipModifyRecord
*/
private enum OkToSkipResult {
SKIP_RECORD, // Skip reading the key and value
SKIP_VALUE, // skip reading just the value
DONT_SKIP; // don't skip the record
public boolean skip() {
return this != DONT_SKIP;
}
public boolean skipKey() {
return this == SKIP_RECORD;
}
}
/**
* Used when creating a KRF to keep track of what DiskRegionView a DiskEntry belongs to.
*/
private static class KRFEntry {
private final DiskEntry de;
private final DiskRegionView drv;
/**
* Fix for 42733 - a stable snapshot of the offset so we can sort It doesn't matter that this is
* stale, we'll filter out these entries later.
*/
private final long offsetInOplog;
private VersionHolder versionTag;
public KRFEntry(DiskRegionView drv, DiskEntry de, VersionHolder tag) {
this.de = de;
this.drv = drv;
DiskId diskId = de.getDiskId();
this.offsetInOplog = diskId != null ? diskId.getOffsetInOplog() : 0;
this.versionTag = tag;
}
public DiskEntry getDiskEntry() {
return this.de;
}
public DiskRegionView getDiskRegionView() {
return this.drv;
}
public long getOffsetInOplogForSorting() {
return offsetInOplog;
}
}
private static class OplogFile {
public File f;
public UninterruptibleRandomAccessFile raf;
public volatile boolean RAFClosed = true;
public UninterruptibleFileChannel channel;
public ByteBuffer writeBuf;
public long currSize;
public long bytesFlushed;
public boolean unpreblown;
}
private static class KRFile {
public File f;
FileOutputStream fos;
BufferedOutputStream bos;
DataOutputStream dos;
long lastOffset = 0;
int keyNum = 0;
}
/**
* Holds all the state for the current operation. Since an oplog can only have one operation in
* progress at any given time we only need a single instance of this class per oplog.
*/
private class OpState {
private byte opCode;
private byte userBits;
private boolean notToUseUserBits; // currently only DestroyFromDisk will not
// use userBits
/**
* How many bytes it will be when serialized
*/
private int size;
private boolean needsValue;
private ValueWrapper value;
private int drIdLength; // 1..9
private final byte[] drIdBytes = new byte[DiskInitFile.DR_ID_MAX_BYTES];
private byte[] keyBytes;
private final byte[] deltaIdBytes = new byte[8];
private int deltaIdBytesLength;
private long newEntryBase;
private DiskStoreID diskStoreId;
private OPLOG_TYPE magic;
private byte[] versionsBytes;
private short gfversion;
public int getSize() {
return this.size;
}
public String debugStr() {
StringBuilder sb = new StringBuilder();
sb.append(" opcode=").append(this.opCode).append(" len=").append(this.value.getLength())
.append(" vb=").append(this.value.getBytesAsString());
return sb.toString();
}
private void write(OplogFile olf, ValueWrapper vw) throws IOException {
vw.sendTo(olf.writeBuf, Oplog.this);
}
private void write(OplogFile olf, byte[] bytes, int byteLength) throws IOException {
int offset = 0;
ByteBuffer bb = olf.writeBuf;
while (offset < byteLength) {
int bytesThisTime = byteLength - offset;
boolean needsFlush = false;
if (bytesThisTime > bb.remaining()) {
needsFlush = true;
bytesThisTime = bb.remaining();
}
bb.put(bytes, offset, bytesThisTime);
offset += bytesThisTime;
if (needsFlush) {
flushNoSync(olf);
}
}
}
private void writeByte(OplogFile olf, byte v) throws IOException {
ByteBuffer bb = olf.writeBuf;
if (1 > bb.remaining()) {
flushNoSync(olf);
}
bb.put(v);
}
private void writeOrdinal(OplogFile olf, short ordinal) throws IOException {
ByteBuffer bb = olf.writeBuf;
if (3 > bb.remaining()) {
flushNoSync(olf);
}
// don't compress since we setup fixed size of buffers
VersioningIO.writeOrdinal(bb, ordinal, false);
}
private void writeInt(OplogFile olf, int v) throws IOException {
ByteBuffer bb = olf.writeBuf;
if (4 > bb.remaining()) {
flushNoSync(olf);
}
bb.putInt(v);
}
private void writeLong(OplogFile olf, long v) throws IOException {
ByteBuffer bb = olf.writeBuf;
if (8 > bb.remaining()) {
flushNoSync(olf);
}
bb.putLong(v);
}
public void initialize(long newEntryBase) {
this.opCode = OPLOG_NEW_ENTRY_BASE_ID;
this.newEntryBase = newEntryBase;
this.size = OPLOG_NEW_ENTRY_BASE_REC_SIZE;
}
public void initialize(short gfversion) {
this.opCode = OPLOG_GEMFIRE_VERSION;
this.gfversion = gfversion;
this.size = OPLOG_GEMFIRE_VERSION_REC_SIZE;
}
public void initialize(DiskStoreID diskStoreId) {
this.opCode = OPLOG_DISK_STORE_ID;
this.diskStoreId = diskStoreId;
this.size = OPLOG_DISK_STORE_REC_SIZE;
}
public void initialize(OPLOG_TYPE magic) {
this.opCode = OPLOG_MAGIC_SEQ_ID;
this.magic = magic;
this.size = OPLOG_MAGIC_SEQ_REC_SIZE;
}
public void initialize(Map<Long, AbstractDiskRegion> drMap, boolean gcRVV) throws IOException {
this.opCode = OPLOG_RVV;
byte[] rvvBytes = serializeRVVs(drMap, gcRVV);
this.value = new DiskEntry.Helper.ByteArrayValueWrapper(true, rvvBytes);
// Size is opCode + length + end of record
this.size = 1 + rvvBytes.length + 1;
}
public void initialize(long oplogKeyId, byte[] keyBytes, byte[] valueBytes, byte userBits,
long drId, VersionTag tag, boolean notToUseUserBits) throws IOException {
this.opCode = OPLOG_MOD_ENTRY_WITH_KEY_1ID;
this.size = 1;// for the opcode
saveUserBits(notToUseUserBits, userBits);
this.keyBytes = keyBytes;
this.value = new DiskEntry.Helper.CompactorValueWrapper(valueBytes, valueBytes.length);
if (this.userBits == 1 && this.value.getLength() == 0) {
throw new IllegalStateException("userBits==1 and valueLength is 0");
}
this.needsValue = EntryBits.isNeedsValue(this.userBits);
this.size += (4 + this.keyBytes.length);
saveDrId(drId);
initVersionsBytes(tag);
if (this.needsValue) {
this.size += 4 + this.value.getLength();
}
this.deltaIdBytesLength = 0;
{
long delta = calcDelta(oplogKeyId, this.opCode);
this.deltaIdBytesLength = bytesNeeded(delta);
this.size += this.deltaIdBytesLength;
this.opCode += (byte) ((this.deltaIdBytesLength & 0xFF) - 1);
for (int i = this.deltaIdBytesLength - 1; i >= 0; i--) {
this.deltaIdBytes[i] = (byte) (delta & 0xFF);
delta >>= 8;
}
}
this.size++; // for END_OF_RECORD_ID
}
private void initVersionsBytes(VersionTag tag) throws IOException {
if (EntryBits.isWithVersions(this.userBits)) {
this.versionsBytes = serializeVersionTag(tag);
this.size += this.versionsBytes.length;
}
}
private void initVersionsBytes(DiskEntry entry) throws IOException {
// persist entry version, region version and memberId
// The versions in entry are initialized to 0. So we will not persist the
// 3
// types of data if region version is 0.
// TODO: This method will be called 2 times, one for persisting into crf
// another for persisting into krf, since we did not save the byte arrary
// for the verstion tag.
VersionStamp stamp = entry.getVersionStamp();
if (EntryBits.isWithVersions(this.userBits)) {
assert (stamp != null);
this.versionsBytes = serializeVersionTag(stamp);
this.size += this.versionsBytes.length;
}
}
public void initialize(byte opCode, DiskRegionView dr, DiskEntry entry, ValueWrapper value,
byte userBits, boolean notToUseUserBits) throws IOException {
this.opCode = opCode;
this.size = 1;// for the opcode
saveUserBits(notToUseUserBits, userBits);
this.value = value;
if (this.userBits == 1 && this.value.getLength() == 0) {
throw new IllegalStateException("userBits==1 and valueLength is 0");
}
boolean needsKey = false;
if (this.opCode == OPLOG_MOD_ENTRY_1ID) {
if (modNeedsKey(entry)) {
needsKey = true;
this.opCode = OPLOG_MOD_ENTRY_WITH_KEY_1ID;
}
this.needsValue = EntryBits.isNeedsValue(this.userBits);
initVersionsBytes(entry);
} else if (this.opCode == OPLOG_NEW_ENTRY_0ID) {
needsKey = true;
this.needsValue = EntryBits.isNeedsValue(this.userBits);
initVersionsBytes(entry);
} else if (this.opCode == OPLOG_DEL_ENTRY_1ID) {
needsKey = false;
this.needsValue = false;
}
if (needsKey) {
Object key = entry.getKey();
this.keyBytes = EntryEventImpl.serialize(key);
this.size += (4 + this.keyBytes.length);
} else {
this.keyBytes = null;
}
if (this.opCode == OPLOG_DEL_ENTRY_1ID) {
this.drIdLength = 0;
} else {
long drId = dr.getId();
saveDrId(drId);
}
if (this.needsValue) {
this.size += 4 + this.value.getLength();
}
this.deltaIdBytesLength = 0;
if (this.opCode != OPLOG_NEW_ENTRY_0ID) {
// if (this.opCode == OPLOG_DEL_ENTRY_1ID) {
// this.newEntryBase =
// writeDelEntryId/*abs(entry.getDiskId().getKeyId())*/; this.size += 8;
// // HACK DEBUG
// } else {
// this.newEntryBase =
// writeModEntryId/*abs(entry.getDiskId().getKeyId())*/; this.size += 8;
// // HACK DEBUG
// }
long keyId = entry.getDiskId().getKeyId();
if (keyId == 0) {
Assert.fail("Attempting to write an entry with keyId=0 to oplog. Entry key="
+ entry.getKey() + " diskId=" + entry.getDiskId() + " region=" + dr);
}
long delta = calcDelta(keyId, this.opCode);
this.deltaIdBytesLength = bytesNeeded(delta);
this.size += this.deltaIdBytesLength;
this.opCode += (byte) ((this.deltaIdBytesLength & 0xFF) - 1);
for (int i = this.deltaIdBytesLength - 1; i >= 0; i--) {
this.deltaIdBytes[i] = (byte) (delta & 0xFF);
delta >>= 8;
}
}
this.size++; // for END_OF_RECORD_ID
}
private void saveUserBits(boolean notToUseUserBits, byte userBits) {
this.notToUseUserBits = notToUseUserBits;
if (notToUseUserBits) {
this.userBits = 0;
} else {
this.userBits = EntryBits.getPersistentBits(userBits);
this.size++; // for the userBits
}
}
private void saveDrId(long drId) {
// If the drId is <= 255 (max unsigned byte) then
// encode it as a single byte.
// Otherwise write a byte whose value is the number of bytes
// it will be encoded by and then follow it with that many bytes.
// Note that drId are not allowed to have a value in the range 1..8
// inclusive.
if (drId >= 0 && drId <= 255) {
this.drIdLength = 1;
this.drIdBytes[0] = (byte) drId;
} else {
byte bytesNeeded = bytesNeeded(drId);
this.drIdLength = bytesNeeded + 1;
this.drIdBytes[0] = bytesNeeded;
for (int i = bytesNeeded; i >= 1; i--) {
this.drIdBytes[i] = (byte) (drId & 0xFF);
drId >>= 8;
}
}
this.size += this.drIdLength;
}
public void initialize(byte opCode, long drId, VersionTag tag) throws IOException {
this.opCode = opCode;
assert this.opCode == OPLOG_CONFLICT_VERSION;
this.size = 1;// for the opcode
saveDrId(drId);
this.versionsBytes = serializeVersionTag(tag);
this.size += this.versionsBytes.length;
this.size++; // for END_OF_RECORD_ID
}
/**
* Returns the offset to the first byte of the value bytes.
*/
public int getValueOffset() {
if (!this.needsValue)
return 0;
int result = this.deltaIdBytesLength
// + 8 /* HACK DEBUG */
+ this.drIdLength + 1/* opcode */
+ 4/* value length */;
if (this.notToUseUserBits == false) {
result++;
}
if (EntryBits.isWithVersions(this.userBits) && this.versionsBytes != null) {
result += this.versionsBytes.length;
}
return result;
}
public long write(OplogFile olf) throws IOException {
long bytesWritten = 0;
writeByte(olf, this.opCode);
bytesWritten++;
if (this.opCode == OPLOG_NEW_ENTRY_BASE_ID) {
writeLong(olf, this.newEntryBase);
bytesWritten += 8;
} else if (this.opCode == OPLOG_DISK_STORE_ID) {
writeLong(olf, this.diskStoreId.getLeastSignificantBits());
writeLong(olf, this.diskStoreId.getMostSignificantBits());
bytesWritten += 16;
} else if (this.opCode == OPLOG_MAGIC_SEQ_ID) {
write(olf, this.magic.getBytes(), OPLOG_TYPE.getLen());
bytesWritten += OPLOG_TYPE.getLen();
} else if (this.opCode == OPLOG_RVV) {
write(olf, this.value);
bytesWritten += this.value.getLength();
} else if (this.opCode == OPLOG_GEMFIRE_VERSION) {
writeOrdinal(olf, this.gfversion);
bytesWritten++;
} else if (this.opCode == OPLOG_CONFLICT_VERSION) {
if (this.drIdLength > 0) {
write(olf, this.drIdBytes, this.drIdLength);
bytesWritten += this.drIdLength;
}
if (this.versionsBytes != null && this.versionsBytes.length > 0) {
write(olf, this.versionsBytes, this.versionsBytes.length);
bytesWritten += this.versionsBytes.length;
}
} else {
if (!this.notToUseUserBits) {
writeByte(olf, this.userBits);
bytesWritten++;
}
if (this.deltaIdBytesLength > 0) {
write(olf, this.deltaIdBytes, this.deltaIdBytesLength);
bytesWritten += this.deltaIdBytesLength;
// writeLong(olf, this.newEntryBase); bytesWritten += 8; // HACK DEBUG
}
if (this.drIdLength > 0) {
write(olf, this.drIdBytes, this.drIdLength);
bytesWritten += this.drIdLength;
}
if (EntryBits.isWithVersions(this.userBits) && this.versionsBytes != null
&& this.opCode != OPLOG_DEL_ENTRY_1ID) {
write(olf, this.versionsBytes, this.versionsBytes.length);
bytesWritten += this.versionsBytes.length;
}
if (this.needsValue) {
int len = this.value.getLength();
writeInt(olf, len);
bytesWritten += 4;
if (len > 0) {
write(olf, this.value);
bytesWritten += len;
}
}
if (this.keyBytes != null) {
writeInt(olf, this.keyBytes.length);
bytesWritten += 4;
if (this.keyBytes.length > 0) {
write(olf, this.keyBytes, this.keyBytes.length);
bytesWritten += this.keyBytes.length;
}
}
}
writeByte(olf, END_OF_RECORD_ID);
bytesWritten++;
return bytesWritten;
}
/**
* Free up any references to possibly large data.
*/
public void clear() {
this.value = null;
this.keyBytes = null;
this.notToUseUserBits = false;
this.versionsBytes = null;
}
}
/**
* Fake disk entry used to implement the circular linked list of entries an oplog has. Each Oplog
* will have one OplogDiskEntry whose prev and next fields point to the actual DiskEntrys
* currently stored in its crf. Items are added at "next" so the most recent entry written will be
* at next and the oldest item written will be at "prev".
*/
static class OplogDiskEntry implements DiskEntry, RegionEntry {
private DiskEntry next = this;
private DiskEntry prev = this;
@Override
public synchronized DiskEntry getPrev() {
return this.prev;
}
@Override
public synchronized void setPrev(DiskEntry v) {
this.prev = v;
}
@Override
public synchronized DiskEntry getNext() {
return this.next;
}
@Override
public synchronized void setNext(DiskEntry v) {
this.next = v;
}
/**
* returns the number of entries cleared
*/
public synchronized int clear(RegionVersionVector rvv,
Map<DiskEntry, VersionHolder> pendingKrfTags) {
if (rvv == null) {
if (pendingKrfTags != null) {
pendingKrfTags.clear();
}
return clear();
} else {
// Clearing the list is handled in AbstractRegionMap.clear for RVV
// based clears, because it removes each entry.
// It needs to be handled there because the entry is synched at that
// point
return 0;
}
}
/**
* Clear using an RVV. Remove live entries that are contained within the clear RVV.
*/
private int clearWithRVV(RegionVersionVector rvv, Map<DiskEntry, VersionTag> pendingKrfTags) {
// TODO this doesn't work, because we can end up removing entries from
// here before
// they are removed from the region map. Reverting this to the old, leaky,
// behavior
// until I fix the region map code.
return 0;
}
/**
* Clear without an RVV. Empties the entire list.
*/
private int clear() {
int result = 0;
// Need to iterate over the list and set each prev field to null
// so that if remove is called it will know that the DiskEntry
// has already been removed.
DiskEntry n = getNext();
setNext(this);
setPrev(this);
while (n != this) {
result++;
n.setPrev(null);
n = n.getNext();
}
return result;
}
public synchronized boolean remove(DiskEntry v) {
DiskEntry p = v.getPrev();
if (p != null) {
v.setPrev(null);
DiskEntry n = v.getNext();
v.setNext(null);
n.setPrev(p);
p.setNext(n);
return true;
} else {
return false;
}
}
public synchronized void insert(DiskEntry v) {
assert v.getPrev() == null;
// checkForDuplicate(v);
DiskEntry n = getNext();
setNext(v);
n.setPrev(v);
v.setNext(n);
v.setPrev(this);
}
public synchronized void replace(DiskEntry old, DiskEntry v) {
DiskEntry p = old.getPrev();
if (p != null) {
old.setPrev(null);
v.setPrev(p);
p.setNext(v);
}
DiskEntry n = old.getNext();
if (n != null) {
old.setNext(null);
v.setNext(n);
n.setPrev(v);
}
if (getNext() == old) {
setNext(v);
}
}
@Override
public Object getKey() {
throw new IllegalStateException();
}
@Override
public Object getValue() {
throw new IllegalStateException();
}
@Override
public Token getValueAsToken() {
throw new IllegalStateException();
}
@Override
public void handleValueOverflow(RegionEntryContext context) {
throw new IllegalStateException();
}
@Override
public Object prepareValueForCache(RegionEntryContext context, Object value,
boolean isEntryUpdate) {
throw new IllegalStateException("Should never be called");
}
@Override
public Object getValueRetain(RegionEntryContext context, boolean decompress) {
throw new IllegalStateException();
}
@Override
public void setValueWithContext(RegionEntryContext context, Object value) {
throw new IllegalStateException();
}
public void _removePhase1() {
throw new IllegalStateException();
}
@Override
public DiskId getDiskId() {
throw new IllegalStateException();
}
@Override
public long getLastModified() {
throw new IllegalStateException();
}
public boolean isRecovered() {
throw new IllegalStateException();
}
@Override
public boolean isValueNull() {
throw new IllegalStateException();
}
@Override
public boolean isRemovedFromDisk() {
throw new IllegalStateException();
}
@Override
public int updateAsyncEntrySize(EvictionController capacityController) {
throw new IllegalStateException();
}
/**
* Adds any live entries in this list to liveEntries and returns the index of the next free
* slot.
*
* @param liveEntries the array to fill with the live entries
* @param idx the first free slot in liveEntries
* @param drv the disk region these entries are on
* @return the next free slot in liveEntries
*/
public synchronized int addLiveEntriesToList(KRFEntry[] liveEntries, int idx,
DiskRegionView drv, Map<DiskEntry, VersionHolder> pendingKrfTags) {
DiskEntry de = getPrev();
while (de != this) {
VersionHolder tag = null;
if (pendingKrfTags != null) {
tag = pendingKrfTags.get(de);
}
liveEntries[idx] = new KRFEntry(drv, de, tag);
idx++;
de = de.getPrev();
}
return idx;
}
/*
* (non-Javadoc)
*
* @see org.apache.geode.internal.cache.DiskEntry#getVersionStamp()
*/
@Override
public VersionStamp getVersionStamp() {
// dummy entry as start of live list
return null;
}
@Override
public boolean hasStats() {
return false;
}
@Override
public long getLastAccessed() throws InternalStatisticsDisabledException {
return 0;
}
@Override
public long getHitCount() throws InternalStatisticsDisabledException {
return 0;
}
@Override
public long getMissCount() throws InternalStatisticsDisabledException {
return 0;
}
@Override
public void updateStatsForPut(long lastModifiedTime, long lastAccessedTime) {
// nothing
}
@Override
public VersionTag generateVersionTag(VersionSource member, boolean withDelta,
InternalRegion region, EntryEventImpl event) {
return null;
}
@Override
public boolean dispatchListenerEvents(EntryEventImpl event) throws InterruptedException {
return false;
}
@Override
public void setRecentlyUsed(RegionEntryContext context) {
// nothing
}
@Override
public void updateStatsForGet(boolean hit, long time) {
// nothing
}
@Override
public void txDidDestroy(long currentTime) {
// nothing
}
@Override
public void resetCounts() throws InternalStatisticsDisabledException {
// nothing
}
@Override
public void makeTombstone(InternalRegion region, VersionTag version)
throws RegionClearedException {
// nothing
}
@Override
public void removePhase1(InternalRegion region, boolean clear) throws RegionClearedException {
// nothing
}
@Override
public void removePhase2() {
// nothing
}
@Override
public boolean isRemoved() {
return false;
}
@Override
public boolean isRemovedPhase2() {
return false;
}
@Override
public boolean isTombstone() {
return false;
}
@Override
public boolean fillInValue(InternalRegion region, InitialImageOperation.Entry entry,
ByteArrayDataInput in, DistributionManager distributionManager,
final KnownVersion version) {
return false;
}
@Override
public boolean isOverflowedToDisk(InternalRegion region, DiskPosition diskPosition) {
return false;
}
@Override
public Object getValue(RegionEntryContext context) {
return null;
}
@Override
public Object getValueRetain(RegionEntryContext context) {
return null;
}
@Override
public void setValue(RegionEntryContext context, Object value) throws RegionClearedException {
// nothing
}
@Override
public void setValueWithTombstoneCheck(Object value, EntryEvent event)
throws RegionClearedException {
// nothing
}
@Override
public Object getTransformedValue() {
return null;
}
@Override
public Object getValueInVM(RegionEntryContext context) {
return null;
}
@Override
public Object getValueOnDisk(InternalRegion region) throws EntryNotFoundException {
return null;
}
@Override
public Object getValueOnDiskOrBuffer(InternalRegion region) throws EntryNotFoundException {
return null;
}
@Override
public boolean initialImagePut(InternalRegion region, long lastModified, Object newValue,
boolean wasRecovered, boolean acceptedVersionTag) throws RegionClearedException {
return false;
}
@Override
public boolean initialImageInit(InternalRegion region, long lastModified, Object newValue,
boolean create, boolean wasRecovered, boolean acceptedVersionTag)
throws RegionClearedException {
return false;
}
@Override
public boolean destroy(InternalRegion region, EntryEventImpl event, boolean inTokenMode,
boolean cacheWrite, Object expectedOldValue, boolean forceDestroy,
boolean removeRecoveredEntry) throws CacheWriterException, EntryNotFoundException,
TimeoutException, RegionClearedException {
return false;
}
@Override
public boolean getValueWasResultOfSearch() {
return false;
}
@Override
public void setValueResultOfSearch(boolean value) {
// nothing
}
@Override
public Object getSerializedValueOnDisk(InternalRegion region) {
return null;
}
@Override
public Object getValueInVMOrDiskWithoutFaultIn(InternalRegion region) {
return null;
}
@Override
public Object getValueOffHeapOrDiskWithoutFaultIn(InternalRegion region) {
return null;
}
@Override
public boolean isUpdateInProgress() {
return false;
}
@Override
public void setUpdateInProgress(boolean underUpdate) {
// nothing
}
@Override
public boolean isInvalid() {
return false;
}
@Override
public boolean isDestroyed() {
return false;
}
@Override
public boolean isDestroyedOrRemoved() {
return false;
}
@Override
public boolean isDestroyedOrRemovedButNotTombstone() {
return false;
}
@Override
public boolean isInvalidOrRemoved() {
return false;
}
@Override
public void setValueToNull() {
// nothing
}
@Override
public void returnToPool() {
// nothing
}
@Override
public boolean isCacheListenerInvocationInProgress() {
return false;
}
@Override
public void setCacheListenerInvocationInProgress(boolean isListenerInvoked) {
// nothing
}
@Override
public void setValue(RegionEntryContext context, Object value, EntryEventImpl event)
throws RegionClearedException {}
@Override
public boolean isInUseByTransaction() {
return false;
}
@Override
public void incRefCount() {
// nothing
}
@Override
public void decRefCount(EvictionList lruList, InternalRegion region) {
// nothing
}
@Override
public void resetRefCount(EvictionList lruList) {
// nothing
}
@Override
public Object prepareValueForCache(RegionEntryContext context, Object value,
EntryEventImpl event, boolean isEntryUpdate) {
throw new IllegalStateException("Should never be called");
}
@Override
public boolean isEvicted() {
return false;
}
}
/**
* Used as the value in the regionMap. Tracks information about what the region has in this oplog.
*/
private interface DiskRegionInfo {
DiskRegionView getDiskRegion();
int addLiveEntriesToList(KRFEntry[] liveEntries, int idx);
void addLive(DiskEntry de);
void update(DiskEntry entry);
void replaceLive(DiskEntry old, DiskEntry de);
boolean rmLive(DiskEntry de, Oplog oplog);
DiskEntry getNextLiveEntry();
void setDiskRegion(DiskRegionView dr);
long clear(RegionVersionVector rvv);
/**
* Return true if we are the first to set it to true
*/
boolean testAndSetUnrecovered();
boolean getUnrecovered();
/**
* Return true if we are the first to set it to false
*/
boolean testAndSetRecovered(DiskRegionView dr);
/**
* Callback to indicate that this oplog has created a krf.
*/
void afterKrfCreated();
}
private abstract static class AbstractDiskRegionInfo implements DiskRegionInfo {
private DiskRegionView dr;
private boolean unrecovered = false;
public AbstractDiskRegionInfo(DiskRegionView dr) {
this.dr = dr;
}
@Override
public abstract void addLive(DiskEntry de);
@Override
public abstract DiskEntry getNextLiveEntry();
@Override
public abstract long clear(RegionVersionVector rvv);
@Override
public DiskRegionView getDiskRegion() {
return this.dr;
}
@Override
public void setDiskRegion(DiskRegionView dr) {
this.dr = dr;
}
@Override
public synchronized boolean testAndSetUnrecovered() {
boolean result = !this.unrecovered;
if (result) {
this.unrecovered = true;
this.dr = null;
}
return result;
}
@Override
public synchronized boolean getUnrecovered() {
return this.unrecovered;
}
@Override
public synchronized boolean testAndSetRecovered(DiskRegionView dr) {
boolean result = this.unrecovered;
if (result) {
this.unrecovered = false;
this.dr = dr;
}
return result;
}
}
private static class DiskRegionInfoNoList extends AbstractDiskRegionInfo {
private final AtomicInteger liveCount = new AtomicInteger();
public DiskRegionInfoNoList(DiskRegionView dr) {
super(dr);
}
@Override
public void addLive(DiskEntry de) {
this.liveCount.incrementAndGet();
}
@Override
public void update(DiskEntry entry) {
// nothing to do
}
@Override
public void replaceLive(DiskEntry old, DiskEntry de) {}
@Override
public boolean rmLive(DiskEntry de, Oplog oplog) {
return this.liveCount.decrementAndGet() >= 0;
}
@Override
public DiskEntry getNextLiveEntry() {
return null;
}
@Override
public long clear(RegionVersionVector rvv) {
return this.liveCount.getAndSet(0);
}
@Override
public int addLiveEntriesToList(KRFEntry[] liveEntries, int idx) {
// nothing needed since no linked list
return idx;
}
@Override
public void afterKrfCreated() {
// do nothing
}
}
private static class DiskRegionInfoWithList extends AbstractDiskRegionInfo {
/**
* A linked list of the live entries in this oplog. Updates to pendingKrfTags are protected by
* synchronizing on object.
*/
private final OplogDiskEntry liveEntries = new OplogDiskEntry();
/**
* A map of DiskEntry to the VersionTag that is written to disk associated with this tag. Only
* needed for async regions so that we can generate a krf with a version tag that matches the
* the tag we have written to disk for this oplog.
*/
private Map<DiskEntry, VersionHolder> pendingKrfTags;
public DiskRegionInfoWithList(DiskRegionView dr, boolean couldHaveKrf, boolean krfExists) {
super(dr);
// we need to keep track of the version tags for entries so that we write
// the correct entry to the krf
// both in sync and async disk write cases
if (!krfExists && couldHaveKrf) {
pendingKrfTags = new HashMap<DiskEntry, VersionHolder>(200);
} else {
pendingKrfTags = null;
}
}
@Override
public void addLive(DiskEntry de) {
synchronized (liveEntries) {
this.liveEntries.insert(de);
if (pendingKrfTags != null && de.getVersionStamp() != null) {
// Remember the version tag of the entry as it was written to the crf.
pendingKrfTags.put(de, new CompactVersionHolder(de.getVersionStamp()));
}
}
}
@Override
public void update(DiskEntry entry) {
if (pendingKrfTags != null && entry.getVersionStamp() != null) {
// Remember the version tag of the entry as it was written to the crf.
pendingKrfTags.put(entry, new CompactVersionHolder(entry.getVersionStamp()));
}
}
@Override
public void replaceLive(DiskEntry old, DiskEntry de) {
synchronized (liveEntries) {
this.liveEntries.replace(old, de);
if (pendingKrfTags != null && de.getVersionStamp() != null) {
// Remember the version tag of the entry as it was written to the crf.
pendingKrfTags.remove(old);
pendingKrfTags.put(de, new CompactVersionHolder(de.getVersionStamp()));
}
}
}
@Override
public boolean rmLive(DiskEntry de, Oplog oplog) {
synchronized (liveEntries) {
boolean removed = this.liveEntries.remove(de);
if (removed && pendingKrfTags != null) {
pendingKrfTags.remove(de);
}
if (removed) {
oplog.decLiveCount();
}
return removed;
}
}
@Override
public DiskEntry getNextLiveEntry() {
DiskEntry result = this.liveEntries.getPrev();
if (result == this.liveEntries) {
result = null;
}
return result;
}
@Override
public long clear(RegionVersionVector rvv) {
synchronized (this.liveEntries) {
return this.liveEntries.clear(rvv, this.pendingKrfTags);
}
}
/**
* Return true if we are the first to set it to true
*/
@Override
public synchronized boolean testAndSetUnrecovered() {
boolean result = super.testAndSetUnrecovered();
if (result) {
this.liveEntries.clear();
}
return result;
}
@Override
public int addLiveEntriesToList(KRFEntry[] liveEntries, int idx) {
synchronized (liveEntries) {
return this.liveEntries.addLiveEntriesToList(liveEntries, idx, getDiskRegion(),
pendingKrfTags);
}
}
@Override
public void afterKrfCreated() {
synchronized (liveEntries) {
this.pendingKrfTags = null;
}
}
}
/**
* Used during offline compaction to hold information that may need to be copied forward.
*/
private static class CompactionRecord {
private final byte[] keyBytes;
private long offset;
public CompactionRecord(byte[] kb, long offset) {
this.keyBytes = kb;
this.offset = offset;
}
public void update(long offset) {
this.offset = offset;
}
public byte[] getKeyBytes() {
return this.keyBytes;
}
public long getOffset() {
return this.offset;
}
}
/**
* Mpa of OplogEntryIds (longs). Memory is optimized by using an int[] for ids in the unsigned int
* range.
*/
static class OplogEntryIdMap {
private final Int2ObjectOpenHashMap ints =
new Int2ObjectOpenHashMap((int) DiskStoreImpl.INVALID_ID);
private final Long2ObjectOpenHashMap longs =
new Long2ObjectOpenHashMap((int) DiskStoreImpl.INVALID_ID);
public Object put(long id, Object v) {
Object result;
if (id == 0) {
throw new IllegalArgumentException();
} else if (id > 0 && id <= 0x00000000FFFFFFFFL) {
result = this.ints.put((int) id, v);
} else {
result = this.longs.put(id, v);
}
return result;
}
public int size() {
return this.ints.size() + this.longs.size();
}
public Object get(long id) {
Object result;
if (id >= 0 && id <= 0x00000000FFFFFFFFL) {
result = this.ints.get((int) id);
} else {
result = this.longs.get(id);
}
return result;
}
public Iterator iterator() {
return new Iterator();
}
class Iterator {
private boolean doingInt = true;
ObjectIterator<Int2ObjectMap.Entry<?>> intIt = ints.int2ObjectEntrySet().fastIterator();
ObjectIterator<Long2ObjectMap.Entry<?>> longIt = longs.long2ObjectEntrySet().fastIterator();
Int2ObjectMap.Entry<?> nextIntEntry;
Long2ObjectMap.Entry<?> nextLongEntry;
public boolean hasNext() {
if (this.intIt.hasNext()) {
return true;
} else {
doingInt = false;
return this.longIt.hasNext();
}
}
public void advance() {
if (doingInt) {
nextIntEntry = intIt.next();
} else {
nextLongEntry = longIt.next();
}
}
public long key() {
if (doingInt) {
return nextIntEntry.getKey();
} else {
return nextLongEntry.getKey();
}
}
public Object value() {
if (doingInt) {
return nextIntEntry.getValue();
} else {
return nextLongEntry.getValue();
}
}
}
}
/**
* Used in offline mode to prevent pdx deserialization of keys. The raw bytes are a serialized
* pdx.
*
* @since GemFire 6.6
*/
private static class RawByteKey implements Sendable {
final byte[] bytes;
final int hashCode;
public RawByteKey(byte[] keyBytes) {
this.bytes = keyBytes;
this.hashCode = Arrays.hashCode(keyBytes);
}
@Override
public int hashCode() {
return this.hashCode;
}
@Override
public boolean equals(Object obj) {
if (!(obj instanceof RawByteKey)) {
return false;
}
return Arrays.equals(this.bytes, ((RawByteKey) obj).bytes);
}
@Override
public void sendTo(DataOutput out) throws IOException {
out.write(this.bytes);
}
}
}