blob: ac93cdc45dac604f44f1b40c5a2a8a7437b06693 [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.partitioned;
import java.io.ByteArrayInputStream;
import java.io.DataInput;
import java.io.DataInputStream;
import java.io.DataOutput;
import java.io.IOException;
import java.util.Collections;
import java.util.HashMap;
import java.util.Iterator;
import java.util.Map;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import org.apache.logging.log4j.Logger;
import org.apache.geode.CancelException;
import org.apache.geode.DataSerializer;
import org.apache.geode.annotations.Immutable;
import org.apache.geode.cache.CacheException;
import org.apache.geode.distributed.internal.ClusterDistributionManager;
import org.apache.geode.distributed.internal.DistributionManager;
import org.apache.geode.distributed.internal.DistributionMessage;
import org.apache.geode.distributed.internal.DistributionStats;
import org.apache.geode.distributed.internal.InternalDistributedSystem;
import org.apache.geode.distributed.internal.ReplyException;
import org.apache.geode.distributed.internal.ReplyMessage;
import org.apache.geode.distributed.internal.ReplyProcessor21;
import org.apache.geode.distributed.internal.membership.InternalDistributedMember;
import org.apache.geode.internal.Assert;
import org.apache.geode.internal.HeapDataOutputStream;
import org.apache.geode.internal.Version;
import org.apache.geode.internal.cache.BucketDump;
import org.apache.geode.internal.cache.BucketRegion;
import org.apache.geode.internal.cache.CachedDeserializable;
import org.apache.geode.internal.cache.ForceReattemptException;
import org.apache.geode.internal.cache.InitialImageOperation;
import org.apache.geode.internal.cache.LocalRegion;
import org.apache.geode.internal.cache.NonTXEntry;
import org.apache.geode.internal.cache.PartitionedRegion;
import org.apache.geode.internal.cache.PartitionedRegionDataStore;
import org.apache.geode.internal.cache.RegionEntry;
import org.apache.geode.internal.cache.Token;
import org.apache.geode.internal.cache.versions.RegionVersionVector;
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.LogService;
import org.apache.geode.internal.logging.log4j.LogMarker;
import org.apache.geode.internal.offheap.OffHeapHelper;
import org.apache.geode.internal.util.ObjectIntProcedure;
public class FetchEntriesMessage extends PartitionMessage {
private static final Logger logger = LogService.getLogger();
private int bucketId;
public FetchEntriesMessage() {}
private FetchEntriesMessage(InternalDistributedMember recipient, int regionId,
ReplyProcessor21 processor, int bucketId) {
super(recipient, regionId, processor);
this.bucketId = bucketId;
}
/**
* Sends a PartitionedRegion message to fetch all the entries for a bucketId
*
* @param recipient the member that the fetch keys message is sent to
* @param r the PartitionedRegion that contains the bucket
* @param bucketId the identity of the bucket that contains the entries to be returned
* @return the processor used to read the returned entries
* @throws ForceReattemptException if the peer is no longer available
*/
public static FetchEntriesResponse send(InternalDistributedMember recipient, PartitionedRegion r,
int bucketId) throws ForceReattemptException {
Assert.assertTrue(recipient != null, "FetchEntriesMessage NULL reply message");
FetchEntriesResponse p = new FetchEntriesResponse(r.getSystem(), r, recipient, bucketId);
FetchEntriesMessage m = new FetchEntriesMessage(recipient, r.getPRId(), p, bucketId);
m.setTransactionDistributed(r.getCache().getTxManager().isDistributed());
Set failures = r.getDistributionManager().putOutgoing(m);
if (failures != null && failures.size() > 0) {
throw new ForceReattemptException(
String.format("Failed sending < %s >", m));
}
return p;
}
@Override
protected boolean operateOnPartitionedRegion(ClusterDistributionManager dm, PartitionedRegion pr,
long startTime) throws CacheException, ForceReattemptException {
if (logger.isTraceEnabled(LogMarker.DM_VERBOSE)) {
logger.trace(LogMarker.DM_VERBOSE, "FetchEntriesMessage operateOnRegion: {}",
pr.getFullPath());
}
PartitionedRegionDataStore ds = pr.getDataStore();
BucketRegion entries = null;
if (ds != null) {
entries = ds.handleRemoteGetEntries(this.bucketId);
if (logger.isTraceEnabled(LogMarker.DM_VERBOSE)) {
logger.trace(LogMarker.DM_VERBOSE, "FetchKeysMessage send keys back using processorId: {}",
getProcessorId());
}
} else {
logger.warn("FetchKeysMessage: data store not configured for this member");
}
pr.getPrStats().endPartitionMessagesProcessing(startTime);
FetchEntriesReplyMessage.send(getSender(), getProcessorId(), dm, this.bucketId, entries);
return false;
}
@Override
protected void appendFields(StringBuilder buff) {
super.appendFields(buff);
buff.append("; bucketId=").append(this.bucketId);
buff.append("; recipient=").append(this.getRecipient());
}
@Override
public int getDSFID() {
return PR_FETCH_ENTRIES_MESSAGE;
}
@Override
public void fromData(DataInput in) throws IOException, ClassNotFoundException {
super.fromData(in);
this.bucketId = in.readInt();
}
@Override
public void toData(DataOutput out) throws IOException {
super.toData(out);
out.writeInt(this.bucketId);
}
public static class FetchEntriesReplyMessage extends ReplyMessage {
/** The bucket id */
int bucketId;
/** The number of the series */
int seriesNum;
/** The message number in the series */
int msgNum;
/** The total number of series */
int numSeries;
/** Whether this is the last of a series */
boolean lastInSeries;
/** array holding chunk of entries received */
transient byte[] chunk;
/** stream holding chunk of entries to send */
transient HeapDataOutputStream chunkStream;
private boolean hasRVV;
/** The versions in which this message was modified */
@Immutable
private static final Version[] dsfidVersions = null;
@Override
public Version[] getSerializationVersions() {
return dsfidVersions;
}
/**
* Empty constructor to conform to DataSerializable interface
*/
public FetchEntriesReplyMessage() {}
protected FetchEntriesReplyMessage(InternalDistributedMember dest, int processorId, int buckId,
HeapDataOutputStream chunk, int seriesNum, int msgNum, int numSeries, boolean lastInSeries,
boolean hasRVV) {
setRecipient(dest);
setProcessorId(processorId);
this.bucketId = buckId;
this.seriesNum = seriesNum;
this.msgNum = msgNum;
this.numSeries = numSeries;
this.lastInSeries = lastInSeries;
this.chunkStream = chunk;
this.hasRVV = hasRVV;
}
/**
* Send an ack
*
* @throws ForceReattemptException if the peer is no longer available
*/
public static void send(final InternalDistributedMember recipient, final int processorId,
final DistributionManager dm, final int bucketId, BucketRegion keys)
throws ForceReattemptException {
Assert.assertTrue(recipient != null, "FetchEntriesReplyMessage NULL reply message");
final int numSeries = 1;
final int seriesNum = 0;
final RegionVersionVector rvv = keys.getVersionVector();
if (rvv != null) {
RegionVersionVector clone = rvv.getCloneForTransmission();
ReplyMessage.send(recipient, processorId, clone, dm);
}
// chunkEntries returns false if didn't finish
if (logger.isDebugEnabled()) {
logger.debug("Starting PR entries chunking for {} entries", keys.size());
}
try {
boolean finished = chunkMap(recipient, keys, InitialImageOperation.CHUNK_SIZE_IN_BYTES,
false, new ObjectIntProcedure() {
int msgNum = 0;
boolean last = false;
/**
* @param a byte[] chunk
* @param b positive if last chunk
* @return true to continue to next chunk
*/
@Override
public boolean executeWith(Object a, int b) {
HeapDataOutputStream chunk = (HeapDataOutputStream) a;
this.last = b > 0;
try {
boolean okay = sendChunk(recipient, processorId, bucketId, dm, chunk, seriesNum,
msgNum++, numSeries, this.last, rvv != null);
return okay;
} catch (CancelException e) {
return false;
}
}
});
if (logger.isDebugEnabled()) {
logger.debug("{} PR entries chunking", (finished ? "Finished" : "DID NOT complete"));
}
} catch (IOException io) {
// This is a little odd, since we're trying to send a reply.
// One does not normally force a reply. Is this correct?
throw new ForceReattemptException(
"Unable to send response to fetch-entries request",
io);
}
}
static boolean sendChunk(InternalDistributedMember recipient, int processorId, int bucketId,
DistributionManager dm, HeapDataOutputStream chunk, int seriesNum, int msgNum,
int numSeries, boolean lastInSeries, boolean hasRVV) {
FetchEntriesReplyMessage reply = new FetchEntriesReplyMessage(recipient, processorId,
bucketId, chunk, seriesNum, msgNum, numSeries, lastInSeries, hasRVV);
Set failures = dm.putOutgoing(reply);
return (failures == null) || (failures.size() == 0);
}
/**
* Serialize the given map's entries into byte[] chunks, calling proc for each one. proc args:
* the byte[] chunk and an int indicating whether it is the last chunk (positive means last
* chunk, zero othewise). The return value of proc indicates whether to continue to the next
* chunk (true) or abort (false).
*
* @return true if finished all chunks, false if stopped early
*/
static boolean chunkMap(InternalDistributedMember receiver, BucketRegion map,
int CHUNK_SIZE_IN_BYTES, boolean includeValues, ObjectIntProcedure proc)
throws IOException {
Iterator it = map.entrySet().iterator();
boolean keepGoing = true;
boolean sentLastChunk = false;
// always write at least one chunk
final HeapDataOutputStream mos = new HeapDataOutputStream(
InitialImageOperation.CHUNK_SIZE_IN_BYTES + 2048, receiver.getVersionObject());
do {
mos.reset();
int avgItemSize = 0;
int itemCount = 0;
while ((mos.size() + avgItemSize) < InitialImageOperation.CHUNK_SIZE_IN_BYTES
&& it.hasNext()) {
NonTXEntry entry = (NonTXEntry) it.next();
RegionEntry re = entry.getRegionEntry();
synchronized (re) {
Object value = re.getValueRetain(map, true);
try {
if (value == null) {
// only possible for disk entry
value = re.getSerializedValueOnDisk((LocalRegion) entry.getRegion());
}
if (!Token.isRemoved(value)) {
DataSerializer.writeObject(re.getKey(), mos);
if (Token.isInvalid(value)) {
value = null;
}
VersionStamp stamp = re.getVersionStamp();
VersionTag versionTag = stamp != null ? stamp.asVersionTag() : null;
if (versionTag != null) {
versionTag.replaceNullIDs(map.getVersionMember());
}
DataSerializer.writeObject(value, mos);
DataSerializer.writeObject(versionTag, mos);
// Note we track the itemCount so we can compute avgItemSize
itemCount++;
// Note we track avgItemSize to help us not to always go one item
// past the max chunk size. When we go past it causes us to grow
// the ByteBuffer that the chunk is stored in resulting in a copy
// of the data.
avgItemSize = mos.size() / itemCount;
}
} finally {
OffHeapHelper.release(value);
}
}
}
// Write "end of chunk" entry to indicate end of chunk
DataSerializer.writeObject((Object) null, mos);
// send 1 for last message if no more data
int lastMsg = it.hasNext() ? 0 : 1;
keepGoing = proc.executeWith(mos, lastMsg);
sentLastChunk = lastMsg == 1 && keepGoing;
// if this region is destroyed while we are sending data, then abort.
} while (keepGoing && it.hasNext());
// return false if we were told to abort
return sentLastChunk;
}
/**
* Processes this message. This method is invoked by the receiver of the message.
*
* @param dm the distribution manager that is processing the message.
*/
@Override
public void process(final DistributionManager dm, final ReplyProcessor21 p) {
final long startTime = getTimestamp();
FetchEntriesResponse processor = (FetchEntriesResponse) p;
if (processor == null) {
if (logger.isTraceEnabled(LogMarker.DM_VERBOSE)) {
logger.trace(LogMarker.DM_VERBOSE, "FetchEntriesReplyMessage processor not found");
}
return;
}
processor.processChunk(this);
if (logger.isTraceEnabled(LogMarker.DM_VERBOSE)) {
logger.trace(LogMarker.DM_VERBOSE, "{} processed {}", processor, this);
}
dm.getStats().incReplyMessageTime(DistributionStats.getStatTime() - startTime);
}
@Override
public void toData(DataOutput out) throws IOException {
super.toData(out);
out.writeInt(this.bucketId);
out.writeInt(this.seriesNum);
out.writeInt(this.msgNum);
out.writeInt(this.numSeries);
out.writeBoolean(this.lastInSeries);
DataSerializer.writeObjectAsByteArray(this.chunkStream, out);
out.writeBoolean(this.hasRVV);
}
@Override
public int getDSFID() {
return PR_FETCH_ENTRIES_REPLY_MESSAGE;
}
@Override
public void fromData(DataInput in) throws IOException, ClassNotFoundException {
super.fromData(in);
this.bucketId = in.readInt();
this.seriesNum = in.readInt();
this.msgNum = in.readInt();
this.numSeries = in.readInt();
this.lastInSeries = in.readBoolean();
this.chunk = DataSerializer.readByteArray(in);
hasRVV = in.readBoolean();
}
@Override
public String toString() {
StringBuilder sb = new StringBuilder();
sb.append("FetchEntriesReplyMessage ").append("processorid=").append(this.processorId)
.append(",bucketId=").append(this.bucketId);
if (getSender() != null) {
sb.append(",sender=").append(this.getSender());
}
sb.append(",seriesNum=").append(seriesNum).append(",msgNum=").append(msgNum)
.append(",numSeries=").append(numSeries).append(",lastInSeries=").append(lastInSeries);
if (chunk != null) {
sb.append(",size=").append(chunk.length);
} else if (chunkStream != null) {
sb.append(",size=").append(chunkStream.size());
}
if (getException() != null) {
sb.append(",exception=").append(getException());
}
return sb.toString();
}
}
/**
* A processor to capture the value returned by
* {@link org.apache.geode.internal.cache.partitioned.GetMessage.GetReplyMessage}
*
* @since GemFire 5.0
*/
public static class FetchEntriesResponse extends ReplyProcessor21 {
private final PartitionedRegion pr;
protected volatile RegionVersionVector returnRVV;
protected final HashMap<Object, Object> returnValue;
protected final HashMap<Object, VersionTag> returnVersions = new HashMap();
private final Map<VersionSource, VersionSource> canonicalMembers =
new ConcurrentHashMap<VersionSource, VersionSource>();
/** lock used to synchronize chunk processing */
private final Object endLock = new Object();
/** number of chunks processed */
private volatile int chunksProcessed;
/** chunks expected (set when last chunk has been processed */
private volatile int chunksExpected;
/** whether the last chunk has been processed */
private volatile boolean lastChunkReceived;
private int bucketId;
private InternalDistributedMember recipient;
public FetchEntriesResponse(InternalDistributedSystem ds, final PartitionedRegion pr,
final InternalDistributedMember recipient, final int bucketId) {
super(ds, Collections.singleton(recipient));
this.pr = pr;
this.bucketId = bucketId;
this.recipient = recipient;
this.returnValue = new HashMap<Object, Object>() {
private static final long serialVersionUID = 0L;
@Override
public String toString() {
return "Bucket id = " + bucketId + " from member = " + recipient + ": "
+ super.toString();
}
};
}
@Override
public void process(DistributionMessage msg) {
// If the reply is a region version vector, store it in our RVV field.
if (msg instanceof ReplyMessage) {
ReplyMessage reply = (ReplyMessage) msg;
Object returnValue = reply.getReturnValue();
if (returnValue instanceof RegionVersionVector) {
this.returnRVV = (RegionVersionVector) returnValue;
synchronized (this.endLock) {
if (allMessagesReceived(true)) {
super.process(msg);
}
}
return;
}
}
super.process(msg);
}
void processChunk(FetchEntriesReplyMessage msg) {
// this processing algorighm won't work well if there are multiple recipients. currently the
// retry logic for failed recipients is in PartitionedRegion. If we parallelize the sending
// of this message, we'll need to handle failover in this processor class and track results
// differently.
final boolean isDebugEnabled = logger.isTraceEnabled(LogMarker.DM_VERBOSE);
boolean doneProcessing = false;
if (msg.getException() != null) {
process(msg);
} else {
boolean deserializingKey = true;
try {
ByteArrayInputStream byteStream = new ByteArrayInputStream(msg.chunk);
DataInputStream in = new DataInputStream(byteStream);
Object key;
while (in.available() > 0) {
deserializingKey = true;
key = DataSerializer.readObject(in);
if (key != null) {
deserializingKey = false;
Object value = DataSerializer.readObject(in);
VersionTag versionTag = DataSerializer.readObject(in);
// Fix for 47260 - canonicalize the mebmer ids to avoid an OOME
VersionSource id = versionTag == null ? null : versionTag.getMemberID();
if (id != null) {
if (canonicalMembers.containsKey(id)) {
versionTag.setMemberID(canonicalMembers.get(id));
} else {
canonicalMembers.put(id, id);
}
}
synchronized (returnValue) {
returnValue.put(key, value);
returnVersions.put(key, versionTag);
}
} else {
// null should signal the end of the set of keys
Assert.assertTrue(in.available() == 0);
}
}
synchronized (this.endLock) {
chunksProcessed = chunksProcessed + 1;
if (((msg.seriesNum + 1) == msg.numSeries) && msg.lastInSeries) {
chunksExpected = msg.msgNum + 1;
lastChunkReceived = true;
}
if (allMessagesReceived(msg.hasRVV)) {
doneProcessing = true;
}
if (isDebugEnabled) {
logger.trace(LogMarker.DM_VERBOSE,
"{} chunksProcessed={},lastChunkReceived={},chunksExpected={},done={}", this,
chunksProcessed, lastChunkReceived, chunksExpected, doneProcessing);
}
}
} catch (Exception e) {
if (deserializingKey) {
processException(new ReplyException(
"Error deserializing keys",
e));
} else {
processException(new ReplyException(
"Error deserializing values",
e)); // for bug 41202
}
checkIfDone(); // fix for hang in 41202
}
// if all chunks have been received, wake up the waiting thread
if (doneProcessing) {
process(msg);
}
}
}
private boolean allMessagesReceived(boolean hasRVV) {
synchronized (this.endLock) {
return lastChunkReceived && (chunksExpected == chunksProcessed)
&& (!hasRVV || returnRVV != null);
}
}
/**
* @return Set the keys associated with the bucketid of the {@link FetchKeysMessage}
* @throws ForceReattemptException if the peer is no longer available
*/
public BucketDump waitForEntries() throws ForceReattemptException {
try {
waitForRepliesUninterruptibly();
} catch (ReplyException e) {
Throwable t = e.getCause();
if (t instanceof CancelException) {
logger.debug("FetchKeysResponse got remote cancellation; forcing reattempt. {}",
t.getMessage(), t);
throw new ForceReattemptException(
"FetchKeysResponse got remote cancellation; forcing reattempt.",
t);
} else if (t instanceof ForceReattemptException) {
// Not sure this is necessary, but it is possible for
// FetchEntriesMessage to marshal a ForceReattemptException, so...
throw new ForceReattemptException(
"Peer requests reattempt", t);
}
e.handleCause();
}
if (!this.lastChunkReceived) {
throw new ForceReattemptException(
"No replies received");
}
// Deserialize all CachedDeserializable here so we have access to applications thread context
// class loader
Iterator it = this.returnValue.entrySet().iterator();
while (it.hasNext()) {
Map.Entry entry = (Map.Entry) it.next();
Object value = entry.getValue();
if (value instanceof CachedDeserializable) {
entry.setValue(((CachedDeserializable) value).getDeserializedValue(null, null));
}
}
return new BucketDump(bucketId, recipient, returnRVV, returnValue, returnVersions);
}
}
}