src/java/org/apache/cassandra/cache/AutoSavingCache.java

/*
 * 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.cassandra.cache;

import java.io.FileNotFoundException;
import java.io.IOException;
import java.nio.file.NoSuchFileException;
import java.util.ArrayDeque;
import java.util.Iterator;
import java.util.LinkedHashMap;
import java.util.Set;
import java.util.UUID;
import java.util.concurrent.ScheduledFuture;
import java.util.concurrent.TimeUnit;
import javax.annotation.concurrent.NotThreadSafe;

import org.cliffc.high_scale_lib.NonBlockingHashSet;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import org.apache.cassandra.concurrent.ExecutorPlus;
import org.apache.cassandra.concurrent.ScheduledExecutors;
import org.apache.cassandra.config.DatabaseDescriptor;
import org.apache.cassandra.db.ColumnFamilyStore;
import org.apache.cassandra.db.compaction.CompactionInfo;
import org.apache.cassandra.db.compaction.CompactionInfo.Unit;
import org.apache.cassandra.db.compaction.CompactionManager;
import org.apache.cassandra.db.compaction.OperationType;
import org.apache.cassandra.io.FSWriteError;
import org.apache.cassandra.io.util.ChecksummedRandomAccessReader;
import org.apache.cassandra.io.util.ChecksummedSequentialWriter;
import org.apache.cassandra.io.util.CorruptFileException;
import org.apache.cassandra.io.util.DataInputPlus;
import org.apache.cassandra.io.util.DataInputPlus.DataInputStreamPlus;
import org.apache.cassandra.io.util.DataOutputPlus;
import org.apache.cassandra.io.util.DataOutputStreamPlus;
import org.apache.cassandra.io.util.File;
import org.apache.cassandra.io.util.FileInputStreamPlus;
import org.apache.cassandra.io.util.FileOutputStreamPlus;
import org.apache.cassandra.io.util.FileUtils;
import org.apache.cassandra.io.util.SequentialWriterOption;
import org.apache.cassandra.io.util.WrappedDataOutputStreamPlus;
import org.apache.cassandra.schema.Schema;
import org.apache.cassandra.schema.SchemaConstants;
import org.apache.cassandra.schema.TableId;
import org.apache.cassandra.schema.TableMetadata;
import org.apache.cassandra.service.CacheService;
import org.apache.cassandra.utils.JVMStabilityInspector;
import org.apache.cassandra.utils.Pair;
import org.apache.cassandra.utils.concurrent.Future;

import static org.apache.cassandra.concurrent.ExecutorFactory.Global.executorFactory;
import static org.apache.cassandra.utils.Clock.Global.nanoTime;
import static org.apache.cassandra.utils.TimeUUID.Generator.nextTimeUUID;

public class AutoSavingCache<K extends CacheKey, V> extends InstrumentingCache<K, V>
{
    public interface IStreamFactory
    {
        DataInputStreamPlus getInputStream(File dataPath, File crcPath) throws IOException;

        DataOutputStreamPlus getOutputStream(File dataPath, File crcPath);
    }

    private static final Logger logger = LoggerFactory.getLogger(AutoSavingCache.class);

    /** True if a cache flush is currently executing: only one may execute at a time. */
    public static final Set<CacheService.CacheType> flushInProgress = new NonBlockingHashSet<CacheService.CacheType>();

    protected volatile ScheduledFuture<?> saveTask;
    protected final CacheService.CacheType cacheType;

    private final CacheSerializer<K, V> cacheLoader;

    /*
     * CASSANDRA-10155 required a format change to fix 2i indexes and caching.
     * 2.2 is already at version "c" and 3.0 is at "d".
     *
     * Since cache versions match exactly and there is no partial fallback just add
     * a minor version letter.
     *
     * Sticking with "d" is fine for 3.0 since it has never been released or used by another version
     *
     * "e" introduced with CASSANDRA-11206, omits IndexInfo from key-cache, stores offset into index-file
     *
     * "f" introduced with CASSANDRA-9425, changes "keyspace.table.index" in cache keys to TableMetadata.id+TableMetadata.indexName
     *
     * "g" introduced an explicit sstable format type ordinal number so that the entry can be skipped regardless of the actual implementation and used serializer
     */
    private static final String CURRENT_VERSION = "g";

    private static volatile IStreamFactory streamFactory = new IStreamFactory()
    {
        private final SequentialWriterOption writerOption = SequentialWriterOption.newBuilder()
                                                                    .trickleFsync(DatabaseDescriptor.getTrickleFsync())
                                                                    .trickleFsyncByteInterval(DatabaseDescriptor.getTrickleFsyncIntervalInKiB() * 1024)
                                                                    .finishOnClose(true).build();

        public DataInputStreamPlus getInputStream(File dataPath, File crcPath) throws IOException
        {
            return ChecksummedRandomAccessReader.open(dataPath, crcPath);
        }

        public DataOutputStreamPlus getOutputStream(File dataPath, File crcPath)
        {
            return new ChecksummedSequentialWriter(dataPath, crcPath, null, writerOption);
        }
    };

    // Unused, but exposed for a reason. See CASSANDRA-8096.
    public static void setStreamFactory(IStreamFactory streamFactory)
    {
        AutoSavingCache.streamFactory = streamFactory;
    }

    public AutoSavingCache(ICache<K, V> cache, CacheService.CacheType cacheType, CacheSerializer<K, V> cacheloader)
    {
        super(cacheType.toString(), cache);
        this.cacheType = cacheType;
        this.cacheLoader = cacheloader;
    }

    public File getCacheDataPath(String version)
    {
        return DatabaseDescriptor.getSerializedCachePath(cacheType, version, "db");
    }

    public File getCacheCrcPath(String version)
    {
        return DatabaseDescriptor.getSerializedCachePath(cacheType, version, "crc");
    }

    public File getCacheMetadataPath(String version)
    {
        return DatabaseDescriptor.getSerializedCachePath(cacheType, version, "metadata");
    }

    public Writer getWriter(int keysToSave)
    {
        return new Writer(keysToSave);
    }

    public void scheduleSaving(int savePeriodInSeconds, final int keysToSave)
    {
        if (saveTask != null)
        {
            saveTask.cancel(false); // Do not interrupt an in-progress save
            saveTask = null;
        }
        if (savePeriodInSeconds > 0)
        {
            Runnable runnable = new Runnable()
            {
                public void run()
                {
                    submitWrite(keysToSave);
                }
            };
            saveTask = ScheduledExecutors.optionalTasks.scheduleWithFixedDelay(runnable,
                                                                               savePeriodInSeconds,
                                                                               savePeriodInSeconds,
                                                                               TimeUnit.SECONDS);
        }
    }

    public Future<Integer> loadSavedAsync()
    {
        final ExecutorPlus es = executorFactory().sequential("loadSavedCache");
        final long start = nanoTime();

        Future<Integer> cacheLoad = es.submit(this::loadSaved);
        cacheLoad.addListener(() -> {
            if (size() > 0)
                logger.info("Completed loading ({} ms; {} keys) {} cache",
                        TimeUnit.NANOSECONDS.toMillis(nanoTime() - start),
                        CacheService.instance.keyCache.size(),
                        cacheType);
            es.shutdown();
        });

        return cacheLoad;
    }

    @SuppressWarnings("resource")
    public int loadSaved()
    {
        int count = 0;
        long start = nanoTime();

        // modern format, allows both key and value (so key cache load can be purely sequential)
        File dataPath = getCacheDataPath(CURRENT_VERSION);
        File crcPath = getCacheCrcPath(CURRENT_VERSION);
        File metadataPath = getCacheMetadataPath(CURRENT_VERSION);
        if (dataPath.exists() && crcPath.exists() && metadataPath.exists())
        {
            DataInputStreamPlus in = null;
            try
            {
                logger.info("Reading saved cache: {}, {}, {}", dataPath, crcPath, metadataPath);
                try (FileInputStreamPlus metadataIn = metadataPath.newInputStream())
                {
                    cacheLoader.deserializeMetadata(metadataIn);
                }

                in = streamFactory.getInputStream(dataPath, crcPath);

                //Check the schema has not changed since CFs are looked up by name which is ambiguous
                UUID schemaVersion = new UUID(in.readLong(), in.readLong());
                if (!schemaVersion.equals(Schema.instance.getVersion()))
                    throw new RuntimeException("Cache schema version "
                                               + schemaVersion
                                               + " does not match current schema version "
                                               + Schema.instance.getVersion());

                ArrayDeque<Future<Pair<K, V>>> futures = new ArrayDeque<>();
                long loadByNanos = start + TimeUnit.SECONDS.toNanos(DatabaseDescriptor.getCacheLoadTimeout());
                while (nanoTime() < loadByNanos && in.available() > 0)
                {
                    Future<Pair<K, V>> entryFuture = cacheLoader.deserialize(in);
                    // Key cache entry can return null, if the SSTable doesn't exist.
                    if (entryFuture == null)
                        continue;

                    futures.offer(entryFuture);
                    count++;

                    /*
                     * Kind of unwise to accrue an unbounded number of pending futures
                     * So now there is this loop to keep a bounded number pending.
                     */
                    do
                    {
                        while (futures.peek() != null && futures.peek().isDone())
                        {
                            Future<Pair<K, V>> future = futures.poll();
                            Pair<K, V> entry = future.get();
                            if (entry != null && entry.right != null)
                                put(entry.left, entry.right);
                        }

                        if (futures.size() > 1000)
                            Thread.yield();
                    } while(futures.size() > 1000);
                }

                Future<Pair<K, V>> future = null;
                while ((future = futures.poll()) != null)
                {
                    Pair<K, V> entry = future.get();
                    if (entry != null && entry.right != null)
                        put(entry.left, entry.right);
                }
            }
            catch (CorruptFileException e)
            {
                JVMStabilityInspector.inspectThrowable(e);
                logger.warn(String.format("Non-fatal checksum error reading saved cache %s", dataPath.absolutePath()), e);
            }
            catch (Throwable t)
            {
                JVMStabilityInspector.inspectThrowable(t);
                logger.info(String.format("Harmless error reading saved cache %s", dataPath.absolutePath()), t);
            }
            finally
            {
                FileUtils.closeQuietly(in);
                cacheLoader.cleanupAfterDeserialize();
            }
        }
        if (logger.isTraceEnabled())
            logger.trace("completed reading ({} ms; {} keys) saved cache {}",
                         TimeUnit.NANOSECONDS.toMillis(nanoTime() - start), count, dataPath);
        return count;
    }

    public Future<?> submitWrite(int keysToSave)
    {
        return CompactionManager.instance.submitCacheWrite(getWriter(keysToSave));
    }

    public class Writer extends CompactionInfo.Holder
    {
        private final Iterator<K> keyIterator;
        private final CompactionInfo info;
        private long keysWritten;
        private final long keysEstimate;

        protected Writer(int keysToSave)
        {
            int size = size();
            if (keysToSave >= size || keysToSave == 0)
            {
                keyIterator = keyIterator();
                keysEstimate = size;
            }
            else
            {
                keyIterator = hotKeyIterator(keysToSave);
                keysEstimate = keysToSave;
            }

            OperationType type;
            if (cacheType == CacheService.CacheType.KEY_CACHE)
                type = OperationType.KEY_CACHE_SAVE;
            else if (cacheType == CacheService.CacheType.ROW_CACHE)
                type = OperationType.ROW_CACHE_SAVE;
            else if (cacheType == CacheService.CacheType.COUNTER_CACHE)
                type = OperationType.COUNTER_CACHE_SAVE;
            else
                type = OperationType.UNKNOWN;

            info = CompactionInfo.withoutSSTables(TableMetadata.minimal(SchemaConstants.SYSTEM_KEYSPACE_NAME, cacheType.toString()),
                                                  type,
                                                  0,
                                                  keysEstimate,
                                                  Unit.KEYS,
                                                  nextTimeUUID(),
                                                  getCacheDataPath(CURRENT_VERSION).toPath().toString());
        }

        public CacheService.CacheType cacheType()
        {
            return cacheType;
        }

        public CompactionInfo getCompactionInfo()
        {
            // keyset can change in size, thus total can too
            // TODO need to check for this one... was: info.forProgress(keysWritten, Math.max(keysWritten, keys.size()));
            return info.forProgress(keysWritten, Math.max(keysWritten, keysEstimate));
        }

        public void saveCache()
        {
            logger.trace("Deleting old {} files.", cacheType);
            deleteOldCacheFiles();

            if (!keyIterator.hasNext())
            {
                logger.trace("Skipping {} save, cache is empty.", cacheType);
                return;
            }

            long start = nanoTime();

            File dataTmpFile = getTempCacheFile(getCacheDataPath(CURRENT_VERSION));
            File crcTmpFile = getTempCacheFile(getCacheCrcPath(CURRENT_VERSION));
            File metadataTmpFile = getTempCacheFile(getCacheMetadataPath(CURRENT_VERSION));

            try (WrappedDataOutputStreamPlus writer = new WrappedDataOutputStreamPlus(streamFactory.getOutputStream(dataTmpFile, crcTmpFile));
                 FileOutputStreamPlus metadataWriter = metadataTmpFile.newOutputStream(File.WriteMode.OVERWRITE))
            {

                //Need to be able to check schema version because CF names are ambiguous
                UUID schemaVersion = Schema.instance.getVersion();
                writer.writeLong(schemaVersion.getMostSignificantBits());
                writer.writeLong(schemaVersion.getLeastSignificantBits());

                while (keyIterator.hasNext())
                {
                    K key = keyIterator.next();

                    ColumnFamilyStore cfs = Schema.instance.getColumnFamilyStoreInstance(key.tableId);
                    if (cfs == null)
                        continue; // the table or 2i has been dropped.
                    if (key.indexName != null)
                        cfs = cfs.indexManager.getIndexByName(key.indexName).getBackingTable().orElse(null);

                    cacheLoader.serialize(key, writer, cfs);

                    keysWritten++;
                    if (keysWritten >= keysEstimate)
                        break;
                }

                cacheLoader.serializeMetadata(metadataWriter);
                metadataWriter.sync();
            }
            catch (FileNotFoundException | NoSuchFileException e)
            {
                throw new RuntimeException(e);
            }
            catch (IOException e)
            {
                throw new FSWriteError(e, dataTmpFile);
            }
            finally
            {
                cacheLoader.cleanupAfterSerialize();
            }

            File dataFile = getCacheDataPath(CURRENT_VERSION);
            File crcFile = getCacheCrcPath(CURRENT_VERSION);
            File metadataFile = getCacheMetadataPath(CURRENT_VERSION);

            dataFile.tryDelete(); // ignore error if it didn't exist
            crcFile.tryDelete();
            metadataFile.tryDelete();

            if (!dataTmpFile.tryMove(dataFile))
                logger.error("Unable to rename {} to {}", dataTmpFile, dataFile);

            if (!crcTmpFile.tryMove(crcFile))
                logger.error("Unable to rename {} to {}", crcTmpFile, crcFile);

            if (!metadataTmpFile.tryMove(metadataFile))
                logger.error("Unable to rename {} to {}", metadataTmpFile, metadataFile);

            logger.info("Saved {} ({} items) in {} ms to {} : {} MB", cacheType, keysWritten, TimeUnit.NANOSECONDS.toMillis(nanoTime() - start), dataFile.toPath(), dataFile.length() / (1 << 20));
        }

        private File getTempCacheFile(File cacheFile)
        {
            return FileUtils.createTempFile(cacheFile.name(), null, cacheFile.parent());
        }

        private void deleteOldCacheFiles()
        {
            File savedCachesDir = new File(DatabaseDescriptor.getSavedCachesLocation());
            assert savedCachesDir.exists() && savedCachesDir.isDirectory();
            File[] files = savedCachesDir.tryList();
            if (files != null)
            {
                String cacheNameFormat = String.format("%s-%s.db", cacheType.toString(), CURRENT_VERSION);
                for (File file : files)
                {
                    if (!file.isFile())
                        continue; // someone's been messing with our directory.  naughty!

                    if (file.name().endsWith(cacheNameFormat)
                     || file.name().endsWith(cacheType.toString()))
                    {
                        if (!file.tryDelete())
                            logger.warn("Failed to delete {}", file.absolutePath());
                    }
                }
            }
            else
            {
                logger.warn("Could not list files in {}", savedCachesDir);
            }
        }

        public boolean isGlobal()
        {
            return false;
        }
    }

    /**
     * A base cache serializer that is used to serialize/deserialize a cache to/from disk.
     * <p>
     * It expects the following lifecycle:
     * Serializations:
     * 1. {@link #serialize(CacheKey, DataOutputPlus, ColumnFamilyStore)} is called for each key in the cache.
     * 2. {@link #serializeMetadata(DataOutputPlus)} is called to serialize any metadata.
     * 3. {@link #cleanupAfterSerialize()} is called to clean up any resources allocated for serialization.
     * <p>
     * Deserializations:
     * 1. {@link #deserializeMetadata(DataInputPlus)} is called to deserialize any metadata.
     * 2. {@link #deserialize(DataInputPlus)} is called for each key in the cache.
     * 3. {@link #cleanupAfterDeserialize()} is called to clean up any resources allocated for deserialization.
     * <p>
     * This abstract class provides the default implementation for the metadata serialization/deserialization.
     * The metadata includes a dictionary of column family stores collected during serialization whenever
     * {@link #writeCFS(DataOutputPlus, ColumnFamilyStore)} or {@link #getOrCreateCFSOrdinal(ColumnFamilyStore)}
     * are called. When such metadata is deserialized, the implementation of {@link #deserialize(DataInputPlus)} may
     * use {@link #readCFS(DataInputPlus)} method to read the ColumnFamilyStore stored with
     * {@link #writeCFS(DataOutputPlus, ColumnFamilyStore)}.
     */
    @NotThreadSafe
    public static abstract class CacheSerializer<K extends CacheKey, V>
    {
        private ColumnFamilyStore[] cfStores;

        private final LinkedHashMap<Pair<TableId, String>, Integer> cfsOrdinals = new LinkedHashMap<>();

        protected final int getOrCreateCFSOrdinal(ColumnFamilyStore cfs)
        {
            Integer ordinal = cfsOrdinals.putIfAbsent(Pair.create(cfs.metadata().id, cfs.metadata().indexName().orElse("")), cfsOrdinals.size());
            if (ordinal == null)
                ordinal = cfsOrdinals.size() - 1;
            return ordinal;
        }

        protected ColumnFamilyStore readCFS(DataInputPlus in) throws IOException
        {
            return cfStores[in.readUnsignedVInt32()];
        }

        protected void writeCFS(DataOutputPlus out, ColumnFamilyStore cfs) throws IOException
        {
            out.writeUnsignedVInt32(getOrCreateCFSOrdinal(cfs));
        }

        public void serializeMetadata(DataOutputPlus out) throws IOException
        {
            // write the table ids
            out.writeUnsignedVInt32(cfsOrdinals.size());
            for (Pair<TableId, String> tableAndIndex : cfsOrdinals.keySet())
            {
                tableAndIndex.left.serialize(out);
                out.writeUTF(tableAndIndex.right);
            }
        }

        public void deserializeMetadata(DataInputPlus in) throws IOException
        {
            int tableEntries = in.readUnsignedVInt32();
            if (tableEntries == 0)
                return;
            cfStores = new ColumnFamilyStore[tableEntries];
            for (int i = 0; i < tableEntries; i++)
            {
                TableId tableId = TableId.deserialize(in);
                String indexName = in.readUTF();
                cfStores[i] = Schema.instance.getColumnFamilyStoreInstance(tableId);
                if (cfStores[i] != null && !indexName.isEmpty())
                    cfStores[i] = cfStores[i].indexManager.getIndexByName(indexName).getBackingTable().orElse(null);
            }
        }

        public abstract void serialize(K key, DataOutputPlus out, ColumnFamilyStore cfs) throws IOException;

        public abstract Future<Pair<K, V>> deserialize(DataInputPlus in) throws IOException;

        public void cleanupAfterSerialize()
        {
            cfsOrdinals.clear();
        }

        public void cleanupAfterDeserialize()
        {
            cfStores = null;
        }
    }
}