/**
* 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.db;
import java.io.File;
import java.io.FilenameFilter;
import java.io.IOError;
import java.io.IOException;
import java.lang.management.ManagementFactory;
import java.util.*;
import java.util.concurrent.*;
import java.util.concurrent.atomic.AtomicInteger;
import java.util.concurrent.atomic.AtomicReference;
import com.google.common.collect.Iterables;
import org.apache.commons.collections.IteratorUtils;
import org.apache.commons.lang.ArrayUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.apache.cassandra.concurrent.JMXEnabledThreadPoolExecutor;
import org.apache.cassandra.concurrent.NamedThreadFactory;
import org.apache.cassandra.concurrent.StageManager;
import org.apache.cassandra.config.CFMetaData;
import org.apache.cassandra.config.ColumnDefinition;
import org.apache.cassandra.config.DatabaseDescriptor;
import org.apache.cassandra.db.IClock.ClockRelationship;
import org.apache.cassandra.db.clock.TimestampReconciler;
import org.apache.cassandra.db.columniterator.IColumnIterator;
import org.apache.cassandra.db.columniterator.IdentityQueryFilter;
import org.apache.cassandra.db.commitlog.CommitLog;
import org.apache.cassandra.db.commitlog.CommitLogSegment;
import org.apache.cassandra.db.filter.*;
import org.apache.cassandra.db.marshal.AbstractType;
import org.apache.cassandra.db.marshal.BytesType;
import org.apache.cassandra.db.marshal.LocalByPartionerType;
import org.apache.cassandra.dht.*;
import org.apache.cassandra.io.sstable.Component;
import org.apache.cassandra.io.sstable.Descriptor;
import org.apache.cassandra.io.sstable.SSTable;
import org.apache.cassandra.io.sstable.SSTableReader;
import org.apache.cassandra.io.sstable.SSTableTracker;
import org.apache.cassandra.io.util.FileUtils;
import org.apache.cassandra.service.StorageService;
import org.apache.cassandra.thrift.IndexClause;
import org.apache.cassandra.thrift.IndexExpression;
import org.apache.cassandra.thrift.IndexOperator;
import org.apache.cassandra.utils.FBUtilities;
import org.apache.cassandra.utils.LatencyTracker;
import org.apache.cassandra.utils.Pair;
import org.apache.cassandra.utils.WrappedRunnable;
import javax.management.MBeanServer;
import javax.management.ObjectName;
public class ColumnFamilyStore implements ColumnFamilyStoreMBean
{
private static Logger logger = LoggerFactory.getLogger(ColumnFamilyStore.class);
/*
* submitFlush first puts [Binary]Memtable.getSortedContents on the flushSorter executor,
* which then puts the sorted results on the writer executor. This is because sorting is CPU-bound,
* and writing is disk-bound; we want to be able to do both at once. When the write is complete,
* we turn the writer into an SSTableReader and add it to ssTables_ where it is available for reads.
*
* For BinaryMemtable that's about all that happens. For live Memtables there are two other things
* that switchMemtable does (which should be the only caller of submitFlush in this case).
* First, it puts the Memtable into memtablesPendingFlush, where it stays until the flush is complete
* and it's been added as an SSTableReader to ssTables_. Second, it adds an entry to commitLogUpdater
* that waits for the flush to complete, then calls onMemtableFlush. This allows multiple flushes
* to happen simultaneously on multicore systems, while still calling onMF in the correct order,
* which is necessary for replay in case of a restart since CommitLog assumes that when onMF is
* called, all data up to the given context has been persisted to SSTables.
*/
private static final ExecutorService flushSorter
= new JMXEnabledThreadPoolExecutor(1,
Runtime.getRuntime().availableProcessors(),
StageManager.KEEPALIVE,
TimeUnit.SECONDS,
new LinkedBlockingQueue<Runnable>(Runtime.getRuntime().availableProcessors()),
new NamedThreadFactory("FLUSH-SORTER-POOL"));
private static final ExecutorService flushWriter
= new JMXEnabledThreadPoolExecutor(1,
DatabaseDescriptor.getFlushWriters(),
StageManager.KEEPALIVE,
TimeUnit.SECONDS,
new LinkedBlockingQueue<Runnable>(DatabaseDescriptor.getFlushWriters()),
new NamedThreadFactory("FLUSH-WRITER-POOL"));
public static final ExecutorService postFlushExecutor = new JMXEnabledThreadPoolExecutor("MEMTABLE-POST-FLUSHER");
private Set<Memtable> memtablesPendingFlush = new ConcurrentSkipListSet<Memtable>();
public final String table;
public final String columnFamily;
public final IPartitioner partitioner;
private final String mbeanName;
private volatile int memtableSwitchCount = 0;
/* This is used to generate the next index for a SSTable */
private AtomicInteger fileIndexGenerator = new AtomicInteger(0);
/* active memtable associated with this ColumnFamilyStore. */
private Memtable memtable;
private final SortedMap<byte[], ColumnFamilyStore> indexedColumns;
// TODO binarymemtable ops are not threadsafe (do they need to be?)
private AtomicReference<BinaryMemtable> binaryMemtable;
/* SSTables on disk for this column family */
private SSTableTracker ssTables;
private LatencyTracker readStats = new LatencyTracker();
private LatencyTracker writeStats = new LatencyTracker();
public final CFMetaData metadata;
private ColumnFamilyStore(String table, String columnFamilyName, IPartitioner partitioner, int generation, CFMetaData metadata)
{
assert metadata != null : "null metadata for " + table + ":" + columnFamilyName;
this.table = table;
columnFamily = columnFamilyName;
this.metadata = metadata;
this.partitioner = partitioner;
fileIndexGenerator.set(generation);
memtable = new Memtable(this, this.partitioner);
binaryMemtable = new AtomicReference<BinaryMemtable>(new BinaryMemtable(this));
if (logger.isDebugEnabled())
logger.debug("Starting CFS {}", columnFamily);
// scan for sstables corresponding to this cf and load them
List<SSTableReader> sstables = new ArrayList<SSTableReader>();
for (Map.Entry<Descriptor,Set<Component>> sstableFiles : files(table, columnFamilyName).entrySet())
{
SSTableReader sstable;
try
{
sstable = SSTableReader.open(sstableFiles.getKey(), sstableFiles.getValue(), metadata, this.partitioner);
}
catch (IOException ex)
{
logger.error("Corrupt sstable " + sstableFiles + "; skipped", ex);
continue;
}
sstables.add(sstable);
}
ssTables = new SSTableTracker(table, columnFamilyName);
ssTables.add(sstables);
indexedColumns = new TreeMap<byte[], ColumnFamilyStore>(getComparator());
for (Map.Entry<byte[], ColumnDefinition> entry : metadata.column_metadata.entrySet())
{
byte[] column = entry.getKey();
ColumnDefinition info = entry.getValue();
if (info.index_type == null)
continue;
String indexedCfName = columnFamily + "." + (info.index_name == null ? FBUtilities.bytesToHex(column) : info.index_name);
IPartitioner rowPartitioner = StorageService.getPartitioner();
AbstractType columnComparator = (rowPartitioner instanceof OrderPreservingPartitioner || rowPartitioner instanceof ByteOrderedPartitioner)
? BytesType.instance
: new LocalByPartionerType(StorageService.getPartitioner());
CFMetaData indexedCfMetadata = new CFMetaData(table,
indexedCfName,
ColumnFamilyType.Standard,
ClockType.Timestamp,
columnComparator,
null,
TimestampReconciler.instance,
"",
0,
false,
0,
0,
CFMetaData.DEFAULT_GC_GRACE_SECONDS,
BytesType.instance,
Collections.<byte[], ColumnDefinition>emptyMap());
ColumnFamilyStore indexedCfs = ColumnFamilyStore.createColumnFamilyStore(table,
indexedCfName,
new LocalPartitioner(metadata.column_metadata.get(column).validator),
indexedCfMetadata);
indexedColumns.put(column, indexedCfs);
}
String type = this.partitioner instanceof LocalPartitioner ? "IndexColumnFamilies" : "ColumnFamilies";
mbeanName = "org.apache.cassandra.db:type=" + type + ",keyspace=" + this.table + ",columnfamily=" + columnFamily;
try
{
MBeanServer mbs = ManagementFactory.getPlatformMBeanServer();
ObjectName nameObj = new ObjectName(mbeanName);
mbs.registerMBean(this, nameObj);
}
catch (Exception e)
{
throw new RuntimeException(e);
}
}
// called when dropping or renaming a CF. Performs mbean housekeeping.
void unregisterMBean()
{
try
{
MBeanServer mbs = ManagementFactory.getPlatformMBeanServer();
ObjectName nameObj = new ObjectName(mbeanName);
if (mbs.isRegistered(nameObj))
mbs.unregisterMBean(nameObj);
for (ColumnFamilyStore index : indexedColumns.values())
index.unregisterMBean();
}
catch (Exception e)
{
// this shouldn't block anything.
logger.warn(e.getMessage(), e);
}
}
public long getMinRowSize()
{
long min = 0;
for (SSTableReader sstable : ssTables)
{
if (min == 0 || sstable.getEstimatedRowSize().min() < min)
min = sstable.getEstimatedRowSize().min();
}
return min;
}
public long getMaxRowSize()
{
long max = 0;
for (SSTableReader sstable : ssTables)
{
if (sstable.getEstimatedRowSize().max() > max)
max = sstable.getEstimatedRowSize().max();
}
return max;
}
public long getMeanRowSize()
{
long sum = 0;
long count = 0;
for (SSTableReader sstable : ssTables)
{
sum += sstable.getEstimatedRowSize().median();
count++;
}
return count > 0 ? sum / count : 0;
}
public int getMeanColumns()
{
long sum = 0;
int count = 0;
for (SSTableReader sstable : ssTables)
{
sum += sstable.getEstimatedColumnCount().median();
count++;
}
return count > 0 ? (int) (sum / count) : 0;
}
public static ColumnFamilyStore createColumnFamilyStore(String table, String columnFamily)
{
return createColumnFamilyStore(table, columnFamily, StorageService.getPartitioner(), DatabaseDescriptor.getCFMetaData(table, columnFamily));
}
public static synchronized ColumnFamilyStore createColumnFamilyStore(String table, String columnFamily, IPartitioner partitioner, CFMetaData metadata)
{
// get the max generation number, to prevent generation conflicts
List<Integer> generations = new ArrayList<Integer>();
for (Descriptor desc : files(table, columnFamily).keySet())
generations.add(desc.generation);
Collections.sort(generations);
int value = (generations.size() > 0) ? (generations.get(generations.size() - 1)) : 0;
return new ColumnFamilyStore(table, columnFamily, partitioner, value, metadata);
}
/**
* Removes unnecessary files from the cf directory at startup: these include temp files, orphans, zero-length files
* and compacted sstables. Files that cannot be recognized will be ignored.
* @return A list of Descriptors that were removed.
*/
static void scrubDataDirectories(String table, String columnFamily)
{
for (Map.Entry<Descriptor,Set<Component>> sstableFiles : files(table, columnFamily).entrySet())
{
Descriptor desc = sstableFiles.getKey();
Set<Component> components = sstableFiles.getValue();
if (SSTable.conditionalDelete(desc, components))
// was compacted or temporary: deleted.
continue;
File dataFile = new File(desc.filenameFor(Component.DATA));
if (components.contains(Component.DATA) && dataFile.length() > 0)
// everything appears to be in order... moving on.
continue;
// missing the DATA file! all components are orphaned
logger.warn("Removing orphans for {}: {}", desc, components);
for (Component component : components)
{
try
{
FileUtils.deleteWithConfirm(desc.filenameFor(component));
}
catch (IOException e)
{
throw new IOError(e);
}
}
}
}
/**
* Collects a map of sstable components.
*/
private static Map<Descriptor,Set<Component>> files(String keyspace, final String columnFamily)
{
final Map<Descriptor,Set<Component>> sstables = new HashMap<Descriptor,Set<Component>>();
for (String directory : DatabaseDescriptor.getAllDataFileLocationsForTable(keyspace))
{
// NB: we never "accept" a file in the FilenameFilter sense: they are added to the sstable map
new File(directory).list(new FilenameFilter()
{
@Override
public boolean accept(File dir, String name)
{
Pair<Descriptor,Component> component = SSTable.tryComponentFromFilename(dir, name);
if (component != null && component.left.cfname.equals(columnFamily))
{
Set<Component> components = sstables.get(component.left);
if (components == null)
{
components = new HashSet<Component>();
sstables.put(component.left, components);
}
components.add(component.right);
}
return false;
}
});
}
return sstables;
}
/**
* @return the name of the column family
*/
public String getColumnFamilyName()
{
return columnFamily;
}
/*
* @return a temporary file name for an sstable.
* When the sstable object is closed, it will be renamed to a non-temporary
* format, so incomplete sstables can be recognized and removed on startup.
*/
public String getFlushPath()
{
long guessedSize = 2 * DatabaseDescriptor.getMemtableThroughput() * 1024*1024; // 2* adds room for keys, column indexes
String location = DatabaseDescriptor.getDataFileLocationForTable(table, guessedSize);
if (location == null)
throw new RuntimeException("Insufficient disk space to flush");
return getTempSSTablePath(location);
}
public String getTempSSTablePath(String directory)
{
Descriptor desc = new Descriptor(new File(directory),
table,
columnFamily,
fileIndexGenerator.incrementAndGet(),
true);
return desc.filenameFor(Component.DATA);
}
/** flush the given memtable and swap in a new one for its CFS, if it hasn't been frozen already. threadsafe. */
Future<?> maybeSwitchMemtable(Memtable oldMemtable, final boolean writeCommitLog)
{
/**
* If we can get the writelock, that means no new updates can come in and
* all ongoing updates to memtables have completed. We can get the tail
* of the log and use it as the starting position for log replay on recovery.
*/
Table.flusherLock.writeLock().lock();
try
{
if (oldMemtable.isFrozen())
return null;
assert memtable == oldMemtable;
memtable.freeze();
final CommitLogSegment.CommitLogContext ctx = writeCommitLog ? CommitLog.instance().getContext() : null;
logger.info("switching in a fresh Memtable for " + columnFamily + " at " + ctx);
// submit the memtable for any indexed sub-cfses, and our own
final CountDownLatch latch = new CountDownLatch(1 + indexedColumns.size());
for (ColumnFamilyStore cfs : Iterables.concat(indexedColumns.values(), Collections.singleton(this)))
{
submitFlush(cfs.memtable, latch);
cfs.memtable = new Memtable(cfs, cfs.partitioner);
}
// when all the memtables have been written, including for indexes, mark the flush in the commitlog header.
// a second executor makes sure the onMemtableFlushes get called in the right order,
// while keeping the wait-for-flush (future.get) out of anything latency-sensitive.
return postFlushExecutor.submit(new WrappedRunnable()
{
public void runMayThrow() throws InterruptedException, IOException
{
latch.await();
if (writeCommitLog)
{
// if we're not writing to the commit log, we are replaying the log, so marking
// the log header with "you can discard anything written before the context" is not valid
logger.debug("Discarding {}", metadata.cfId);
CommitLog.instance().discardCompletedSegments(metadata.cfId, ctx);
}
}
});
}
finally
{
Table.flusherLock.writeLock().unlock();
if (memtableSwitchCount == Integer.MAX_VALUE)
{
memtableSwitchCount = 0;
}
memtableSwitchCount++;
}
}
void switchBinaryMemtable(DecoratedKey key, byte[] buffer)
{
binaryMemtable.set(new BinaryMemtable(this));
binaryMemtable.get().put(key, buffer);
}
public void forceFlushIfExpired()
{
if (memtable.isExpired())
forceFlush();
}
public Future<?> forceFlush()
{
if (memtable.isClean())
return null;
return maybeSwitchMemtable(memtable, true);
}
public void forceBlockingFlush() throws ExecutionException, InterruptedException
{
Future<?> future = forceFlush();
if (future != null)
future.get();
}
public void forceFlushBinary()
{
if (binaryMemtable.get().isClean())
return;
submitFlush(binaryMemtable.get(), new CountDownLatch(1));
}
/**
* Insert/Update the column family for this key.
* Caller is responsible for acquiring Table.flusherLock!
* param @ lock - lock that needs to be used.
* param @ key - key for update/insert
* param @ columnFamily - columnFamily changes
*/
Memtable apply(DecoratedKey key, ColumnFamily columnFamily)
{
long start = System.nanoTime();
boolean flushRequested = memtable.isThresholdViolated();
memtable.put(key, columnFamily);
writeStats.addNano(System.nanoTime() - start);
return flushRequested ? memtable : null;
}
/*
* Insert/Update the column family for this key. param @ lock - lock that
* needs to be used. param @ key - key for update/insert param @
* columnFamily - columnFamily changes
*/
void applyBinary(DecoratedKey key, byte[] buffer)
{
long start = System.nanoTime();
binaryMemtable.get().put(key, buffer);
writeStats.addNano(System.nanoTime() - start);
}
public static ColumnFamily removeDeletedCF(ColumnFamily cf, int gcBefore)
{
// in case of a timestamp tie, tombstones get priority over non-tombstones.
// (we want this to be deterministic to avoid confusion.)
if (cf.getColumnCount() == 0 && cf.getLocalDeletionTime() <= gcBefore)
return null;
return cf;
}
/*
This is complicated because we need to preserve deleted columns, supercolumns, and columnfamilies
until they have been deleted for at least GC_GRACE_IN_SECONDS. But, we do not need to preserve
their contents; just the object itself as a "tombstone" that can be used to repair other
replicas that do not know about the deletion.
*/
public static ColumnFamily removeDeleted(ColumnFamily cf, int gcBefore)
{
if (cf == null)
{
return null;
}
removeDeletedColumnsOnly(cf, gcBefore);
return removeDeletedCF(cf, gcBefore);
}
private static void removeDeletedColumnsOnly(ColumnFamily cf, int gcBefore)
{
if (cf.isSuper())
removeDeletedSuper(cf, gcBefore);
else
removeDeletedStandard(cf, gcBefore);
}
private static void removeDeletedStandard(ColumnFamily cf, int gcBefore)
{
for (Map.Entry<byte[], IColumn> entry : cf.getColumnsMap().entrySet())
{
byte[] cname = entry.getKey();
IColumn c = entry.getValue();
// remove columns if
// (a) the column itself is tombstoned or
// (b) the CF is tombstoned and the column is not newer than it
// (we split the test to avoid computing ClockRelationship if not necessary)
if ((c.isMarkedForDelete() && c.getLocalDeletionTime() <= gcBefore))
{
cf.remove(cname);
}
else
{
ClockRelationship rel = c.clock().compare(cf.getMarkedForDeleteAt());
if ((ClockRelationship.LESS_THAN == rel) || (ClockRelationship.EQUAL == rel))
{
cf.remove(cname);
}
}
}
}
private static void removeDeletedSuper(ColumnFamily cf, int gcBefore)
{
// TODO assume deletion means "most are deleted?" and add to clone, instead of remove from original?
// this could be improved by having compaction, or possibly even removeDeleted, r/m the tombstone
// once gcBefore has passed, so if new stuff is added in it doesn't used the wrong algorithm forever
for (Map.Entry<byte[], IColumn> entry : cf.getColumnsMap().entrySet())
{
SuperColumn c = (SuperColumn) entry.getValue();
List<IClock> clocks = Arrays.asList(cf.getMarkedForDeleteAt());
IClock minClock = c.getMarkedForDeleteAt().getSuperset(clocks);
for (IColumn subColumn : c.getSubColumns())
{
// remove subcolumns if
// (a) the subcolumn itself is tombstoned or
// (b) the supercolumn is tombstoned and the subcolumn is not newer than it
// (we split the test to avoid computing ClockRelationship if not necessary)
if (subColumn.isMarkedForDelete() && subColumn.getLocalDeletionTime() <= gcBefore)
{
c.remove(subColumn.name());
}
else
{
ClockRelationship subRel = subColumn.clock().compare(minClock);
if ((ClockRelationship.LESS_THAN == subRel) || (ClockRelationship.EQUAL == subRel))
{
c.remove(subColumn.name());
}
}
}
if (c.getSubColumns().isEmpty() && c.getLocalDeletionTime() <= gcBefore)
{
cf.remove(c.name());
}
}
}
/**
* Uses bloom filters to check if key may be present in any sstable in this
* ColumnFamilyStore, minus a set of provided ones.
*
* Because BFs are checked, negative returns ensure that the key is not
* present in the checked SSTables, but positive ones doesn't ensure key
* presence.
*/
public boolean isKeyInRemainingSSTables(DecoratedKey key, Set<SSTable> sstablesToIgnore)
{
for (SSTableReader sstable : ssTables)
{
if (!sstablesToIgnore.contains(sstable) && sstable.getBloomFilter().isPresent(key.key))
return true;
}
return false;
}
/*
* Called after the Memtable flushes its in-memory data, or we add a file
* via bootstrap. This information is
* cached in the ColumnFamilyStore. This is useful for reads because the
* ColumnFamilyStore first looks in the in-memory store and the into the
* disk to find the key. If invoked during recoveryMode the
* onMemtableFlush() need not be invoked.
*
* param @ filename - filename just flushed to disk
*/
public void addSSTable(SSTableReader sstable)
{
ssTables.add(Arrays.asList(sstable));
CompactionManager.instance.submitMinorIfNeeded(this);
}
/*
* Add up all the files sizes this is the worst case file
* size for compaction of all the list of files given.
*/
long getExpectedCompactedFileSize(Iterable<SSTableReader> sstables)
{
long expectedFileSize = 0;
for (SSTableReader sstable : sstables)
{
long size = sstable.length();
expectedFileSize = expectedFileSize + size;
}
return expectedFileSize;
}
/*
* Find the maximum size file in the list .
*/
SSTableReader getMaxSizeFile(Iterable<SSTableReader> sstables)
{
long maxSize = 0L;
SSTableReader maxFile = null;
for (SSTableReader sstable : sstables)
{
if (sstable.length() > maxSize)
{
maxSize = sstable.length();
maxFile = sstable;
}
}
return maxFile;
}
void forceCleanup()
{
CompactionManager.instance.submitCleanup(ColumnFamilyStore.this);
}
public Table getTable()
{
return Table.open(table);
}
void markCompacted(Collection<SSTableReader> sstables)
{
ssTables.markCompacted(sstables);
}
boolean isCompleteSSTables(Collection<SSTableReader> sstables)
{
return ssTables.getSSTables().equals(new HashSet<SSTableReader>(sstables));
}
void replaceCompactedSSTables(Collection<SSTableReader> sstables, Iterable<SSTableReader> replacements)
{
ssTables.replace(sstables, replacements);
}
public void removeAllSSTables()
{
ssTables.replace(ssTables.getSSTables(), Collections.<SSTableReader>emptyList());
for (ColumnFamilyStore indexedCfs : indexedColumns.values())
{
indexedCfs.removeAllSSTables();
}
}
/**
* submits flush sort on the flushSorter executor, which will in turn submit to flushWriter when sorted.
* TODO because our executors use CallerRunsPolicy, when flushSorter fills up, no writes will proceed
* because the next flush will start executing on the caller, mutation-stage thread that has the
* flush write lock held. (writes aquire this as a read lock before proceeding.)
* This is good, because it backpressures flushes, but bad, because we can't write until that last
* flushing thread finishes sorting, which will almost always be longer than any of the flushSorter threads proper
* (since, by definition, it started last).
*/
void submitFlush(IFlushable flushable, CountDownLatch latch)
{
logger.info("Enqueuing flush of {}", flushable);
flushable.flushAndSignal(latch, flushSorter, flushWriter);
}
public int getMemtableColumnsCount()
{
return getMemtableThreadSafe().getCurrentOperations();
}
public int getMemtableDataSize()
{
return getMemtableThreadSafe().getCurrentThroughput();
}
public int getMemtableSwitchCount()
{
return memtableSwitchCount;
}
/**
* get the current memtable in a threadsafe fashion. note that simply "return memtable_" is
* incorrect; you need to lock to introduce a thread safe happens-before ordering.
*
* do NOT use this method to do either a put or get on the memtable object, since it could be
* flushed in the meantime (and its executor terminated).
*
* also do NOT make this method public or it will really get impossible to reason about these things.
* @return
*/
private Memtable getMemtableThreadSafe()
{
Table.flusherLock.readLock().lock();
try
{
return memtable;
}
finally
{
Table.flusherLock.readLock().unlock();
}
}
public Collection<SSTableReader> getSSTables()
{
return ssTables.getSSTables();
}
public long getReadCount()
{
return readStats.getOpCount();
}
public double getRecentReadLatencyMicros()
{
return readStats.getRecentLatencyMicros();
}
public long[] getLifetimeReadLatencyHistogramMicros()
{
return readStats.getTotalLatencyHistogramMicros();
}
public long[] getRecentReadLatencyHistogramMicros()
{
return readStats.getRecentLatencyHistogramMicros();
}
public long getTotalReadLatencyMicros()
{
return readStats.getTotalLatencyMicros();
}
// TODO this actually isn't a good meature of pending tasks
public int getPendingTasks()
{
return Table.flusherLock.getQueueLength();
}
public long getWriteCount()
{
return writeStats.getOpCount();
}
public long getTotalWriteLatencyMicros()
{
return writeStats.getTotalLatencyMicros();
}
public double getRecentWriteLatencyMicros()
{
return writeStats.getRecentLatencyMicros();
}
public long[] getLifetimeWriteLatencyHistogramMicros()
{
return writeStats.getTotalLatencyHistogramMicros();
}
public long[] getRecentWriteLatencyHistogramMicros()
{
return writeStats.getRecentLatencyHistogramMicros();
}
public ColumnFamily getColumnFamily(DecoratedKey key, QueryPath path, byte[] start, byte[] finish, boolean reversed, int limit)
{
return getColumnFamily(QueryFilter.getSliceFilter(key, path, start, finish, reversed, limit));
}
/**
* get a list of columns starting from a given column, in a specified order.
* only the latest version of a column is returned.
* @return null if there is no data and no tombstones; otherwise a ColumnFamily
*/
public ColumnFamily getColumnFamily(QueryFilter filter)
{
return getColumnFamily(filter, gcBefore());
}
public int gcBefore()
{
return (int) (System.currentTimeMillis() / 1000) - metadata.gcGraceSeconds;
}
private ColumnFamily cacheRow(DecoratedKey key)
{
ColumnFamily cached;
if ((cached = ssTables.getRowCache().get(key)) == null)
{
cached = getTopLevelColumns(QueryFilter.getIdentityFilter(key, new QueryPath(columnFamily)), Integer.MIN_VALUE);
if (cached == null)
return null;
ssTables.getRowCache().put(key, cached);
}
return cached;
}
private ColumnFamily getColumnFamily(QueryFilter filter, int gcBefore)
{
assert columnFamily.equals(filter.getColumnFamilyName()) : filter.getColumnFamilyName();
long start = System.nanoTime();
try
{
if (ssTables.getRowCache().getCapacity() == 0)
{
ColumnFamily cf = getTopLevelColumns(filter, gcBefore);
// TODO this is necessary because when we collate supercolumns together, we don't check
// their subcolumns for relevance, so we need to do a second prune post facto here.
return cf.isSuper() ? removeDeleted(cf, gcBefore) : removeDeletedCF(cf, gcBefore);
}
ColumnFamily cached = cacheRow(filter.key);
if (cached == null)
return null;
return filterColumnFamily(cached, filter, gcBefore);
}
finally
{
readStats.addNano(System.nanoTime() - start);
}
}
/** filter a cached row, which will not be modified by the filter, but may be modified by throwing out
* tombstones that are no longer relevant. */
ColumnFamily filterColumnFamily(ColumnFamily cached, QueryFilter filter, int gcBefore)
{
// special case slicing the entire row:
// we can skip the filter step entirely, and we can help out removeDeleted by re-caching the result
// if any tombstones have aged out since last time. (This means that the row cache will treat gcBefore as
// max(gcBefore, all previous gcBefore), which is fine for correctness.)
//
// But, if the filter is asking for less columns than we have cached, we fall back to the slow path
// since we have to copy out a subset.
if (filter.filter instanceof SliceQueryFilter)
{
SliceQueryFilter sliceFilter = (SliceQueryFilter) filter.filter;
if (sliceFilter.start.length == 0 && sliceFilter.finish.length == 0)
{
if (cached.isSuper() && filter.path.superColumnName != null)
{
// subcolumns from named supercolumn
IColumn sc = cached.getColumn(filter.path.superColumnName);
if (sc == null || sliceFilter.count >= sc.getSubColumns().size())
{
ColumnFamily cf = cached.cloneMeShallow();
if (sc != null)
cf.addColumn(sc);
return removeDeleted(cf, gcBefore);
}
}
else
{
// top-level columns
if (sliceFilter.count >= cached.getColumnCount())
{
removeDeletedColumnsOnly(cached, gcBefore);
return removeDeletedCF(cached, gcBefore);
}
}
}
}
IColumnIterator ci = filter.getMemtableColumnIterator(cached, null, getComparator());
ColumnFamily cf = null;
try
{
cf = ci.getColumnFamily().cloneMeShallow();
}
catch (IOException e)
{
throw new IOError(e);
}
filter.collectCollatedColumns(cf, ci, gcBefore);
// TODO this is necessary because when we collate supercolumns together, we don't check
// their subcolumns for relevance, so we need to do a second prune post facto here.
return cf.isSuper() ? removeDeleted(cf, gcBefore) : removeDeletedCF(cf, gcBefore);
}
private ColumnFamily getTopLevelColumns(QueryFilter filter, int gcBefore)
{
// we are querying top-level columns, do a merging fetch with indexes.
List<IColumnIterator> iterators = new ArrayList<IColumnIterator>();
final ColumnFamily returnCF = ColumnFamily.create(metadata);
try
{
IColumnIterator iter;
/* add the current memtable */
iter = filter.getMemtableColumnIterator(getMemtableThreadSafe(), getComparator());
if (iter != null)
{
returnCF.delete(iter.getColumnFamily());
iterators.add(iter);
}
/* add the memtables being flushed */
for (Memtable memtable : memtablesPendingFlush)
{
iter = filter.getMemtableColumnIterator(memtable, getComparator());
if (iter != null)
{
returnCF.delete(iter.getColumnFamily());
iterators.add(iter);
}
}
/* add the SSTables on disk */
for (SSTableReader sstable : ssTables)
{
iter = filter.getSSTableColumnIterator(sstable);
if (iter.getColumnFamily() != null)
{
returnCF.delete(iter.getColumnFamily());
iterators.add(iter);
}
}
Comparator<IColumn> comparator = filter.filter.getColumnComparator(getComparator());
Iterator collated = IteratorUtils.collatedIterator(comparator, iterators);
filter.collectCollatedColumns(returnCF, collated, gcBefore);
// Caller is responsible for final removeDeletedCF. This is important for cacheRow to work correctly:
// we need to distinguish between "there is no data at all for this row" (BF will let us rebuild that efficiently)
// and "there used to be data, but it's gone now" (we should cache the empty CF so we don't need to rebuild that slower)
return returnCF;
}
catch (IOException e)
{
throw new IOError(e);
}
finally
{
/* close all cursors */
for (IColumnIterator ci : iterators)
{
try
{
ci.close();
}
catch (Throwable th)
{
logger.error("error closing " + ci, th);
}
}
}
}
/**
* Fetch a range of rows and columns from memtables/sstables.
*
* @param superColumn optional SuperColumn to slice subcolumns of; null to slice top-level columns
* @param range Either a Bounds, which includes start key, or a Range, which does not.
* @param maxResults Maximum rows to return
* @param columnFilter description of the columns we're interested in for each row
* @return true if we found all keys we were looking for, otherwise false
*/
public List<Row> getRangeSlice(byte[] superColumn, final AbstractBounds range, int maxResults, IFilter columnFilter)
throws ExecutionException, InterruptedException
{
assert range instanceof Bounds
|| (!((Range)range).isWrapAround() || range.right.equals(StorageService.getPartitioner().getMinimumToken()))
: range;
List<Row> rows = new ArrayList<Row>();
DecoratedKey startWith = new DecoratedKey(range.left, (byte[])null);
DecoratedKey stopAt = new DecoratedKey(range.right, (byte[])null);
QueryFilter filter = new QueryFilter(null, new QueryPath(columnFamily, superColumn, null), columnFilter);
Collection<Memtable> memtables = new ArrayList<Memtable>();
memtables.add(getMemtableThreadSafe());
memtables.addAll(memtablesPendingFlush);
Collection<SSTableReader> sstables = new ArrayList<SSTableReader>();
Iterables.addAll(sstables, ssTables);
RowIterator iterator = RowIteratorFactory.getIterator(memtables, sstables, startWith, stopAt, filter, getComparator(), this);
try
{
// pull rows out of the iterator
boolean first = true;
while(iterator.hasNext())
{
Row current = iterator.next();
DecoratedKey key = current.key;
if (!stopAt.isEmpty() && stopAt.compareTo(key) < 0)
return rows;
// skip first one
if(range instanceof Bounds || !first || !key.equals(startWith))
{
rows.add(current);
if (logger.isDebugEnabled())
logger.debug("scanned " + key);
}
first = false;
if (rows.size() >= maxResults)
return rows;
}
}
finally
{
try
{
iterator.close();
}
catch (IOException e)
{
throw new IOError(e);
}
}
return rows;
}
public List<Row> scan(IndexClause clause, AbstractBounds range, IFilter dataFilter)
{
// Start with the most-restrictive indexed clause, then apply remaining clauses
// to each row matching that clause.
// TODO: allow merge join instead of just one index + loop
IndexExpression primary = highestSelectivityPredicate(clause);
ColumnFamilyStore indexCFS = getIndexedColumnFamilyStore(primary.column_name);
assert indexCFS != null;
DecoratedKey indexKey = indexCFS.partitioner.decorateKey(primary.value);
// if the slicepredicate doesn't contain all the columns for which we have expressions to evaluate,
// it needs to be expanded to include those too
IFilter firstFilter = dataFilter;
NamesQueryFilter extraFilter = null;
if (clause.expressions.size() > 1)
{
if (dataFilter instanceof SliceQueryFilter)
{
// if we have a high chance of getting all the columns in a single index slice, do that.
// otherwise, create an extraFilter to fetch by name the columns referenced by the additional expressions.
if (getMaxRowSize() < DatabaseDescriptor.getColumnIndexSize())
{
firstFilter = new SliceQueryFilter(ArrayUtils.EMPTY_BYTE_ARRAY,
ArrayUtils.EMPTY_BYTE_ARRAY,
((SliceQueryFilter) dataFilter).reversed,
Integer.MAX_VALUE);
}
else
{
SortedSet<byte[]> columns = new TreeSet<byte[]>(getComparator());
for (IndexExpression expr : clause.expressions)
{
if (expr == primary)
continue;
columns.add(expr.column_name);
}
extraFilter = new NamesQueryFilter(columns);
}
}
else
{
// just add in columns that are not part of the resultset
assert dataFilter instanceof NamesQueryFilter;
SortedSet<byte[]> columns = new TreeSet<byte[]>(getComparator());
for (IndexExpression expr : clause.expressions)
{
if (expr == primary || ((NamesQueryFilter) dataFilter).columns.contains(expr.column_name))
continue;
columns.add(expr.column_name);
}
if (columns.size() > 0)
{
columns.addAll(((NamesQueryFilter) dataFilter).columns);
firstFilter = new NamesQueryFilter(columns);
}
}
}
List<Row> rows = new ArrayList<Row>();
byte[] startKey = clause.start_key;
QueryPath path = new QueryPath(columnFamily);
// fetch row keys matching the primary expression, fetch the slice predicate for each
// and filter by remaining expressions. repeat until finished w/ assigned range or index row is exhausted.
outer:
while (true)
{
/* we don't have a way to get the key back from the DK -- we just have a token --
* so, we need to loop after starting with start_key, until we get to keys in the given `range`.
* But, if the calling StorageProxy is doing a good job estimating data from each range, the range
* should be pretty close to `start_key`. */
QueryFilter indexFilter = QueryFilter.getSliceFilter(indexKey,
new QueryPath(indexCFS.getColumnFamilyName()),
startKey,
ArrayUtils.EMPTY_BYTE_ARRAY,
false,
clause.count);
ColumnFamily indexRow = indexCFS.getColumnFamily(indexFilter);
if (indexRow == null)
break;
byte[] dataKey = null;
int n = 0;
for (IColumn column : indexRow.getSortedColumns())
{
if (column.isMarkedForDelete())
continue;
dataKey = column.name();
n++;
DecoratedKey dk = partitioner.decorateKey(dataKey);
if (!range.right.equals(partitioner.getMinimumToken()) && range.right.compareTo(dk.token) < 0)
break outer;
if (!range.contains(dk.token))
continue;
// get the row columns requested, and additional columns for the expressions if necessary
ColumnFamily data = getColumnFamily(new QueryFilter(dk, path, firstFilter));
if (extraFilter != null)
{
// we might have gotten the expression columns in with the main data slice, but
// we can't know for sure until that slice is done. So, we'll do the extra query
// if we go through and any expression columns are not present.
for (IndexExpression expr : clause.expressions)
{
if (expr != primary && data.getColumn(expr.column_name) == null)
{
data.addAll(getColumnFamily(new QueryFilter(dk, path, extraFilter)));
break;
}
}
}
if (satisfies(data, clause, primary))
{
// cut the resultset back to what was requested, if necessary
if (firstFilter != dataFilter)
{
ColumnFamily expandedData = data;
data = expandedData.cloneMeShallow();
IColumnIterator iter = dataFilter.getMemtableColumnIterator(expandedData, dk, getComparator());
new QueryFilter(dk, path, dataFilter).collectCollatedColumns(data, iter, gcBefore());
}
rows.add(new Row(dk, data));
}
if (rows.size() == clause.count)
break outer;
}
startKey = dataKey;
if (n < clause.count)
break;
}
return rows;
}
private IndexExpression highestSelectivityPredicate(IndexClause clause)
{
IndexExpression best = null;
int bestMeanCount = Integer.MAX_VALUE;
for (IndexExpression expression : clause.expressions)
{
ColumnFamilyStore cfs = getIndexedColumnFamilyStore(expression.column_name);
if (cfs == null || !expression.op.equals(IndexOperator.EQ))
continue;
int columns = cfs.getMeanColumns();
if (columns < bestMeanCount)
{
best = expression;
bestMeanCount = columns;
}
}
return best;
}
private static boolean satisfies(ColumnFamily data, IndexClause clause, IndexExpression first)
{
for (IndexExpression expression : clause.expressions)
{
// (we can skip "first" since we already know it's satisfied)
if (expression == first)
continue;
// check column data vs expression
IColumn column = data.getColumn(expression.column_name);
if (column == null)
continue;
int v = data.getComparator().compare(column.value(), expression.value);
if (!satisfies(v, expression.op))
return false;
}
return true;
}
private static boolean satisfies(int comparison, IndexOperator op)
{
switch (op)
{
case EQ:
return comparison == 0;
case GTE:
return comparison >= 0;
case GT:
return comparison > 0;
case LTE:
return comparison <= 0;
case LT:
return comparison < 0;
default:
throw new IllegalStateException();
}
}
public AbstractType getComparator()
{
return metadata.comparator;
}
/**
* Take a snap shot of this columnfamily store.
*
* @param snapshotName the name of the associated with the snapshot
*/
public void snapshot(String snapshotName)
{
try
{
forceBlockingFlush();
}
catch (ExecutionException e)
{
throw new RuntimeException(e);
}
catch (InterruptedException e)
{
throw new AssertionError(e);
}
for (SSTableReader ssTable : ssTables)
{
try
{
// mkdir
File dataDirectory = ssTable.getDescriptor().directory.getParentFile();
String snapshotDirectoryPath = Table.getSnapshotPath(dataDirectory.getAbsolutePath(), table, snapshotName);
FileUtils.createDirectory(snapshotDirectoryPath);
// hard links
for (Component component : ssTable.getComponents())
{
File sourceFile = new File(ssTable.getDescriptor().filenameFor(component));
File targetLink = new File(snapshotDirectoryPath, sourceFile.getName());
FileUtils.createHardLink(sourceFile, targetLink);
}
if (logger.isDebugEnabled())
logger.debug("Snapshot for " + table + " keyspace data file " + ssTable.getFilename() +
" created in " + snapshotDirectoryPath);
}
catch (IOException e)
{
throw new IOError(e);
}
}
}
public void loadRowCache()
{
if (metadata.preloadRowCache)
{
logger.debug(String.format("Loading cache for keyspace/columnfamily %s/%s", table, columnFamily));
int ROWS = 4096;
Token min = partitioner.getMinimumToken();
Token start = min;
long i = 0;
while (i < ssTables.getRowCache().getCapacity())
{
List<Row> result;
try
{
result = getRangeSlice(null, new Bounds(start, min), ROWS, new IdentityQueryFilter());
}
catch (Exception e)
{
throw new RuntimeException(e);
}
for (Row row : result)
ssTables.getRowCache().put(row.key, row.cf);
i += result.size();
if (result.size() < ROWS)
break;
start = partitioner.getToken(result.get(ROWS - 1).key.key);
}
logger.info(String.format("Loaded %s rows into the %s cache", i, columnFamily));
}
}
public boolean hasUnreclaimedSpace()
{
return ssTables.getLiveSize() < ssTables.getTotalSize();
}
public long getTotalDiskSpaceUsed()
{
return ssTables.getTotalSize();
}
public long getLiveDiskSpaceUsed()
{
return ssTables.getLiveSize();
}
public int getLiveSSTableCount()
{
return ssTables.size();
}
/** raw cached row -- does not fetch the row if it is not present. not counted in cache statistics. */
public ColumnFamily getRawCachedRow(DecoratedKey key)
{
return ssTables.getRowCache().getCapacity() == 0 ? null : ssTables.getRowCache().getInternal(key);
}
void invalidateCachedRow(DecoratedKey key)
{
ssTables.getRowCache().remove(key);
}
public void forceMajorCompaction()
{
CompactionManager.instance.submitMajor(this);
}
public void invalidateRowCache()
{
ssTables.getRowCache().clear();
}
public int getKeyCacheSize()
{
return ssTables.getKeyCache().getCapacity();
}
public static Iterable<ColumnFamilyStore> all()
{
Iterable<ColumnFamilyStore>[] stores = new Iterable[DatabaseDescriptor.getTables().size()];
int i = 0;
for (Table table : Table.all())
{
stores[i++] = table.getColumnFamilyStores();
}
return Iterables.concat(stores);
}
public Iterable<DecoratedKey> allKeySamples()
{
Collection<SSTableReader> sstables = getSSTables();
Iterable<DecoratedKey>[] samples = new Iterable[sstables.size()];
int i = 0;
for (SSTableReader sstable: sstables)
{
samples[i++] = sstable.getKeySamples();
}
return Iterables.concat(samples);
}
/**
* for testing. no effort is made to clear historical memtables.
*/
void clearUnsafe()
{
memtable.clearUnsafe();
ssTables.clearUnsafe();
}
public Set<Memtable> getMemtablesPendingFlush()
{
return memtablesPendingFlush;
}
/**
* Truncate practically deletes the entire column family's data
* @return a Future to the delete operation. Call the future's get() to make
* sure the column family has been deleted
*/
public Future<?> truncate() throws IOException
{
// snapshot will also flush, but we want to truncate the most possible, and anything in a flush written
// after truncateAt won't be truncated.
try
{
forceBlockingFlush();
}
catch (Exception e)
{
throw new RuntimeException(e);
}
final long truncatedAt = System.currentTimeMillis();
snapshot(Table.getTimestampedSnapshotName("before-truncate"));
Runnable runnable = new WrappedRunnable()
{
public void runMayThrow() throws InterruptedException, IOException
{
// putting markCompacted on the commitlogUpdater thread ensures it will run
// after any compactions that were in progress when truncate was called, are finished
List<SSTableReader> truncatedSSTables = new ArrayList<SSTableReader>();
for (SSTableReader sstable : ssTables.getSSTables())
{
if (!sstable.newSince(truncatedAt))
truncatedSSTables.add(sstable);
}
markCompacted(truncatedSSTables);
// Invalidate row cache
invalidateRowCache();
}
};
return postFlushExecutor.submit(runnable);
}
// if this errors out, we are in a world of hurt.
public void renameSSTables(String newCfName) throws IOException
{
// complete as much of the job as possible. Don't let errors long the way prevent as much renaming as possible
// from happening.
IOException mostRecentProblem = null;
for (File existing : DefsTable.getFiles(table, columnFamily))
{
try
{
String newFileName = existing.getName().replaceFirst("\\w+-", newCfName + "-");
FileUtils.renameWithConfirm(existing, new File(existing.getParent(), newFileName));
}
catch (IOException ex)
{
mostRecentProblem = ex;
}
}
if (mostRecentProblem != null)
throw new IOException("One or more IOExceptions encountered while renaming files. Most recent problem is included.", mostRecentProblem);
for (ColumnFamilyStore indexedCfs : indexedColumns.values())
{
indexedCfs.renameSSTables(indexedCfs.columnFamily.replace(columnFamily, newCfName));
}
}
public static Future<?> submitPostFlush(Runnable runnable)
{
return postFlushExecutor.submit(runnable);
}
public long getBloomFilterFalsePositives()
{
long count = 0L;
for (SSTableReader sstable: getSSTables())
{
count += sstable.getBloomFilterFalsePositiveCount();
}
return count;
}
public long getRecentBloomFilterFalsePositives()
{
long count = 0L;
for (SSTableReader sstable: getSSTables())
{
count += sstable.getRecentBloomFilterFalsePositiveCount();
}
return count;
}
public double getBloomFilterFalseRatio()
{
long falseCount = 0L;
long trueCount = 0L;
for (SSTableReader sstable: getSSTables())
{
falseCount += sstable.getBloomFilterFalsePositiveCount();
trueCount += sstable.getBloomFilterTruePositiveCount();
}
if (falseCount == 0L && trueCount == 0L)
return 0d;
return (double) falseCount / (trueCount + falseCount);
}
public double getRecentBloomFilterFalseRatio()
{
long falseCount = 0L;
long trueCount = 0L;
for (SSTableReader sstable: getSSTables())
{
falseCount += sstable.getRecentBloomFilterFalsePositiveCount();
trueCount += sstable.getRecentBloomFilterTruePositiveCount();
}
if (falseCount == 0L && trueCount == 0L)
return 0d;
return (double) falseCount / (trueCount + falseCount);
}
public SortedSet<byte[]> getIndexedColumns()
{
return (SortedSet<byte[]>) indexedColumns.keySet();
}
public ColumnFamilyStore getIndexedColumnFamilyStore(byte[] column)
{
return indexedColumns.get(column);
}
public ColumnFamily newIndexedColumnFamily(byte[] column)
{
return ColumnFamily.create(indexedColumns.get(column).metadata);
}
public DecoratedKey<LocalToken> getIndexKeyFor(byte[] name, byte[] value)
{
return indexedColumns.get(name).partitioner.decorateKey(value);
}
@Override
public String toString()
{
return "ColumnFamilyStore(" +
"table='" + table + '\'' +
", columnFamily='" + columnFamily + '\'' +
')';
}
}