/*
* 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.io.sstable;
import java.io.BufferedInputStream;
import java.io.DataInputStream;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.RandomAccessFile;
import java.nio.ByteBuffer;
import java.util.*;
import java.util.concurrent.*;
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.concurrent.atomic.AtomicLong;
import com.google.common.annotations.VisibleForTesting;
import com.google.common.base.Predicate;
import com.google.common.collect.Iterators;
import com.google.common.collect.Ordering;
import com.google.common.primitives.Longs;
import com.google.common.util.concurrent.RateLimiter;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import com.clearspring.analytics.stream.cardinality.CardinalityMergeException;
import com.clearspring.analytics.stream.cardinality.HyperLogLogPlus;
import com.clearspring.analytics.stream.cardinality.ICardinality;
import org.apache.cassandra.cache.CachingOptions;
import org.apache.cassandra.cache.InstrumentingCache;
import org.apache.cassandra.cache.KeyCacheKey;
import org.apache.cassandra.concurrent.DebuggableThreadPoolExecutor;
import org.apache.cassandra.concurrent.ScheduledExecutors;
import org.apache.cassandra.config.CFMetaData;
import org.apache.cassandra.config.ColumnDefinition;
import org.apache.cassandra.config.Config;
import org.apache.cassandra.config.DatabaseDescriptor;
import org.apache.cassandra.config.Schema;
import org.apache.cassandra.db.ColumnFamilyStore;
import org.apache.cassandra.db.DataRange;
import org.apache.cassandra.db.DataTracker;
import org.apache.cassandra.db.DecoratedKey;
import org.apache.cassandra.db.Keyspace;
import org.apache.cassandra.db.RowIndexEntry;
import org.apache.cassandra.db.RowPosition;
import org.apache.cassandra.db.SystemKeyspace;
import org.apache.cassandra.db.commitlog.ReplayPosition;
import org.apache.cassandra.db.index.SecondaryIndex;
import org.apache.cassandra.dht.AbstractBounds;
import org.apache.cassandra.dht.IPartitioner;
import org.apache.cassandra.dht.LocalPartitioner;
import org.apache.cassandra.dht.Range;
import org.apache.cassandra.dht.Token;
import org.apache.cassandra.io.compress.CompressionMetadata;
import org.apache.cassandra.io.sstable.metadata.CompactionMetadata;
import org.apache.cassandra.io.sstable.metadata.MetadataComponent;
import org.apache.cassandra.io.sstable.metadata.MetadataType;
import org.apache.cassandra.io.sstable.metadata.StatsMetadata;
import org.apache.cassandra.io.sstable.metadata.ValidationMetadata;
import org.apache.cassandra.io.util.BufferedSegmentedFile;
import org.apache.cassandra.io.util.CompressedSegmentedFile;
import org.apache.cassandra.io.util.DataOutputStreamAndChannel;
import org.apache.cassandra.io.util.FileDataInput;
import org.apache.cassandra.io.util.FileUtils;
import org.apache.cassandra.io.util.ICompressedFile;
import org.apache.cassandra.io.util.RandomAccessReader;
import org.apache.cassandra.io.util.SegmentedFile;
import org.apache.cassandra.metrics.RestorableMeter;
import org.apache.cassandra.metrics.StorageMetrics;
import org.apache.cassandra.service.ActiveRepairService;
import org.apache.cassandra.service.CacheService;
import org.apache.cassandra.service.StorageService;
import org.apache.cassandra.tracing.Tracing;
import org.apache.cassandra.utils.ByteBufferUtil;
import org.apache.cassandra.utils.CLibrary;
import org.apache.cassandra.utils.EstimatedHistogram;
import org.apache.cassandra.utils.FBUtilities;
import org.apache.cassandra.utils.FilterFactory;
import org.apache.cassandra.utils.IFilter;
import org.apache.cassandra.utils.Pair;
import org.apache.cassandra.utils.concurrent.OpOrder;
import org.apache.cassandra.utils.concurrent.Ref;
import org.apache.cassandra.utils.concurrent.RefCounted;
import org.apache.cassandra.utils.concurrent.SelfRefCounted;
import static org.apache.cassandra.db.Directories.SECONDARY_INDEX_NAME_SEPARATOR;
/**
* An SSTableReader can be constructed in a number of places, but typically is either
* read from disk at startup, or constructed from a flushed memtable, or after compaction
* to replace some existing sstables. However once created, an sstablereader may also be modified.
*
* A reader's OpenReason describes its current stage in its lifecycle, as follows:
*
* NORMAL
* From: None => Reader has been read from disk, either at startup or from a flushed memtable
* EARLY => Reader is the final result of a compaction
* MOVED_START => Reader WAS being compacted, but this failed and it has been restored to NORMAL status
*
* EARLY
* From: None => Reader is a compaction replacement that is either incomplete and has been opened
* to represent its partial result status, or has been finished but the compaction
* it is a part of has not yet completed fully
* EARLY => Same as from None, only it is not the first time it has been
*
* MOVED_START
* From: NORMAL => Reader is being compacted. This compaction has not finished, but the compaction result
* is either partially or fully opened, to either partially or fully replace this reader.
* This reader's start key has been updated to represent this, so that reads only hit
* one or the other reader.
*
* METADATA_CHANGE
* From: NORMAL => Reader has seen low traffic and the amount of memory available for index summaries is
* constrained, so its index summary has been downsampled.
* METADATA_CHANGE => Same
*
* Note that in parallel to this, there are two different Descriptor types; TMPLINK and FINAL; the latter corresponds
* to NORMAL state readers and all readers that replace a NORMAL one. TMPLINK is used for EARLY state readers and
* no others.
*
* When a reader is being compacted, if the result is large its replacement may be opened as EARLY before compaction
* completes in order to present the result to consumers earlier. In this case the reader will itself be changed to
* a MOVED_START state, where its start no longer represents its on-disk minimum key. This is to permit reads to be
* directed to only one reader when the two represent the same data. The EARLY file can represent a compaction result
* that is either partially complete and still in-progress, or a complete and immutable sstable that is part of a larger
* macro compaction action that has not yet fully completed.
*
* Currently ALL compaction results at least briefly go through an EARLY open state prior to completion, regardless
* of if early opening is enabled.
*
* Since a reader can be created multiple times over the same shared underlying resources, and the exact resources
* it shares between each instance differ subtly, we track the lifetime of any underlying resource with its own
* reference count, which each instance takes a Ref to. Each instance then tracks references to itself, and once these
* all expire it releases its Refs to these underlying resources.
*
* There is some shared cleanup behaviour needed only once all sstablereaders in a certain stage of their lifecycle
* (i.e. EARLY or NORMAL opening), and some that must only occur once all readers of any kind over a single logical
* sstable have expired. These are managed by the TypeTidy and GlobalTidy classes at the bottom, and are effectively
* managed as another resource each instance tracks its own Ref instance to, to ensure all of these resources are
* cleaned up safely and can be debugged otherwise.
*
* TODO: fill in details about DataTracker and lifecycle interactions for tools, and for compaction strategies
*/
public class SSTableReader extends SSTable implements SelfRefCounted<SSTableReader>
{
private static final Logger logger = LoggerFactory.getLogger(SSTableReader.class);
private static final ScheduledThreadPoolExecutor syncExecutor = new ScheduledThreadPoolExecutor(1);
private static final RateLimiter meterSyncThrottle = RateLimiter.create(100.0);
public static final Comparator<SSTableReader> maxTimestampComparator = new Comparator<SSTableReader>()
{
public int compare(SSTableReader o1, SSTableReader o2)
{
long ts1 = o1.getMaxTimestamp();
long ts2 = o2.getMaxTimestamp();
return (ts1 > ts2 ? -1 : (ts1 == ts2 ? 0 : 1));
}
};
public static final Comparator<SSTableReader> sstableComparator = new Comparator<SSTableReader>()
{
public int compare(SSTableReader o1, SSTableReader o2)
{
return o1.first.compareTo(o2.first);
}
};
public static final Ordering<SSTableReader> sstableOrdering = Ordering.from(sstableComparator);
/**
* maxDataAge is a timestamp in local server time (e.g. System.currentTimeMilli) which represents an upper bound
* to the newest piece of data stored in the sstable. In other words, this sstable does not contain items created
* later than maxDataAge.
*
* The field is not serialized to disk, so relying on it for more than what truncate does is not advised.
*
* When a new sstable is flushed, maxDataAge is set to the time of creation.
* When a sstable is created from compaction, maxDataAge is set to max of all merged sstables.
*
* The age is in milliseconds since epoc and is local to this host.
*/
public final long maxDataAge;
public enum OpenReason
{
NORMAL,
EARLY,
METADATA_CHANGE,
MOVED_START,
SHADOWED // => MOVED_START past end
}
public final OpenReason openReason;
// indexfile and datafile: might be null before a call to load()
private SegmentedFile ifile;
private SegmentedFile dfile;
private IndexSummary indexSummary;
private IFilter bf;
private InstrumentingCache<KeyCacheKey, RowIndexEntry> keyCache;
private final BloomFilterTracker bloomFilterTracker = new BloomFilterTracker();
// technically isCompacted is not necessary since it should never be unreferenced unless it is also compacted,
// but it seems like a good extra layer of protection against reference counting bugs to not delete data based on that alone
private AtomicBoolean isSuspect = new AtomicBoolean(false);
// not final since we need to be able to change level on a file.
private volatile StatsMetadata sstableMetadata;
private final AtomicLong keyCacheHit = new AtomicLong(0);
private final AtomicLong keyCacheRequest = new AtomicLong(0);
private final InstanceTidier tidy = new InstanceTidier(descriptor, metadata);
private final Ref<SSTableReader> selfRef = new Ref<>(this, tidy);
private RestorableMeter readMeter;
/**
* Calculate approximate key count.
* If cardinality estimator is available on all given sstables, then this method use them to estimate
* key count.
* If not, then this uses index summaries.
*
* @param sstables SSTables to calculate key count
* @return estimated key count
*/
public static long getApproximateKeyCount(Collection<SSTableReader> sstables)
{
long count = -1;
// check if cardinality estimator is available for all SSTables
boolean cardinalityAvailable = !sstables.isEmpty() && Iterators.all(sstables.iterator(), new Predicate<SSTableReader>()
{
public boolean apply(SSTableReader sstable)
{
return sstable.descriptor.version.newStatsFile;
}
});
// if it is, load them to estimate key count
if (cardinalityAvailable)
{
boolean failed = false;
ICardinality cardinality = null;
for (SSTableReader sstable : sstables)
{
try
{
CompactionMetadata metadata = (CompactionMetadata) sstable.descriptor.getMetadataSerializer().deserialize(sstable.descriptor, MetadataType.COMPACTION);
assert metadata != null : sstable.getFilename();
if (cardinality == null)
cardinality = metadata.cardinalityEstimator;
else
cardinality = cardinality.merge(metadata.cardinalityEstimator);
}
catch (IOException e)
{
logger.warn("Reading cardinality from Statistics.db failed.", e);
failed = true;
break;
}
catch (CardinalityMergeException e)
{
logger.warn("Cardinality merge failed.", e);
failed = true;
break;
}
}
if (cardinality != null && !failed)
count = cardinality.cardinality();
}
// if something went wrong above or cardinality is not available, calculate using index summary
if (count < 0)
{
for (SSTableReader sstable : sstables)
count += sstable.estimatedKeys();
}
return count;
}
/**
* Estimates how much of the keys we would keep if the sstables were compacted together
*/
public static double estimateCompactionGain(Set<SSTableReader> overlapping)
{
Set<ICardinality> cardinalities = new HashSet<>(overlapping.size());
for (SSTableReader sstable : overlapping)
{
try
{
ICardinality cardinality = ((CompactionMetadata) sstable.descriptor.getMetadataSerializer().deserialize(sstable.descriptor, MetadataType.COMPACTION)).cardinalityEstimator;
if (cardinality != null)
cardinalities.add(cardinality);
else
logger.debug("Got a null cardinality estimator in: "+sstable.getFilename());
}
catch (IOException e)
{
logger.warn("Could not read up compaction metadata for " + sstable, e);
}
}
long totalKeyCountBefore = 0;
for (ICardinality cardinality : cardinalities)
{
totalKeyCountBefore += cardinality.cardinality();
}
if (totalKeyCountBefore == 0)
return 1;
long totalKeyCountAfter = mergeCardinalities(cardinalities).cardinality();
logger.debug("Estimated compaction gain: {}/{}={}", totalKeyCountAfter, totalKeyCountBefore, ((double)totalKeyCountAfter)/totalKeyCountBefore);
return ((double)totalKeyCountAfter)/totalKeyCountBefore;
}
private static ICardinality mergeCardinalities(Collection<ICardinality> cardinalities)
{
ICardinality base = new HyperLogLogPlus(13, 25); // see MetadataCollector.cardinality
try
{
base = base.merge(cardinalities.toArray(new ICardinality[cardinalities.size()]));
}
catch (CardinalityMergeException e)
{
logger.warn("Could not merge cardinalities", e);
}
return base;
}
public static SSTableReader open(Descriptor descriptor) throws IOException
{
CFMetaData metadata;
if (descriptor.cfname.contains(SECONDARY_INDEX_NAME_SEPARATOR))
{
int i = descriptor.cfname.indexOf(SECONDARY_INDEX_NAME_SEPARATOR);
String parentName = descriptor.cfname.substring(0, i);
CFMetaData parent = Schema.instance.getCFMetaData(descriptor.ksname, parentName);
ColumnDefinition def = parent.getColumnDefinitionForIndex(descriptor.cfname.substring(i + 1));
metadata = CFMetaData.newIndexMetadata(parent, def, SecondaryIndex.getIndexComparator(parent, def));
}
else
{
metadata = Schema.instance.getCFMetaData(descriptor.ksname, descriptor.cfname);
}
return open(descriptor, metadata);
}
public static SSTableReader open(Descriptor desc, CFMetaData metadata) throws IOException
{
IPartitioner p = desc.cfname.contains(SECONDARY_INDEX_NAME_SEPARATOR)
? new LocalPartitioner(metadata.getKeyValidator())
: StorageService.getPartitioner();
return open(desc, componentsFor(desc), metadata, p);
}
public static SSTableReader open(Descriptor descriptor, Set<Component> components, CFMetaData metadata, IPartitioner partitioner) throws IOException
{
return open(descriptor, components, metadata, partitioner, true);
}
// use only for offline or "Standalone" operations
public static SSTableReader openNoValidation(Descriptor descriptor, Set<Component> components, CFMetaData metadata) throws IOException
{
return open(descriptor, components, metadata, StorageService.getPartitioner(), false);
}
/**
* Open SSTable reader to be used in batch mode(such as sstableloader).
*
* @param descriptor
* @param components
* @param metadata
* @param partitioner
* @return opened SSTableReader
* @throws IOException
*/
public static SSTableReader openForBatch(Descriptor descriptor, Set<Component> components, CFMetaData metadata, IPartitioner partitioner) throws IOException
{
// Minimum components without which we can't do anything
assert components.contains(Component.DATA) : "Data component is missing for sstable" + descriptor;
assert components.contains(Component.PRIMARY_INDEX) : "Primary index component is missing for sstable " + descriptor;
Map<MetadataType, MetadataComponent> sstableMetadata = descriptor.getMetadataSerializer().deserialize(descriptor,
EnumSet.of(MetadataType.VALIDATION, MetadataType.STATS));
ValidationMetadata validationMetadata = (ValidationMetadata) sstableMetadata.get(MetadataType.VALIDATION);
StatsMetadata statsMetadata = (StatsMetadata) sstableMetadata.get(MetadataType.STATS);
// Check if sstable is created using same partitioner.
// Partitioner can be null, which indicates older version of sstable or no stats available.
// In that case, we skip the check.
String partitionerName = partitioner.getClass().getCanonicalName();
if (validationMetadata != null && !partitionerName.equals(validationMetadata.partitioner))
{
logger.error(String.format("Cannot open %s; partitioner %s does not match system partitioner %s. Note that the default partitioner starting with Cassandra 1.2 is Murmur3Partitioner, so you will need to edit that to match your old partitioner if upgrading.",
descriptor, validationMetadata.partitioner, partitionerName));
System.exit(1);
}
logger.info("Opening {} ({} bytes)", descriptor, new File(descriptor.filenameFor(Component.DATA)).length());
SSTableReader sstable = new SSTableReader(descriptor,
components,
metadata,
partitioner,
System.currentTimeMillis(),
statsMetadata,
OpenReason.NORMAL);
// special implementation of load to use non-pooled SegmentedFile builders
SegmentedFile.Builder ibuilder = new BufferedSegmentedFile.Builder();
SegmentedFile.Builder dbuilder = sstable.compression
? new CompressedSegmentedFile.Builder(null)
: new BufferedSegmentedFile.Builder();
if (!sstable.loadSummary(ibuilder, dbuilder))
sstable.buildSummary(false, ibuilder, dbuilder, false, Downsampling.BASE_SAMPLING_LEVEL);
sstable.ifile = ibuilder.complete(sstable.descriptor.filenameFor(Component.PRIMARY_INDEX));
sstable.dfile = dbuilder.complete(sstable.descriptor.filenameFor(Component.DATA));
sstable.bf = FilterFactory.AlwaysPresent;
sstable.setup(true);
return sstable;
}
private static SSTableReader open(Descriptor descriptor,
Set<Component> components,
CFMetaData metadata,
IPartitioner partitioner,
boolean validate) throws IOException
{
// Minimum components without which we can't do anything
assert components.contains(Component.DATA) : "Data component is missing for sstable" + descriptor;
assert components.contains(Component.PRIMARY_INDEX) : "Primary index component is missing for sstable " + descriptor;
Map<MetadataType, MetadataComponent> sstableMetadata = descriptor.getMetadataSerializer().deserialize(descriptor,
EnumSet.of(MetadataType.VALIDATION, MetadataType.STATS));
ValidationMetadata validationMetadata = (ValidationMetadata) sstableMetadata.get(MetadataType.VALIDATION);
StatsMetadata statsMetadata = (StatsMetadata) sstableMetadata.get(MetadataType.STATS);
// Check if sstable is created using same partitioner.
// Partitioner can be null, which indicates older version of sstable or no stats available.
// In that case, we skip the check.
String partitionerName = partitioner.getClass().getCanonicalName();
if (validationMetadata != null && !partitionerName.equals(validationMetadata.partitioner))
{
logger.error(String.format("Cannot open %s; partitioner %s does not match system partitioner %s. Note that the default partitioner starting with Cassandra 1.2 is Murmur3Partitioner, so you will need to edit that to match your old partitioner if upgrading.",
descriptor, validationMetadata.partitioner, partitionerName));
System.exit(1);
}
logger.info("Opening {} ({} bytes)", descriptor, new File(descriptor.filenameFor(Component.DATA)).length());
SSTableReader sstable = new SSTableReader(descriptor,
components,
metadata,
partitioner,
System.currentTimeMillis(),
statsMetadata,
OpenReason.NORMAL);
// load index and filter
long start = System.nanoTime();
sstable.load(validationMetadata);
logger.debug("INDEX LOAD TIME for {}: {} ms.", descriptor, TimeUnit.NANOSECONDS.toMillis(System.nanoTime() - start));
sstable.setup(!validate);
if (validate)
sstable.validate();
if (sstable.getKeyCache() != null)
logger.debug("key cache contains {}/{} keys", sstable.getKeyCache().size(), sstable.getKeyCache().getCapacity());
return sstable;
}
public static void logOpenException(Descriptor descriptor, IOException e)
{
if (e instanceof FileNotFoundException)
logger.error("Missing sstable component in {}; skipped because of {}", descriptor, e.getMessage());
else
logger.error("Corrupt sstable {}; skipped", descriptor, e);
}
public static Collection<SSTableReader> openAll(Set<Map.Entry<Descriptor, Set<Component>>> entries,
final CFMetaData metadata,
final IPartitioner partitioner)
{
final Collection<SSTableReader> sstables = new LinkedBlockingQueue<>();
ExecutorService executor = DebuggableThreadPoolExecutor.createWithFixedPoolSize("SSTableBatchOpen", FBUtilities.getAvailableProcessors());
for (final Map.Entry<Descriptor, Set<Component>> entry : entries)
{
Runnable runnable = new Runnable()
{
public void run()
{
SSTableReader sstable;
try
{
sstable = open(entry.getKey(), entry.getValue(), metadata, partitioner);
}
catch (IOException ex)
{
logger.error("Corrupt sstable {}; skipped", entry, ex);
return;
}
sstables.add(sstable);
}
};
executor.submit(runnable);
}
executor.shutdown();
try
{
executor.awaitTermination(7, TimeUnit.DAYS);
}
catch (InterruptedException e)
{
throw new AssertionError(e);
}
return sstables;
}
/**
* Open a RowIndexedReader which already has its state initialized (by SSTableWriter).
*/
static SSTableReader internalOpen(Descriptor desc,
Set<Component> components,
CFMetaData metadata,
IPartitioner partitioner,
SegmentedFile ifile,
SegmentedFile dfile,
IndexSummary isummary,
IFilter bf,
long maxDataAge,
StatsMetadata sstableMetadata,
OpenReason openReason)
{
assert desc != null && partitioner != null && ifile != null && dfile != null && isummary != null && bf != null && sstableMetadata != null;
return new SSTableReader(desc,
components,
metadata,
partitioner,
ifile, dfile,
isummary,
bf,
maxDataAge,
sstableMetadata,
openReason);
}
private SSTableReader(final Descriptor desc,
Set<Component> components,
CFMetaData metadata,
IPartitioner partitioner,
long maxDataAge,
StatsMetadata sstableMetadata,
OpenReason openReason)
{
super(desc, components, metadata, partitioner);
this.sstableMetadata = sstableMetadata;
this.maxDataAge = maxDataAge;
this.openReason = openReason;
}
private SSTableReader(Descriptor desc,
Set<Component> components,
CFMetaData metadata,
IPartitioner partitioner,
SegmentedFile ifile,
SegmentedFile dfile,
IndexSummary indexSummary,
IFilter bloomFilter,
long maxDataAge,
StatsMetadata sstableMetadata,
OpenReason openReason)
{
this(desc, components, metadata, partitioner, maxDataAge, sstableMetadata, openReason);
this.ifile = ifile;
this.dfile = dfile;
this.indexSummary = indexSummary;
this.bf = bloomFilter;
this.setup(false);
}
public static long getTotalBytes(Iterable<SSTableReader> sstables)
{
long sum = 0;
for (SSTableReader sstable : sstables)
sum += sstable.onDiskLength();
return sum;
}
public static long getTotalUncompressedBytes(Iterable<SSTableReader> sstables)
{
long sum = 0;
for (SSTableReader sstable : sstables)
sum += sstable.uncompressedLength();
return sum;
}
public boolean equals(Object that)
{
return that instanceof SSTableReader && ((SSTableReader) that).descriptor.equals(this.descriptor);
}
public int hashCode()
{
return this.descriptor.hashCode();
}
public String getFilename()
{
return dfile.path;
}
public String getIndexFilename()
{
return ifile.path;
}
public void setTrackedBy(DataTracker tracker)
{
tidy.type.deletingTask.setTracker(tracker);
// under normal operation we can do this at any time, but SSTR is also used outside C* proper,
// e.g. by BulkLoader, which does not initialize the cache. As a kludge, we set up the cache
// here when we know we're being wired into the rest of the server infrastructure.
keyCache = CacheService.instance.keyCache;
}
private void load(ValidationMetadata validation) throws IOException
{
if (metadata.getBloomFilterFpChance() == 1.0)
{
// bf is disabled.
load(false, true);
bf = FilterFactory.AlwaysPresent;
}
else if (!components.contains(Component.FILTER) || validation == null)
{
// bf is enabled, but filter component is missing.
load(true, true);
}
else if (validation.bloomFilterFPChance != metadata.getBloomFilterFpChance())
{
// bf fp chance in sstable metadata and it has changed since compaction.
load(true, true);
}
else
{
// bf is enabled and fp chance matches the currently configured value.
load(false, true);
loadBloomFilter();
}
}
/**
* Load bloom filter from Filter.db file.
*
* @throws IOException
*/
private void loadBloomFilter() throws IOException
{
DataInputStream stream = null;
try
{
stream = new DataInputStream(new BufferedInputStream(new FileInputStream(descriptor.filenameFor(Component.FILTER))));
bf = FilterFactory.deserialize(stream, true);
}
finally
{
FileUtils.closeQuietly(stream);
}
}
/**
* Loads ifile, dfile and indexSummary, and optionally recreates the bloom filter.
* @param saveSummaryIfCreated for bulk loading purposes, if the summary was absent and needed to be built, you can
* avoid persisting it to disk by setting this to false
*/
private void load(boolean recreateBloomFilter, boolean saveSummaryIfCreated) throws IOException
{
SegmentedFile.Builder ibuilder = SegmentedFile.getBuilder(DatabaseDescriptor.getIndexAccessMode());
SegmentedFile.Builder dbuilder = compression
? SegmentedFile.getCompressedBuilder()
: SegmentedFile.getBuilder(DatabaseDescriptor.getDiskAccessMode());
boolean summaryLoaded = loadSummary(ibuilder, dbuilder);
boolean builtSummary = false;
if (recreateBloomFilter || !summaryLoaded)
{
buildSummary(recreateBloomFilter, ibuilder, dbuilder, summaryLoaded, Downsampling.BASE_SAMPLING_LEVEL);
builtSummary = true;
}
ifile = ibuilder.complete(descriptor.filenameFor(Component.PRIMARY_INDEX));
dfile = dbuilder.complete(descriptor.filenameFor(Component.DATA));
// Check for an index summary that was downsampled even though the serialization format doesn't support
// that. If it was downsampled, rebuild it. See CASSANDRA-8993 for details.
if (!descriptor.version.hasSamplingLevel && !builtSummary && !validateSummarySamplingLevel())
{
indexSummary.close();
ifile.close();
dfile.close();
logger.info("Detected erroneously downsampled index summary; will rebuild summary at full sampling");
FileUtils.deleteWithConfirm(new File(descriptor.filenameFor(Component.SUMMARY)));
ibuilder = SegmentedFile.getBuilder(DatabaseDescriptor.getIndexAccessMode());
dbuilder = compression
? SegmentedFile.getCompressedBuilder()
: SegmentedFile.getBuilder(DatabaseDescriptor.getDiskAccessMode());
buildSummary(false, ibuilder, dbuilder, false, Downsampling.BASE_SAMPLING_LEVEL);
ifile = ibuilder.complete(descriptor.filenameFor(Component.PRIMARY_INDEX));
dfile = dbuilder.complete(descriptor.filenameFor(Component.DATA));
saveSummary(ibuilder, dbuilder);
}
else if (saveSummaryIfCreated && builtSummary)
{
saveSummary(ibuilder, dbuilder);
}
}
/**
* Build index summary(and optionally bloom filter) by reading through Index.db file.
*
* @param recreateBloomFilter true if recreate bloom filter
* @param ibuilder
* @param dbuilder
* @param summaryLoaded true if index summary is already loaded and not need to build again
* @throws IOException
*/
private void buildSummary(boolean recreateBloomFilter, SegmentedFile.Builder ibuilder, SegmentedFile.Builder dbuilder, boolean summaryLoaded, int samplingLevel) throws IOException
{
// we read the positions in a BRAF so we don't have to worry about an entry spanning a mmap boundary.
RandomAccessReader primaryIndex = RandomAccessReader.open(new File(descriptor.filenameFor(Component.PRIMARY_INDEX)));
try
{
long indexSize = primaryIndex.length();
long histogramCount = sstableMetadata.estimatedRowSize.count();
long estimatedKeys = histogramCount > 0 && !sstableMetadata.estimatedRowSize.isOverflowed()
? histogramCount
: estimateRowsFromIndex(primaryIndex); // statistics is supposed to be optional
try(IndexSummaryBuilder summaryBuilder = summaryLoaded ? null : new IndexSummaryBuilder(estimatedKeys, metadata.getMinIndexInterval(), samplingLevel))
{
if (recreateBloomFilter)
bf = FilterFactory.getFilter(estimatedKeys, metadata.getBloomFilterFpChance(), true);
long indexPosition;
while ((indexPosition = primaryIndex.getFilePointer()) != indexSize)
{
ByteBuffer key = ByteBufferUtil.readWithShortLength(primaryIndex);
RowIndexEntry indexEntry = metadata.comparator.rowIndexEntrySerializer().deserialize(primaryIndex, descriptor.version);
DecoratedKey decoratedKey = partitioner.decorateKey(key);
if (first == null)
first = decoratedKey;
last = decoratedKey;
if (recreateBloomFilter)
bf.add(decoratedKey.getKey());
// if summary was already read from disk we don't want to re-populate it using primary index
if (!summaryLoaded)
{
summaryBuilder.maybeAddEntry(decoratedKey, indexPosition);
ibuilder.addPotentialBoundary(indexPosition);
dbuilder.addPotentialBoundary(indexEntry.position);
}
}
if (!summaryLoaded)
indexSummary = summaryBuilder.build(partitioner);
}
}
finally
{
FileUtils.closeQuietly(primaryIndex);
}
first = getMinimalKey(first);
last = getMinimalKey(last);
}
/**
* Load index summary from Summary.db file if it exists.
*
* if loaded index summary has different index interval from current value stored in schema,
* then Summary.db file will be deleted and this returns false to rebuild summary.
*
* @param ibuilder
* @param dbuilder
* @return true if index summary is loaded successfully from Summary.db file.
*/
public boolean loadSummary(SegmentedFile.Builder ibuilder, SegmentedFile.Builder dbuilder)
{
File summariesFile = new File(descriptor.filenameFor(Component.SUMMARY));
if (!summariesFile.exists())
return false;
DataInputStream iStream = null;
try
{
iStream = new DataInputStream(new FileInputStream(summariesFile));
indexSummary = IndexSummary.serializer.deserialize(
iStream, partitioner, descriptor.version.hasSamplingLevel,
metadata.getMinIndexInterval(), metadata.getMaxIndexInterval());
first = partitioner.decorateKey(ByteBufferUtil.readWithLength(iStream));
last = partitioner.decorateKey(ByteBufferUtil.readWithLength(iStream));
ibuilder.deserializeBounds(iStream);
dbuilder.deserializeBounds(iStream);
}
catch (IOException e)
{
if (indexSummary != null)
indexSummary.close();
logger.debug("Cannot deserialize SSTable Summary File {}: {}", summariesFile.getPath(), e.getMessage());
// corrupted; delete it and fall back to creating a new summary
FileUtils.closeQuietly(iStream);
// delete it and fall back to creating a new summary
FileUtils.deleteWithConfirm(summariesFile);
return false;
}
finally
{
FileUtils.closeQuietly(iStream);
}
return true;
}
/**
* Validates that an index summary has full sampling, as expected when the serialization format does not support
* persisting the sampling level.
* @return true if the summary has full sampling, false otherwise
*/
private boolean validateSummarySamplingLevel()
{
// We need to check index summary entries against the index to verify that none of them were dropped due to
// downsampling. Downsampling can drop any of the first BASE_SAMPLING_LEVEL entries (repeating that drop pattern
// for the remainder of the summary). Unfortunately, the first entry to be dropped is the entry at
// index (BASE_SAMPLING_LEVEL - 1), so we need to check a full set of BASE_SAMPLING_LEVEL entries.
Iterator<FileDataInput> segments = ifile.iterator(0);
int i = 0;
int summaryEntriesChecked = 0;
int expectedIndexInterval = getMinIndexInterval();
while (segments.hasNext())
{
FileDataInput in = segments.next();
try
{
while (!in.isEOF())
{
ByteBuffer indexKey = ByteBufferUtil.readWithShortLength(in);
if (i % expectedIndexInterval == 0)
{
ByteBuffer summaryKey = ByteBuffer.wrap(indexSummary.getKey(i / expectedIndexInterval));
if (!summaryKey.equals(indexKey))
return false;
summaryEntriesChecked++;
if (summaryEntriesChecked == Downsampling.BASE_SAMPLING_LEVEL)
return true;
}
RowIndexEntry.Serializer.skip(in);
i++;
}
}
catch (IOException e)
{
markSuspect();
throw new CorruptSSTableException(e, in.getPath());
}
finally
{
FileUtils.closeQuietly(in);
}
}
return true;
}
/**
* Save index summary to Summary.db file.
*
* @param ibuilder
* @param dbuilder
*/
public void saveSummary(SegmentedFile.Builder ibuilder, SegmentedFile.Builder dbuilder)
{
saveSummary(ibuilder, dbuilder, indexSummary);
}
private void saveSummary(SegmentedFile.Builder ibuilder, SegmentedFile.Builder dbuilder, IndexSummary summary)
{
File summariesFile = new File(descriptor.filenameFor(Component.SUMMARY));
if (summariesFile.exists())
FileUtils.deleteWithConfirm(summariesFile);
DataOutputStreamAndChannel oStream = null;
try
{
oStream = new DataOutputStreamAndChannel(new FileOutputStream(summariesFile));
IndexSummary.serializer.serialize(summary, oStream, descriptor.version.hasSamplingLevel);
ByteBufferUtil.writeWithLength(first.getKey(), oStream);
ByteBufferUtil.writeWithLength(last.getKey(), oStream);
ibuilder.serializeBounds(oStream);
dbuilder.serializeBounds(oStream);
}
catch (IOException e)
{
logger.debug("Cannot save SSTable Summary: ", e);
// corrupted hence delete it and let it load it now.
if (summariesFile.exists())
FileUtils.deleteWithConfirm(summariesFile);
}
finally
{
FileUtils.closeQuietly(oStream);
}
}
public void setReplacedBy(SSTableReader replacement)
{
synchronized (tidy.global)
{
assert replacement != null;
assert !tidy.isReplaced;
assert tidy.global.live == this;
tidy.isReplaced = true;
tidy.global.live = replacement;
}
}
public SSTableReader cloneWithNewStart(DecoratedKey newStart, final Runnable runOnClose)
{
synchronized (tidy.global)
{
assert openReason != OpenReason.EARLY;
SSTableReader replacement = new SSTableReader(descriptor, components, metadata, partitioner, ifile.sharedCopy(),
dfile.sharedCopy(), indexSummary.sharedCopy(), bf.sharedCopy(),
maxDataAge, sstableMetadata, OpenReason.MOVED_START);
// TODO: make data/index start accurate for compressed files
// TODO: merge with caller's firstKeyBeyond() work,to save time
if (newStart.compareTo(first) > 0)
{
final long dataStart = getPosition(newStart, Operator.EQ).position;
final long indexStart = getIndexScanPosition(newStart);
this.tidy.runOnClose = new Runnable()
{
public void run()
{
dfile.dropPageCache(dataStart);
ifile.dropPageCache(indexStart);
if (runOnClose != null)
runOnClose.run();
}
};
}
replacement.first = newStart;
replacement.last = this.last;
setReplacedBy(replacement);
return replacement;
}
}
public SSTableReader cloneAsShadowed(final Runnable runOnClose)
{
synchronized (tidy.global)
{
assert openReason != OpenReason.EARLY;
this.tidy.runOnClose = new Runnable()
{
public void run()
{
dfile.dropPageCache(0);
ifile.dropPageCache(0);
runOnClose.run();
}
};
SSTableReader replacement = new SSTableReader(descriptor, components, metadata, partitioner, ifile.sharedCopy(),
dfile.sharedCopy(), indexSummary.sharedCopy(), bf.sharedCopy(),
maxDataAge, sstableMetadata, OpenReason.SHADOWED);
replacement.first = first;
replacement.last = last;
setReplacedBy(replacement);
return replacement;
}
}
/**
* Returns a new SSTableReader with the same properties as this SSTableReader except that a new IndexSummary will
* be built at the target samplingLevel. This (original) SSTableReader instance will be marked as replaced, have
* its DeletingTask removed, and have its periodic read-meter sync task cancelled.
* @param samplingLevel the desired sampling level for the index summary on the new SSTableReader
* @return a new SSTableReader
* @throws IOException
*/
public SSTableReader cloneWithNewSummarySamplingLevel(ColumnFamilyStore parent, int samplingLevel) throws IOException
{
assert descriptor.version.hasSamplingLevel;
synchronized (tidy.global)
{
assert openReason != OpenReason.EARLY;
int minIndexInterval = metadata.getMinIndexInterval();
int maxIndexInterval = metadata.getMaxIndexInterval();
double effectiveInterval = indexSummary.getEffectiveIndexInterval();
IndexSummary newSummary;
long oldSize = bytesOnDisk();
// We have to rebuild the summary from the on-disk primary index in three cases:
// 1. The sampling level went up, so we need to read more entries off disk
// 2. The min_index_interval changed (in either direction); this changes what entries would be in the summary
// at full sampling (and consequently at any other sampling level)
// 3. The max_index_interval was lowered, forcing us to raise the sampling level
if (samplingLevel > indexSummary.getSamplingLevel() || indexSummary.getMinIndexInterval() != minIndexInterval || effectiveInterval > maxIndexInterval)
{
newSummary = buildSummaryAtLevel(samplingLevel);
}
else if (samplingLevel < indexSummary.getSamplingLevel())
{
// we can use the existing index summary to make a smaller one
newSummary = IndexSummaryBuilder.downsample(indexSummary, samplingLevel, minIndexInterval, partitioner);
SegmentedFile.Builder ibuilder = SegmentedFile.getBuilder(DatabaseDescriptor.getIndexAccessMode());
SegmentedFile.Builder dbuilder = compression
? SegmentedFile.getCompressedBuilder()
: SegmentedFile.getBuilder(DatabaseDescriptor.getDiskAccessMode());
saveSummary(ibuilder, dbuilder, newSummary);
}
else
{
throw new AssertionError("Attempted to clone SSTableReader with the same index summary sampling level and " +
"no adjustments to min/max_index_interval");
}
long newSize = bytesOnDisk();
StorageMetrics.load.inc(newSize - oldSize);
parent.metric.liveDiskSpaceUsed.inc(newSize - oldSize);
SSTableReader replacement = new SSTableReader(descriptor, components, metadata, partitioner, ifile.sharedCopy(),
dfile.sharedCopy(), newSummary, bf.sharedCopy(), maxDataAge,
sstableMetadata, OpenReason.METADATA_CHANGE);
replacement.first = this.first;
replacement.last = this.last;
setReplacedBy(replacement);
return replacement;
}
}
private IndexSummary buildSummaryAtLevel(int newSamplingLevel) throws IOException
{
// we read the positions in a BRAF so we don't have to worry about an entry spanning a mmap boundary.
RandomAccessReader primaryIndex = RandomAccessReader.open(new File(descriptor.filenameFor(Component.PRIMARY_INDEX)));
try
{
long indexSize = primaryIndex.length();
try (IndexSummaryBuilder summaryBuilder = new IndexSummaryBuilder(estimatedKeys(), metadata.getMinIndexInterval(), newSamplingLevel))
{
long indexPosition;
while ((indexPosition = primaryIndex.getFilePointer()) != indexSize)
{
summaryBuilder.maybeAddEntry(partitioner.decorateKey(ByteBufferUtil.readWithShortLength(primaryIndex)), indexPosition);
RowIndexEntry.Serializer.skip(primaryIndex);
}
return summaryBuilder.build(partitioner);
}
}
finally
{
FileUtils.closeQuietly(primaryIndex);
}
}
public RestorableMeter getReadMeter()
{
return readMeter;
}
public int getIndexSummarySamplingLevel()
{
return indexSummary.getSamplingLevel();
}
public long getIndexSummaryOffHeapSize()
{
return indexSummary.getOffHeapSize();
}
public int getMinIndexInterval()
{
return indexSummary.getMinIndexInterval();
}
public double getEffectiveIndexInterval()
{
return indexSummary.getEffectiveIndexInterval();
}
public void releaseSummary() throws IOException
{
tidy.releaseSummary();
indexSummary = null;
}
private void validate()
{
if (this.first.compareTo(this.last) > 0)
{
selfRef().release();
throw new IllegalStateException(String.format("SSTable first key %s > last key %s", this.first, this.last));
}
}
/**
* Gets the position in the index file to start scanning to find the given key (at most indexInterval keys away,
* modulo downsampling of the index summary). Always returns a value >= 0
*/
public long getIndexScanPosition(RowPosition key)
{
if (openReason == OpenReason.MOVED_START && key.compareTo(first) < 0)
key = first;
return getIndexScanPositionFromBinarySearchResult(indexSummary.binarySearch(key), indexSummary);
}
@VisibleForTesting
public static long getIndexScanPositionFromBinarySearchResult(int binarySearchResult, IndexSummary referencedIndexSummary)
{
if (binarySearchResult == -1)
return 0;
else
return referencedIndexSummary.getPosition(getIndexSummaryIndexFromBinarySearchResult(binarySearchResult));
}
public static int getIndexSummaryIndexFromBinarySearchResult(int binarySearchResult)
{
if (binarySearchResult < 0)
{
// binary search gives us the first index _greater_ than the key searched for,
// i.e., its insertion position
int greaterThan = (binarySearchResult + 1) * -1;
if (greaterThan == 0)
return -1;
return greaterThan - 1;
}
else
{
return binarySearchResult;
}
}
/**
* Returns the compression metadata for this sstable.
* @throws IllegalStateException if the sstable is not compressed
*/
public CompressionMetadata getCompressionMetadata()
{
if (!compression)
throw new IllegalStateException(this + " is not compressed");
CompressionMetadata cmd = ((ICompressedFile) dfile).getMetadata();
//We need the parent cf metadata
String cfName = metadata.isSecondaryIndex() ? metadata.getParentColumnFamilyName() : metadata.cfName;
cmd.parameters.setLiveMetadata(Schema.instance.getCFMetaData(metadata.ksName, cfName));
return cmd;
}
/**
* Returns the amount of memory in bytes used off heap by the compression meta-data.
* @return the amount of memory in bytes used off heap by the compression meta-data
*/
public long getCompressionMetadataOffHeapSize()
{
if (!compression)
return 0;
return getCompressionMetadata().offHeapSize();
}
/**
* For testing purposes only.
*/
public void forceFilterFailures()
{
bf = FilterFactory.AlwaysPresent;
}
public IFilter getBloomFilter()
{
return bf;
}
public long getBloomFilterSerializedSize()
{
return bf.serializedSize();
}
/**
* Returns the amount of memory in bytes used off heap by the bloom filter.
* @return the amount of memory in bytes used off heap by the bloom filter
*/
public long getBloomFilterOffHeapSize()
{
return bf.offHeapSize();
}
/**
* @return An estimate of the number of keys in this SSTable based on the index summary.
*/
public long estimatedKeys()
{
return indexSummary.getEstimatedKeyCount();
}
/**
* @param ranges
* @return An estimate of the number of keys for given ranges in this SSTable.
*/
public long estimatedKeysForRanges(Collection<Range<Token>> ranges)
{
long sampleKeyCount = 0;
List<Pair<Integer, Integer>> sampleIndexes = getSampleIndexesForRanges(indexSummary, ranges);
for (Pair<Integer, Integer> sampleIndexRange : sampleIndexes)
sampleKeyCount += (sampleIndexRange.right - sampleIndexRange.left + 1);
// adjust for the current sampling level: (BSL / SL) * index_interval_at_full_sampling
long estimatedKeys = sampleKeyCount * ((long) Downsampling.BASE_SAMPLING_LEVEL * indexSummary.getMinIndexInterval()) / indexSummary.getSamplingLevel();
return Math.max(1, estimatedKeys);
}
/**
* Returns the number of entries in the IndexSummary. At full sampling, this is approximately 1/INDEX_INTERVALth of
* the keys in this SSTable.
*/
public int getIndexSummarySize()
{
return indexSummary.size();
}
/**
* Returns the approximate number of entries the IndexSummary would contain if it were at full sampling.
*/
public int getMaxIndexSummarySize()
{
return indexSummary.getMaxNumberOfEntries();
}
/**
* Returns the key for the index summary entry at `index`.
*/
public byte[] getIndexSummaryKey(int index)
{
return indexSummary.getKey(index);
}
private static List<Pair<Integer,Integer>> getSampleIndexesForRanges(IndexSummary summary, Collection<Range<Token>> ranges)
{
// use the index to determine a minimal section for each range
List<Pair<Integer,Integer>> positions = new ArrayList<>();
for (Range<Token> range : Range.normalize(ranges))
{
RowPosition leftPosition = range.left.maxKeyBound();
RowPosition rightPosition = range.right.maxKeyBound();
int left = summary.binarySearch(leftPosition);
if (left < 0)
left = (left + 1) * -1;
else
// left range are start exclusive
left = left + 1;
if (left == summary.size())
// left is past the end of the sampling
continue;
int right = Range.isWrapAround(range.left, range.right)
? summary.size() - 1
: summary.binarySearch(rightPosition);
if (right < 0)
{
// range are end inclusive so we use the previous index from what binarySearch give us
// since that will be the last index we will return
right = (right + 1) * -1;
if (right == 0)
// Means the first key is already stricly greater that the right bound
continue;
right--;
}
if (left > right)
// empty range
continue;
positions.add(Pair.create(left, right));
}
return positions;
}
public Iterable<DecoratedKey> getKeySamples(final Range<Token> range)
{
final List<Pair<Integer, Integer>> indexRanges = getSampleIndexesForRanges(indexSummary, Collections.singletonList(range));
if (indexRanges.isEmpty())
return Collections.emptyList();
return new Iterable<DecoratedKey>()
{
public Iterator<DecoratedKey> iterator()
{
return new Iterator<DecoratedKey>()
{
private Iterator<Pair<Integer, Integer>> rangeIter = indexRanges.iterator();
private Pair<Integer, Integer> current;
private int idx;
public boolean hasNext()
{
if (current == null || idx > current.right)
{
if (rangeIter.hasNext())
{
current = rangeIter.next();
idx = current.left;
return true;
}
return false;
}
return true;
}
public DecoratedKey next()
{
byte[] bytes = indexSummary.getKey(idx++);
return partitioner.decorateKey(ByteBuffer.wrap(bytes));
}
public void remove()
{
throw new UnsupportedOperationException();
}
};
}
};
}
/**
* Determine the minimal set of sections that can be extracted from this SSTable to cover the given ranges.
* @return A sorted list of (offset,end) pairs that cover the given ranges in the datafile for this SSTable.
*/
public List<Pair<Long,Long>> getPositionsForRanges(Collection<Range<Token>> ranges)
{
// use the index to determine a minimal section for each range
List<Pair<Long,Long>> positions = new ArrayList<>();
for (Range<Token> range : Range.normalize(ranges))
{
assert !range.isWrapAround() || range.right.isMinimum();
// truncate the range so it at most covers the sstable
AbstractBounds<RowPosition> bounds = range.toRowBounds();
RowPosition leftBound = bounds.left.compareTo(first) > 0 ? bounds.left : first.getToken().minKeyBound();
RowPosition rightBound = bounds.right.isMinimum() ? last.getToken().maxKeyBound() : bounds.right;
if (leftBound.compareTo(last) > 0 || rightBound.compareTo(first) < 0)
continue;
long left = getPosition(leftBound, Operator.GT).position;
long right = (rightBound.compareTo(last) > 0)
? uncompressedLength()
: getPosition(rightBound, Operator.GT).position;
if (left == right)
// empty range
continue;
assert left < right : String.format("Range=%s openReason=%s first=%s last=%s left=%d right=%d", range, openReason, first, last, left, right);
positions.add(Pair.create(left, right));
}
return positions;
}
public void invalidateCacheKey(DecoratedKey key)
{
KeyCacheKey cacheKey = new KeyCacheKey(metadata.cfId, descriptor, key.getKey());
keyCache.remove(cacheKey);
}
public void cacheKey(DecoratedKey key, RowIndexEntry info)
{
CachingOptions caching = metadata.getCaching();
if (!caching.keyCache.isEnabled()
|| keyCache == null
|| keyCache.getCapacity() == 0)
{
return;
}
KeyCacheKey cacheKey = new KeyCacheKey(metadata.cfId, descriptor, key.getKey());
logger.trace("Adding cache entry for {} -> {}", cacheKey, info);
keyCache.put(cacheKey, info);
}
public RowIndexEntry getCachedPosition(DecoratedKey key, boolean updateStats)
{
return getCachedPosition(new KeyCacheKey(metadata.cfId, descriptor, key.getKey()), updateStats);
}
private RowIndexEntry getCachedPosition(KeyCacheKey unifiedKey, boolean updateStats)
{
if (keyCache != null && keyCache.getCapacity() > 0) {
if (updateStats)
{
RowIndexEntry cachedEntry = keyCache.get(unifiedKey);
keyCacheRequest.incrementAndGet();
if (cachedEntry != null)
{
keyCacheHit.incrementAndGet();
bloomFilterTracker.addTruePositive();
}
return cachedEntry;
}
else
{
return keyCache.getInternal(unifiedKey);
}
}
return null;
}
/**
* Get position updating key cache and stats.
* @see #getPosition(org.apache.cassandra.db.RowPosition, org.apache.cassandra.io.sstable.SSTableReader.Operator, boolean)
*/
public RowIndexEntry getPosition(RowPosition key, Operator op)
{
return getPosition(key, op, true);
}
/**
* @param key The key to apply as the rhs to the given Operator. A 'fake' key is allowed to
* allow key selection by token bounds but only if op != * EQ
* @param op The Operator defining matching keys: the nearest key to the target matching the operator wins.
* @param updateCacheAndStats true if updating stats and cache
* @return The index entry corresponding to the key, or null if the key is not present
*/
public RowIndexEntry getPosition(RowPosition key, Operator op, boolean updateCacheAndStats)
{
return getPosition(key, op, updateCacheAndStats, false);
}
private RowIndexEntry getPosition(RowPosition key, Operator op, boolean updateCacheAndStats, boolean permitMatchPastLast)
{
// first, check bloom filter
if (op == Operator.EQ)
{
assert key instanceof DecoratedKey; // EQ only make sense if the key is a valid row key
if (!bf.isPresent(((DecoratedKey)key).getKey()))
{
Tracing.trace("Bloom filter allows skipping sstable {}", descriptor.generation);
return null;
}
}
// next, the key cache (only make sense for valid row key)
if ((op == Operator.EQ || op == Operator.GE) && (key instanceof DecoratedKey))
{
DecoratedKey decoratedKey = (DecoratedKey)key;
KeyCacheKey cacheKey = new KeyCacheKey(metadata.cfId, descriptor, decoratedKey.getKey());
RowIndexEntry cachedPosition = getCachedPosition(cacheKey, updateCacheAndStats);
if (cachedPosition != null)
{
Tracing.trace("Key cache hit for sstable {}", descriptor.generation);
return cachedPosition;
}
}
// check the smallest and greatest keys in the sstable to see if it can't be present
boolean skip = false;
if (key.compareTo(first) < 0)
{
if (op == Operator.EQ)
skip = true;
else
key = first;
op = Operator.EQ;
}
else
{
int l = last.compareTo(key);
// l <= 0 => we may be looking past the end of the file; we then narrow our behaviour to:
// 1) skipping if strictly greater for GE and EQ;
// 2) skipping if equal and searching GT, and we aren't permitting matching past last
skip = l <= 0 && (l < 0 || (!permitMatchPastLast && op == Operator.GT));
}
if (skip)
{
if (op == Operator.EQ && updateCacheAndStats)
bloomFilterTracker.addFalsePositive();
Tracing.trace("Check against min and max keys allows skipping sstable {}", descriptor.generation);
return null;
}
int binarySearchResult = indexSummary.binarySearch(key);
long sampledPosition = getIndexScanPositionFromBinarySearchResult(binarySearchResult, indexSummary);
int sampledIndex = getIndexSummaryIndexFromBinarySearchResult(binarySearchResult);
int effectiveInterval = indexSummary.getEffectiveIndexIntervalAfterIndex(sampledIndex);
// scan the on-disk index, starting at the nearest sampled position.
// The check against IndexInterval is to be exit the loop in the EQ case when the key looked for is not present
// (bloom filter false positive). But note that for non-EQ cases, we might need to check the first key of the
// next index position because the searched key can be greater the last key of the index interval checked if it
// is lesser than the first key of next interval (and in that case we must return the position of the first key
// of the next interval).
int i = 0;
Iterator<FileDataInput> segments = ifile.iterator(sampledPosition);
while (segments.hasNext())
{
FileDataInput in = segments.next();
try
{
while (!in.isEOF())
{
i++;
ByteBuffer indexKey = ByteBufferUtil.readWithShortLength(in);
boolean opSatisfied; // did we find an appropriate position for the op requested
boolean exactMatch; // is the current position an exact match for the key, suitable for caching
// Compare raw keys if possible for performance, otherwise compare decorated keys.
if (op == Operator.EQ && i <= effectiveInterval)
{
opSatisfied = exactMatch = indexKey.equals(((DecoratedKey) key).getKey());
}
else
{
DecoratedKey indexDecoratedKey = partitioner.decorateKey(indexKey);
int comparison = indexDecoratedKey.compareTo(key);
int v = op.apply(comparison);
opSatisfied = (v == 0);
exactMatch = (comparison == 0);
if (v < 0)
{
Tracing.trace("Partition index lookup allows skipping sstable {}", descriptor.generation);
return null;
}
}
if (opSatisfied)
{
// read data position from index entry
RowIndexEntry indexEntry = metadata.comparator.rowIndexEntrySerializer().deserialize(in, descriptor.version);
if (exactMatch && updateCacheAndStats)
{
assert key instanceof DecoratedKey; // key can be == to the index key only if it's a true row key
DecoratedKey decoratedKey = (DecoratedKey)key;
if (logger.isTraceEnabled())
{
// expensive sanity check! see CASSANDRA-4687
FileDataInput fdi = dfile.getSegment(indexEntry.position);
DecoratedKey keyInDisk = partitioner.decorateKey(ByteBufferUtil.readWithShortLength(fdi));
if (!keyInDisk.equals(key))
throw new AssertionError(String.format("%s != %s in %s", keyInDisk, key, fdi.getPath()));
fdi.close();
}
// store exact match for the key
cacheKey(decoratedKey, indexEntry);
}
if (op == Operator.EQ && updateCacheAndStats)
bloomFilterTracker.addTruePositive();
Tracing.trace("Partition index with {} entries found for sstable {}", indexEntry.columnsIndex().size(), descriptor.generation);
return indexEntry;
}
RowIndexEntry.Serializer.skip(in);
}
}
catch (IOException e)
{
markSuspect();
throw new CorruptSSTableException(e, in.getPath());
}
finally
{
FileUtils.closeQuietly(in);
}
}
if (op == Operator.EQ && updateCacheAndStats)
bloomFilterTracker.addFalsePositive();
Tracing.trace("Partition index lookup complete (bloom filter false positive) for sstable {}", descriptor.generation);
return null;
}
/**
* Finds and returns the first key beyond a given token in this SSTable or null if no such key exists.
*/
public DecoratedKey firstKeyBeyond(RowPosition token)
{
if (token.compareTo(first) < 0)
return first;
long sampledPosition = getIndexScanPosition(token);
Iterator<FileDataInput> segments = ifile.iterator(sampledPosition);
while (segments.hasNext())
{
FileDataInput in = segments.next();
try
{
while (!in.isEOF())
{
ByteBuffer indexKey = ByteBufferUtil.readWithShortLength(in);
DecoratedKey indexDecoratedKey = partitioner.decorateKey(indexKey);
if (indexDecoratedKey.compareTo(token) > 0)
return indexDecoratedKey;
RowIndexEntry.Serializer.skip(in);
}
}
catch (IOException e)
{
markSuspect();
throw new CorruptSSTableException(e, in.getPath());
}
finally
{
FileUtils.closeQuietly(in);
}
}
return null;
}
/**
* @return The length in bytes of the data for this SSTable. For
* compressed files, this is not the same thing as the on disk size (see
* onDiskLength())
*/
public long uncompressedLength()
{
return dfile.length;
}
/**
* @return The length in bytes of the on disk size for this SSTable. For
* compressed files, this is not the same thing as the data length (see
* length())
*/
public long onDiskLength()
{
return dfile.onDiskLength;
}
/**
* Mark the sstable as obsolete, i.e., compacted into newer sstables.
*
* When calling this function, the caller must ensure that the SSTableReader is not referenced anywhere
* except for threads holding a reference.
*
* @return true if the this is the first time the file was marked obsolete. Calling this
* multiple times is usually buggy (see exceptions in DataTracker.unmarkCompacting and removeOldSSTablesSize).
*/
public boolean markObsolete()
{
if (logger.isDebugEnabled())
logger.debug("Marking {} compacted", getFilename());
synchronized (tidy.global)
{
assert !tidy.isReplaced;
}
return !tidy.global.isCompacted.getAndSet(true);
}
public boolean isMarkedCompacted()
{
return tidy.global.isCompacted.get();
}
public void markSuspect()
{
if (logger.isDebugEnabled())
logger.debug("Marking {} as a suspect for blacklisting.", getFilename());
isSuspect.getAndSet(true);
}
public boolean isMarkedSuspect()
{
return isSuspect.get();
}
/**
*
* @param dataRange filter to use when reading the columns
* @return A Scanner for seeking over the rows of the SSTable.
*/
public ISSTableScanner getScanner(DataRange dataRange)
{
return SSTableScanner.getScanner(this, dataRange, null);
}
/**
* I/O SSTableScanner
* @return A Scanner for seeking over the rows of the SSTable.
*/
public ISSTableScanner getScanner()
{
return getScanner((RateLimiter) null);
}
public ISSTableScanner getScanner(RateLimiter limiter)
{
return SSTableScanner.getScanner(this, DataRange.allData(partitioner), limiter);
}
/**
* Direct I/O SSTableScanner over a defined range of tokens.
*
* @param range the range of keys to cover
* @return A Scanner for seeking over the rows of the SSTable.
*/
public ISSTableScanner getScanner(Range<Token> range, RateLimiter limiter)
{
if (range == null)
return getScanner(limiter);
return getScanner(Collections.singletonList(range), limiter);
}
/**
* Direct I/O SSTableScanner over a defined collection of ranges of tokens.
*
* @param ranges the range of keys to cover
* @return A Scanner for seeking over the rows of the SSTable.
*/
public ISSTableScanner getScanner(Collection<Range<Token>> ranges, RateLimiter limiter)
{
return SSTableScanner.getScanner(this, ranges, limiter);
}
public FileDataInput getFileDataInput(long position)
{
return dfile.getSegment(position);
}
/**
* Tests if the sstable contains data newer than the given age param (in localhost currentMilli time).
* This works in conjunction with maxDataAge which is an upper bound on the create of data in this sstable.
* @param age The age to compare the maxDataAre of this sstable. Measured in millisec since epoc on this host
* @return True iff this sstable contains data that's newer than the given age parameter.
*/
public boolean newSince(long age)
{
return maxDataAge > age;
}
public void createLinks(String snapshotDirectoryPath)
{
for (Component component : components)
{
File sourceFile = new File(descriptor.filenameFor(component));
File targetLink = new File(snapshotDirectoryPath, sourceFile.getName());
FileUtils.createHardLink(sourceFile, targetLink);
}
}
public boolean isRepaired()
{
return sstableMetadata.repairedAt != ActiveRepairService.UNREPAIRED_SSTABLE;
}
public SSTableReader getCurrentReplacement()
{
return tidy.global.live;
}
/**
* TODO: Move someplace reusable
*/
public abstract static class Operator
{
public static final Operator EQ = new Equals();
public static final Operator GE = new GreaterThanOrEqualTo();
public static final Operator GT = new GreaterThan();
/**
* @param comparison The result of a call to compare/compareTo, with the desired field on the rhs.
* @return less than 0 if the operator cannot match forward, 0 if it matches, greater than 0 if it might match forward.
*/
public abstract int apply(int comparison);
final static class Equals extends Operator
{
public int apply(int comparison) { return -comparison; }
}
final static class GreaterThanOrEqualTo extends Operator
{
public int apply(int comparison) { return comparison >= 0 ? 0 : 1; }
}
final static class GreaterThan extends Operator
{
public int apply(int comparison) { return comparison > 0 ? 0 : 1; }
}
}
public long getBloomFilterFalsePositiveCount()
{
return bloomFilterTracker.getFalsePositiveCount();
}
public long getRecentBloomFilterFalsePositiveCount()
{
return bloomFilterTracker.getRecentFalsePositiveCount();
}
public long getBloomFilterTruePositiveCount()
{
return bloomFilterTracker.getTruePositiveCount();
}
public long getRecentBloomFilterTruePositiveCount()
{
return bloomFilterTracker.getRecentTruePositiveCount();
}
public InstrumentingCache<KeyCacheKey, RowIndexEntry> getKeyCache()
{
return keyCache;
}
public EstimatedHistogram getEstimatedRowSize()
{
return sstableMetadata.estimatedRowSize;
}
public EstimatedHistogram getEstimatedColumnCount()
{
return sstableMetadata.estimatedColumnCount;
}
public double getEstimatedDroppableTombstoneRatio(int gcBefore)
{
return sstableMetadata.getEstimatedDroppableTombstoneRatio(gcBefore);
}
public double getDroppableTombstonesBefore(int gcBefore)
{
return sstableMetadata.getDroppableTombstonesBefore(gcBefore);
}
public double getCompressionRatio()
{
return sstableMetadata.compressionRatio;
}
public ReplayPosition getReplayPosition()
{
return sstableMetadata.replayPosition;
}
public long getMinTimestamp()
{
return sstableMetadata.minTimestamp;
}
public long getMaxTimestamp()
{
return sstableMetadata.maxTimestamp;
}
public Set<Integer> getAncestors()
{
try
{
CompactionMetadata compactionMetadata = (CompactionMetadata) descriptor.getMetadataSerializer().deserialize(descriptor, MetadataType.COMPACTION);
return compactionMetadata.ancestors;
}
catch (IOException e)
{
SSTableReader.logOpenException(descriptor, e);
return Collections.emptySet();
}
}
public int getSSTableLevel()
{
return sstableMetadata.sstableLevel;
}
/**
* Reloads the sstable metadata from disk.
*
* Called after level is changed on sstable, for example if the sstable is dropped to L0
*
* Might be possible to remove in future versions
*
* @throws IOException
*/
public void reloadSSTableMetadata() throws IOException
{
this.sstableMetadata = (StatsMetadata) descriptor.getMetadataSerializer().deserialize(descriptor, MetadataType.STATS);
}
public StatsMetadata getSSTableMetadata()
{
return sstableMetadata;
}
public RandomAccessReader openDataReader(RateLimiter limiter)
{
assert limiter != null;
return dfile.createThrottledReader(limiter);
}
public RandomAccessReader openDataReader()
{
return dfile.createReader();
}
public RandomAccessReader openIndexReader()
{
return ifile.createReader();
}
/**
* @param component component to get timestamp.
* @return last modified time for given component. 0 if given component does not exist or IO error occurs.
*/
public long getCreationTimeFor(Component component)
{
return new File(descriptor.filenameFor(component)).lastModified();
}
/**
* @return Number of key cache hit
*/
public long getKeyCacheHit()
{
return keyCacheHit.get();
}
/**
* @return Number of key cache request
*/
public long getKeyCacheRequest()
{
return keyCacheRequest.get();
}
/**
* Increment the total row read count and read rate for this SSTable. This should not be incremented for range
* slice queries, row cache hits, or non-query reads, like compaction.
*/
public void incrementReadCount()
{
if (readMeter != null)
readMeter.mark();
}
public static class SizeComparator implements Comparator<SSTableReader>
{
public int compare(SSTableReader o1, SSTableReader o2)
{
return Longs.compare(o1.onDiskLength(), o2.onDiskLength());
}
}
public Ref<SSTableReader> tryRef()
{
return selfRef.tryRef();
}
public Ref<SSTableReader> selfRef()
{
return selfRef;
}
public Ref<SSTableReader> ref()
{
return selfRef.ref();
}
void setup(boolean isOffline)
{
tidy.setup(this, isOffline);
this.readMeter = tidy.global.readMeter;
}
@VisibleForTesting
public void overrideReadMeter(RestorableMeter readMeter)
{
this.readMeter = tidy.global.readMeter = readMeter;
}
/**
* One instance per SSTableReader we create. This references the type-shared tidy, which in turn references
* the globally shared tidy, i.e.
*
* InstanceTidier => DescriptorTypeTitdy => GlobalTidy
*
* We can create many InstanceTidiers (one for every time we reopen an sstable with MOVED_START for example), but there can only be
* two DescriptorTypeTidy (FINAL and TEMPLINK) and only one GlobalTidy for one single logical sstable.
*
* When the InstanceTidier cleansup, it releases its reference to its DescriptorTypeTidy; when all InstanceTidiers
* for that type have run, the DescriptorTypeTidy cleansup. DescriptorTypeTidy behaves in the same way towards GlobalTidy.
*
* For ease, we stash a direct reference to both our type-shared and global tidier
*/
private static final class InstanceTidier implements Tidy
{
private final Descriptor descriptor;
private final CFMetaData metadata;
private IFilter bf;
private IndexSummary summary;
private SegmentedFile dfile;
private SegmentedFile ifile;
private Runnable runOnClose;
private boolean isReplaced = false;
// a reference to our shared per-Descriptor.Type tidy instance, that
// we will release when we are ourselves released
private Ref<DescriptorTypeTidy> typeRef;
// a convenience stashing of the shared per-descriptor-type tidy instance itself
// and the per-logical-sstable globally shared state that it is linked to
private DescriptorTypeTidy type;
private GlobalTidy global;
private boolean setup;
void setup(SSTableReader reader, boolean isOffline)
{
this.setup = true;
this.bf = reader.bf;
this.summary = reader.indexSummary;
this.dfile = reader.dfile;
this.ifile = reader.ifile;
// get a new reference to the shared descriptor-type tidy
this.typeRef = DescriptorTypeTidy.get(reader);
this.type = typeRef.get();
this.global = type.globalRef.get();
if (!isOffline)
global.ensureReadMeter();
}
InstanceTidier(Descriptor descriptor, CFMetaData metadata)
{
this.descriptor = descriptor;
this.metadata = metadata;
}
public void tidy()
{
// don't try to cleanup if the sstablereader was never fully constructed
if (!setup)
return;
final ColumnFamilyStore cfs = Schema.instance.getColumnFamilyStoreInstance(metadata.cfId);
final OpOrder.Barrier barrier;
if (cfs != null)
{
barrier = cfs.readOrdering.newBarrier();
barrier.issue();
}
else
barrier = null;
ScheduledExecutors.nonPeriodicTasks.execute(new Runnable()
{
public void run()
{
if (barrier != null)
barrier.await();
bf.close();
if (summary != null)
summary.close();
if (runOnClose != null)
runOnClose.run();
dfile.close();
ifile.close();
typeRef.release();
}
});
}
public String name()
{
return descriptor.toString();
}
void releaseSummary()
{
summary.close();
assert summary.isCleanedUp();
summary = null;
}
}
/**
* One shared between all instances of a given Descriptor.Type.
* Performs only two things: the deletion of the sstables for the type,
* if necessary; and the shared reference to the globally shared state.
*
* All InstanceTidiers, on setup(), ask the static get() method for their shared state,
* and stash a reference to it to be released when they are. Once all such references are
* released, the shared tidy will be performed.
*/
static final class DescriptorTypeTidy implements Tidy
{
// keyed by REAL descriptor (TMPLINK/FINAL), mapping to the shared DescriptorTypeTidy for that descriptor
static final ConcurrentMap<Descriptor, Ref<DescriptorTypeTidy>> lookup = new ConcurrentHashMap<>();
private final Descriptor desc;
private final Ref<GlobalTidy> globalRef;
private final SSTableDeletingTask deletingTask;
DescriptorTypeTidy(Descriptor desc, SSTableReader sstable)
{
this.desc = desc;
this.deletingTask = new SSTableDeletingTask(desc, sstable);
// get a new reference to the shared global tidy
this.globalRef = GlobalTidy.get(sstable);
}
public void tidy()
{
lookup.remove(desc);
boolean isCompacted = globalRef.get().isCompacted.get();
globalRef.release();
switch (desc.type)
{
case FINAL:
if (isCompacted)
deletingTask.run();
break;
case TEMPLINK:
deletingTask.run();
break;
default:
throw new IllegalStateException();
}
}
public String name()
{
return desc.toString();
}
// get a new reference to the shared DescriptorTypeTidy for this sstable
public static Ref<DescriptorTypeTidy> get(SSTableReader sstable)
{
Descriptor desc = sstable.descriptor;
if (sstable.openReason == OpenReason.EARLY)
desc = desc.asType(Descriptor.Type.TEMPLINK);
Ref<DescriptorTypeTidy> refc = lookup.get(desc);
if (refc != null)
return refc.ref();
final DescriptorTypeTidy tidy = new DescriptorTypeTidy(desc, sstable);
refc = new Ref<>(tidy, tidy);
Ref<?> ex = lookup.putIfAbsent(desc, refc);
assert ex == null;
return refc;
}
}
/**
* One instance per logical sstable. This both tracks shared cleanup and some shared state related
* to the sstable's lifecycle. All DescriptorTypeTidy instances, on construction, obtain a reference to us
* via our static get(). There should only ever be at most two such references extant at any one time,
* since only TMPLINK and FINAL type descriptors should be open as readers. When all files of both
* kinds have been released, this shared tidy will be performed.
*/
static final class GlobalTidy implements Tidy
{
// keyed by FINAL descriptor, mapping to the shared GlobalTidy for that descriptor
static final ConcurrentMap<Descriptor, Ref<GlobalTidy>> lookup = new ConcurrentHashMap<>();
private final Descriptor desc;
// a single convenience property for getting the most recent version of an sstable, not related to tidying
private SSTableReader live;
// the readMeter that is shared between all instances of the sstable, and can be overridden in all of them
// at once also, for testing purposes
private RestorableMeter readMeter;
// the scheduled persistence of the readMeter, that we will cancel once all instances of this logical
// sstable have been released
private ScheduledFuture readMeterSyncFuture;
// shared state managing if the logical sstable has been compacted; this is used in cleanup both here
// and in the FINAL type tidier
private final AtomicBoolean isCompacted;
GlobalTidy(final SSTableReader reader)
{
this.desc = reader.descriptor;
this.isCompacted = new AtomicBoolean();
this.live = reader;
}
void ensureReadMeter()
{
if (readMeter != null)
return;
// Don't track read rates for tables in the system keyspace and don't bother trying to load or persist
// the read meter when in client mode.
if (Keyspace.SYSTEM_KS.equals(desc.ksname) || Config.isClientMode())
{
readMeter = null;
readMeterSyncFuture = null;
return;
}
readMeter = SystemKeyspace.getSSTableReadMeter(desc.ksname, desc.cfname, desc.generation);
// sync the average read rate to system.sstable_activity every five minutes, starting one minute from now
readMeterSyncFuture = syncExecutor.scheduleAtFixedRate(new Runnable()
{
public void run()
{
if (!isCompacted.get())
{
meterSyncThrottle.acquire();
SystemKeyspace.persistSSTableReadMeter(desc.ksname, desc.cfname, desc.generation, readMeter);
}
}
}, 1, 5, TimeUnit.MINUTES);
}
public void tidy()
{
lookup.remove(desc);
if (readMeterSyncFuture != null)
readMeterSyncFuture.cancel(true);
if (isCompacted.get())
SystemKeyspace.clearSSTableReadMeter(desc.ksname, desc.cfname, desc.generation);
// don't ideally want to dropPageCache for the file until all instances have been released
CLibrary.trySkipCache(desc.filenameFor(Component.DATA), 0, 0);
CLibrary.trySkipCache(desc.filenameFor(Component.PRIMARY_INDEX), 0, 0);
}
public String name()
{
return desc.toString();
}
// get a new reference to the shared GlobalTidy for this sstable
public static Ref<GlobalTidy> get(SSTableReader sstable)
{
Descriptor descriptor = sstable.descriptor;
Ref<GlobalTidy> refc = lookup.get(descriptor);
if (refc != null)
return refc.ref();
final GlobalTidy tidy = new GlobalTidy(sstable);
refc = new Ref<>(tidy, tidy);
Ref<?> ex = lookup.putIfAbsent(descriptor, refc);
assert ex == null;
return refc;
}
}
}