You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by be...@apache.org on 2015/03/12 13:27:50 UTC
[1/3] cassandra git commit: pass correct (but unused in this context)
value to SSTableWriter.afterAppend, and rename parameter name
Repository: cassandra
Updated Branches:
refs/heads/cassandra-2.1 4831ba14a -> cbd4de8f5
refs/heads/trunk 24915cd48 -> e5d119aab
pass correct (but unused in this context) value to SSTableWriter.afterAppend, and rename parameter name
Project: http://git-wip-us.apache.org/repos/asf/cassandra/repo
Commit: http://git-wip-us.apache.org/repos/asf/cassandra/commit/cbd4de8f
Tree: http://git-wip-us.apache.org/repos/asf/cassandra/tree/cbd4de8f
Diff: http://git-wip-us.apache.org/repos/asf/cassandra/diff/cbd4de8f
Branch: refs/heads/cassandra-2.1
Commit: cbd4de8f51cf0da1579abcd8d93803c4e8845024
Parents: 4831ba1
Author: Benedict Elliott Smith <be...@apache.org>
Authored: Thu Mar 12 12:26:03 2015 +0000
Committer: Benedict Elliott Smith <be...@apache.org>
Committed: Thu Mar 12 12:26:03 2015 +0000
----------------------------------------------------------------------
.../apache/cassandra/io/sstable/SSTableWriter.java | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/cassandra/blob/cbd4de8f/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
----------------------------------------------------------------------
diff --git a/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java b/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
index b35b652..440961f 100644
--- a/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
+++ b/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
@@ -173,7 +173,7 @@ public class SSTableWriter extends SSTable
return (lastWrittenKey == null) ? 0 : dataFile.getFilePointer();
}
- private void afterAppend(DecoratedKey decoratedKey, long dataPosition, RowIndexEntry index)
+ private void afterAppend(DecoratedKey decoratedKey, long dataEnd, RowIndexEntry index)
{
sstableMetadataCollector.addKey(decoratedKey.getKey());
lastWrittenKey = decoratedKey;
@@ -182,9 +182,9 @@ public class SSTableWriter extends SSTable
first = lastWrittenKey;
if (logger.isTraceEnabled())
- logger.trace("wrote " + decoratedKey + " at " + dataPosition);
- iwriter.append(decoratedKey, index, dataPosition);
- dbuilder.addPotentialBoundary(dataPosition);
+ logger.trace("wrote " + decoratedKey + " at " + dataEnd);
+ iwriter.append(decoratedKey, index, dataEnd);
+ dbuilder.addPotentialBoundary(dataEnd);
}
/**
@@ -222,16 +222,18 @@ public class SSTableWriter extends SSTable
}
long startPosition = beforeAppend(decoratedKey);
+ long endPosition;
try
{
RowIndexEntry entry = rawAppend(cf, startPosition, decoratedKey, dataFile.stream);
- afterAppend(decoratedKey, startPosition, entry);
+ endPosition = dataFile.getFilePointer();
+ afterAppend(decoratedKey, endPosition, entry);
}
catch (IOException e)
{
throw new FSWriteError(e, dataFile.getPath());
}
- sstableMetadataCollector.update(dataFile.getFilePointer() - startPosition, cf.getColumnStats());
+ sstableMetadataCollector.update(endPosition - startPosition, cf.getColumnStats());
}
public static RowIndexEntry rawAppend(ColumnFamily cf, long startPosition, DecoratedKey key, DataOutputPlus out) throws IOException
[2/3] cassandra git commit: pass correct (but unused in this context)
value to SSTableWriter.afterAppend, and rename parameter name
Posted by be...@apache.org.
pass correct (but unused in this context) value to SSTableWriter.afterAppend, and rename parameter name
Project: http://git-wip-us.apache.org/repos/asf/cassandra/repo
Commit: http://git-wip-us.apache.org/repos/asf/cassandra/commit/cbd4de8f
Tree: http://git-wip-us.apache.org/repos/asf/cassandra/tree/cbd4de8f
Diff: http://git-wip-us.apache.org/repos/asf/cassandra/diff/cbd4de8f
Branch: refs/heads/trunk
Commit: cbd4de8f51cf0da1579abcd8d93803c4e8845024
Parents: 4831ba1
Author: Benedict Elliott Smith <be...@apache.org>
Authored: Thu Mar 12 12:26:03 2015 +0000
Committer: Benedict Elliott Smith <be...@apache.org>
Committed: Thu Mar 12 12:26:03 2015 +0000
----------------------------------------------------------------------
.../apache/cassandra/io/sstable/SSTableWriter.java | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/cassandra/blob/cbd4de8f/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
----------------------------------------------------------------------
diff --git a/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java b/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
index b35b652..440961f 100644
--- a/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
+++ b/src/java/org/apache/cassandra/io/sstable/SSTableWriter.java
@@ -173,7 +173,7 @@ public class SSTableWriter extends SSTable
return (lastWrittenKey == null) ? 0 : dataFile.getFilePointer();
}
- private void afterAppend(DecoratedKey decoratedKey, long dataPosition, RowIndexEntry index)
+ private void afterAppend(DecoratedKey decoratedKey, long dataEnd, RowIndexEntry index)
{
sstableMetadataCollector.addKey(decoratedKey.getKey());
lastWrittenKey = decoratedKey;
@@ -182,9 +182,9 @@ public class SSTableWriter extends SSTable
first = lastWrittenKey;
if (logger.isTraceEnabled())
- logger.trace("wrote " + decoratedKey + " at " + dataPosition);
- iwriter.append(decoratedKey, index, dataPosition);
- dbuilder.addPotentialBoundary(dataPosition);
+ logger.trace("wrote " + decoratedKey + " at " + dataEnd);
+ iwriter.append(decoratedKey, index, dataEnd);
+ dbuilder.addPotentialBoundary(dataEnd);
}
/**
@@ -222,16 +222,18 @@ public class SSTableWriter extends SSTable
}
long startPosition = beforeAppend(decoratedKey);
+ long endPosition;
try
{
RowIndexEntry entry = rawAppend(cf, startPosition, decoratedKey, dataFile.stream);
- afterAppend(decoratedKey, startPosition, entry);
+ endPosition = dataFile.getFilePointer();
+ afterAppend(decoratedKey, endPosition, entry);
}
catch (IOException e)
{
throw new FSWriteError(e, dataFile.getPath());
}
- sstableMetadataCollector.update(dataFile.getFilePointer() - startPosition, cf.getColumnStats());
+ sstableMetadataCollector.update(endPosition - startPosition, cf.getColumnStats());
}
public static RowIndexEntry rawAppend(ColumnFamily cf, long startPosition, DecoratedKey key, DataOutputPlus out) throws IOException
[3/3] cassandra git commit: Merge branch 'cassandra-2.1' into trunk
Posted by be...@apache.org.
Merge branch 'cassandra-2.1' into trunk
Conflicts:
src/java/org/apache/cassandra/io/sstable/format/big/BigTableWriter.java
Project: http://git-wip-us.apache.org/repos/asf/cassandra/repo
Commit: http://git-wip-us.apache.org/repos/asf/cassandra/commit/e5d119aa
Tree: http://git-wip-us.apache.org/repos/asf/cassandra/tree/e5d119aa
Diff: http://git-wip-us.apache.org/repos/asf/cassandra/diff/e5d119aa
Branch: refs/heads/trunk
Commit: e5d119aab49b1e19cd67f9d9dc0b0cc44bae90bb
Parents: 24915cd cbd4de8
Author: Benedict Elliott Smith <be...@apache.org>
Authored: Thu Mar 12 12:27:36 2015 +0000
Committer: Benedict Elliott Smith <be...@apache.org>
Committed: Thu Mar 12 12:27:36 2015 +0000
----------------------------------------------------------------------
.../io/sstable/format/big/BigTableWriter.java | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/cassandra/blob/e5d119aa/src/java/org/apache/cassandra/io/sstable/format/big/BigTableWriter.java
----------------------------------------------------------------------
diff --cc src/java/org/apache/cassandra/io/sstable/format/big/BigTableWriter.java
index a1b923d,0000000..ea2549d
mode 100644,000000..100644
--- a/src/java/org/apache/cassandra/io/sstable/format/big/BigTableWriter.java
+++ b/src/java/org/apache/cassandra/io/sstable/format/big/BigTableWriter.java
@@@ -1,585 -1,0 +1,587 @@@
+/*
+ * 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.format.big;
+
+import java.io.DataInput;
+import java.io.File;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.util.Collections;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Map;
+import java.util.Set;
+
+import org.apache.cassandra.db.*;
+import org.apache.cassandra.io.sstable.*;
+import org.apache.cassandra.io.sstable.format.SSTableReader;
+import org.apache.cassandra.io.sstable.format.SSTableWriter;
+import org.apache.cassandra.io.sstable.format.Version;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import org.apache.cassandra.config.CFMetaData;
+import org.apache.cassandra.config.DatabaseDescriptor;
+import org.apache.cassandra.db.compaction.AbstractCompactedRow;
+import org.apache.cassandra.dht.IPartitioner;
+import org.apache.cassandra.io.FSWriteError;
+import org.apache.cassandra.io.compress.CompressedSequentialWriter;
+import org.apache.cassandra.io.sstable.metadata.MetadataCollector;
+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.util.DataOutputPlus;
+import org.apache.cassandra.io.util.DataOutputStreamAndChannel;
+import org.apache.cassandra.io.util.FileMark;
+import org.apache.cassandra.io.util.FileUtils;
+import org.apache.cassandra.io.util.SegmentedFile;
+import org.apache.cassandra.io.util.SequentialWriter;
+import org.apache.cassandra.utils.ByteBufferUtil;
+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.StreamingHistogram;
+
+public class BigTableWriter extends SSTableWriter
+{
+ private static final Logger logger = LoggerFactory.getLogger(BigTableWriter.class);
+
+ // not very random, but the only value that can't be mistaken for a legal column-name length
+ public static final int END_OF_ROW = 0x0000;
+
+ private IndexWriter iwriter;
+ private SegmentedFile.Builder dbuilder;
+ private final SequentialWriter dataFile;
+ private DecoratedKey lastWrittenKey;
+ private FileMark dataMark;
+
+ BigTableWriter(Descriptor descriptor, Long keyCount, Long repairedAt, CFMetaData metadata, IPartitioner partitioner, MetadataCollector metadataCollector)
+ {
+ super(descriptor, keyCount, repairedAt, metadata, partitioner, metadataCollector);
+
+ if (compression)
+ {
+ dataFile = SequentialWriter.open(getFilename(),
+ descriptor.filenameFor(Component.COMPRESSION_INFO),
+ metadata.compressionParameters(),
+ metadataCollector);
+ dbuilder = SegmentedFile.getCompressedBuilder((CompressedSequentialWriter) dataFile);
+ }
+ else
+ {
+ dataFile = SequentialWriter.open(new File(getFilename()), new File(descriptor.filenameFor(Component.CRC)));
+ dbuilder = SegmentedFile.getBuilder(DatabaseDescriptor.getDiskAccessMode());
+ }
+ iwriter = new IndexWriter(keyCount, dataFile);
+ }
+
+ public void mark()
+ {
+ dataMark = dataFile.mark();
+ iwriter.mark();
+ }
+
+ public void resetAndTruncate()
+ {
+ dataFile.resetAndTruncate(dataMark);
+ iwriter.resetAndTruncate();
+ }
+
+ /**
+ * Perform sanity checks on @param decoratedKey and @return the position in the data file before any data is written
+ */
+ private long beforeAppend(DecoratedKey decoratedKey)
+ {
+ assert decoratedKey != null : "Keys must not be null"; // empty keys ARE allowed b/c of indexed column values
+ if (lastWrittenKey != null && lastWrittenKey.compareTo(decoratedKey) >= 0)
+ throw new RuntimeException("Last written key " + lastWrittenKey + " >= current key " + decoratedKey + " writing into " + getFilename());
+ return (lastWrittenKey == null) ? 0 : dataFile.getFilePointer();
+ }
+
- private void afterAppend(DecoratedKey decoratedKey, long dataPosition, RowIndexEntry index)
++ private void afterAppend(DecoratedKey decoratedKey, long dataEnd, RowIndexEntry index)
+ {
+ metadataCollector.addKey(decoratedKey.getKey());
+ lastWrittenKey = decoratedKey;
+ last = lastWrittenKey;
+ if (first == null)
+ first = lastWrittenKey;
+
+ if (logger.isTraceEnabled())
- logger.trace("wrote {} at {}", decoratedKey, dataPosition);
- iwriter.append(decoratedKey, index, dataPosition);
- dbuilder.addPotentialBoundary(dataPosition);
++ logger.trace("wrote " + decoratedKey + " at " + dataEnd);
++ iwriter.append(decoratedKey, index, dataEnd);
++ dbuilder.addPotentialBoundary(dataEnd);
+ }
+
+ /**
+ * @param row
+ * @return null if the row was compacted away entirely; otherwise, the PK index entry for this row
+ */
+ public RowIndexEntry append(AbstractCompactedRow row)
+ {
+ long startPosition = beforeAppend(row.key);
+ RowIndexEntry entry;
+ try
+ {
+ entry = row.write(startPosition, dataFile);
+ if (entry == null)
+ return null;
+ }
+ catch (IOException e)
+ {
+ throw new FSWriteError(e, dataFile.getPath());
+ }
+ long endPosition = dataFile.getFilePointer();
+ metadataCollector.update(endPosition - startPosition, row.columnStats());
+ afterAppend(row.key, endPosition, entry);
+ return entry;
+ }
+
+ public void append(DecoratedKey decoratedKey, ColumnFamily cf)
+ {
+ if (decoratedKey.getKey().remaining() > FBUtilities.MAX_UNSIGNED_SHORT)
+ {
+ logger.error("Key size {} exceeds maximum of {}, skipping row",
+ decoratedKey.getKey().remaining(),
+ FBUtilities.MAX_UNSIGNED_SHORT);
+ return;
+ }
+
+ long startPosition = beforeAppend(decoratedKey);
++ long endPosition;
+ try
+ {
+ RowIndexEntry entry = rawAppend(cf, startPosition, decoratedKey, dataFile.stream);
- afterAppend(decoratedKey, startPosition, entry);
++ endPosition = dataFile.getFilePointer();
++ afterAppend(decoratedKey, endPosition, entry);
+ }
+ catch (IOException e)
+ {
+ throw new FSWriteError(e, dataFile.getPath());
+ }
- metadataCollector.update(dataFile.getFilePointer() - startPosition, cf.getColumnStats());
++ metadataCollector.update(endPosition - startPosition, cf.getColumnStats());
+ }
+
+ private static RowIndexEntry rawAppend(ColumnFamily cf, long startPosition, DecoratedKey key, DataOutputPlus out) throws IOException
+ {
+ assert cf.hasColumns() || cf.isMarkedForDelete();
+
+ ColumnIndex.Builder builder = new ColumnIndex.Builder(cf, key.getKey(), out);
+ ColumnIndex index = builder.build(cf);
+
+ out.writeShort(END_OF_ROW);
+ return RowIndexEntry.create(startPosition, cf.deletionInfo().getTopLevelDeletion(), index);
+ }
+
+ /**
+ * @throws IOException if a read from the DataInput fails
+ * @throws FSWriteError if a write to the dataFile fails
+ */
+ public long appendFromStream(DecoratedKey key, CFMetaData metadata, DataInput in, Version version) throws IOException
+ {
+ long currentPosition = beforeAppend(key);
+
+ ColumnStats.MaxLongTracker maxTimestampTracker = new ColumnStats.MaxLongTracker(Long.MAX_VALUE);
+ ColumnStats.MinLongTracker minTimestampTracker = new ColumnStats.MinLongTracker(Long.MIN_VALUE);
+ ColumnStats.MaxIntTracker maxDeletionTimeTracker = new ColumnStats.MaxIntTracker(Integer.MAX_VALUE);
+ List<ByteBuffer> minColumnNames = Collections.emptyList();
+ List<ByteBuffer> maxColumnNames = Collections.emptyList();
+ StreamingHistogram tombstones = new StreamingHistogram(TOMBSTONE_HISTOGRAM_BIN_SIZE);
+ boolean hasLegacyCounterShards = false;
+
+ ColumnFamily cf = ArrayBackedSortedColumns.factory.create(metadata);
+ cf.delete(DeletionTime.serializer.deserialize(in));
+
+ ColumnIndex.Builder columnIndexer = new ColumnIndex.Builder(cf, key.getKey(), dataFile.stream);
+
+ if (cf.deletionInfo().getTopLevelDeletion().localDeletionTime < Integer.MAX_VALUE)
+ {
+ tombstones.update(cf.deletionInfo().getTopLevelDeletion().localDeletionTime);
+ maxDeletionTimeTracker.update(cf.deletionInfo().getTopLevelDeletion().localDeletionTime);
+ minTimestampTracker.update(cf.deletionInfo().getTopLevelDeletion().markedForDeleteAt);
+ maxTimestampTracker.update(cf.deletionInfo().getTopLevelDeletion().markedForDeleteAt);
+ }
+
+ Iterator<RangeTombstone> rangeTombstoneIterator = cf.deletionInfo().rangeIterator();
+ while (rangeTombstoneIterator.hasNext())
+ {
+ RangeTombstone rangeTombstone = rangeTombstoneIterator.next();
+ tombstones.update(rangeTombstone.getLocalDeletionTime());
+ minTimestampTracker.update(rangeTombstone.timestamp());
+ maxTimestampTracker.update(rangeTombstone.timestamp());
+ maxDeletionTimeTracker.update(rangeTombstone.getLocalDeletionTime());
+ minColumnNames = ColumnNameHelper.minComponents(minColumnNames, rangeTombstone.min, metadata.comparator);
+ maxColumnNames = ColumnNameHelper.maxComponents(maxColumnNames, rangeTombstone.max, metadata.comparator);
+ }
+
+ Iterator<OnDiskAtom> iter = AbstractCell.onDiskIterator(in, ColumnSerializer.Flag.PRESERVE_SIZE, Integer.MIN_VALUE, version, metadata.comparator);
+ try
+ {
+ while (iter.hasNext())
+ {
+ OnDiskAtom atom = iter.next();
+ if (atom == null)
+ break;
+
+ if (atom instanceof CounterCell)
+ {
+ atom = ((CounterCell) atom).markLocalToBeCleared();
+ hasLegacyCounterShards = hasLegacyCounterShards || ((CounterCell) atom).hasLegacyShards();
+ }
+
+ int deletionTime = atom.getLocalDeletionTime();
+ if (deletionTime < Integer.MAX_VALUE)
+ tombstones.update(deletionTime);
+ minTimestampTracker.update(atom.timestamp());
+ maxTimestampTracker.update(atom.timestamp());
+ minColumnNames = ColumnNameHelper.minComponents(minColumnNames, atom.name(), metadata.comparator);
+ maxColumnNames = ColumnNameHelper.maxComponents(maxColumnNames, atom.name(), metadata.comparator);
+ maxDeletionTimeTracker.update(atom.getLocalDeletionTime());
+
+ columnIndexer.add(atom); // This write the atom on disk too
+ }
+
+ columnIndexer.maybeWriteEmptyRowHeader();
+ dataFile.stream.writeShort(END_OF_ROW);
+ }
+ catch (IOException e)
+ {
+ throw new FSWriteError(e, dataFile.getPath());
+ }
+
+ metadataCollector.updateMinTimestamp(minTimestampTracker.get())
+ .updateMaxTimestamp(maxTimestampTracker.get())
+ .updateMaxLocalDeletionTime(maxDeletionTimeTracker.get())
+ .addRowSize(dataFile.getFilePointer() - currentPosition)
+ .addColumnCount(columnIndexer.writtenAtomCount())
+ .mergeTombstoneHistogram(tombstones)
+ .updateMinColumnNames(minColumnNames)
+ .updateMaxColumnNames(maxColumnNames)
+ .updateHasLegacyCounterShards(hasLegacyCounterShards);
+
+ afterAppend(key, currentPosition, RowIndexEntry.create(currentPosition, cf.deletionInfo().getTopLevelDeletion(), columnIndexer.build()));
+ return currentPosition;
+ }
+
+ /**
+ * After failure, attempt to close the index writer and data file before deleting all temp components for the sstable
+ */
+ public void abort()
+ {
+ assert descriptor.type.isTemporary;
+ if (iwriter == null && dataFile == null)
+ return;
+
+ if (iwriter != null)
+ iwriter.abort();
+
+ if (dataFile!= null)
+ dataFile.abort();
+
+ Set<Component> components = SSTable.componentsFor(descriptor);
+ try
+ {
+ if (!components.isEmpty())
+ SSTable.delete(descriptor, components);
+ }
+ catch (FSWriteError e)
+ {
+ logger.error(String.format("Failed deleting temp components for %s", descriptor), e);
+ throw e;
+ }
+ }
+
+ // we use this method to ensure any managed data we may have retained references to during the write are no
+ // longer referenced, so that we do not need to enclose the expensive call to closeAndOpenReader() in a transaction
+ public void isolateReferences()
+ {
+ // currently we only maintain references to first/last/lastWrittenKey from the data provided; all other
+ // data retention is done through copying
+ first = getMinimalKey(first);
+ last = lastWrittenKey = getMinimalKey(last);
+ }
+
+ private Descriptor makeTmpLinks()
+ {
+ // create temp links if they don't already exist
+ Descriptor link = descriptor.asType(Descriptor.Type.TEMPLINK);
+ if (!new File(link.filenameFor(Component.PRIMARY_INDEX)).exists())
+ {
+ FileUtils.createHardLink(new File(descriptor.filenameFor(Component.PRIMARY_INDEX)), new File(link.filenameFor(Component.PRIMARY_INDEX)));
+ FileUtils.createHardLink(new File(descriptor.filenameFor(Component.DATA)), new File(link.filenameFor(Component.DATA)));
+ }
+ return link;
+ }
+
+ public SSTableReader openEarly(long maxDataAge)
+ {
+ StatsMetadata sstableMetadata = (StatsMetadata) metadataCollector.finalizeMetadata(partitioner.getClass().getCanonicalName(),
+ metadata.getBloomFilterFpChance(),
+ repairedAt).get(MetadataType.STATS);
+
+ // find the max (exclusive) readable key
+ IndexSummaryBuilder.ReadableBoundary boundary = iwriter.getMaxReadable();
+ if (boundary == null)
+ return null;
+
+ assert boundary.indexLength > 0 && boundary.dataLength > 0;
+ Descriptor link = makeTmpLinks();
+ // open the reader early, giving it a FINAL descriptor type so that it is indistinguishable for other consumers
+ SegmentedFile ifile = iwriter.builder.complete(link.filenameFor(Component.PRIMARY_INDEX), boundary.indexLength);
+ SegmentedFile dfile = dbuilder.complete(link.filenameFor(Component.DATA), boundary.dataLength);
+ SSTableReader sstable = SSTableReader.internalOpen(descriptor.asType(Descriptor.Type.FINAL),
+ components, metadata,
+ partitioner, ifile,
+ dfile, iwriter.summary.build(partitioner, boundary),
+ iwriter.bf.sharedCopy(), maxDataAge, sstableMetadata, SSTableReader.OpenReason.EARLY);
+
+ // now it's open, find the ACTUAL last readable key (i.e. for which the data file has also been flushed)
+ sstable.first = getMinimalKey(first);
+ sstable.last = getMinimalKey(boundary.lastKey);
+ return sstable;
+ }
+
+ public SSTableReader closeAndOpenReader()
+ {
+ return closeAndOpenReader(System.currentTimeMillis());
+ }
+
+ public SSTableReader closeAndOpenReader(long maxDataAge)
+ {
+ return finish(FinishType.NORMAL, maxDataAge, this.repairedAt);
+ }
+
+ public SSTableReader finish(FinishType finishType, long maxDataAge, long repairedAt)
+ {
+ assert finishType != FinishType.CLOSE;
+ Pair<Descriptor, StatsMetadata> p;
+
+ p = close(finishType, repairedAt < 0 ? this.repairedAt : repairedAt);
+ Descriptor desc = p.left;
+ StatsMetadata metadata = p.right;
+
+ if (finishType == FinishType.EARLY)
+ desc = makeTmpLinks();
+
+ // finalize in-memory state for the reader
+ SegmentedFile ifile = iwriter.builder.complete(desc.filenameFor(Component.PRIMARY_INDEX), finishType.isFinal);
+ SegmentedFile dfile = dbuilder.complete(desc.filenameFor(Component.DATA), finishType.isFinal);
+ SSTableReader sstable = SSTableReader.internalOpen(desc.asType(Descriptor.Type.FINAL),
+ components,
+ this.metadata,
+ partitioner,
+ ifile,
+ dfile,
+ iwriter.summary.build(partitioner),
+ iwriter.bf.sharedCopy(),
+ maxDataAge,
+ metadata,
+ finishType.openReason);
+ sstable.first = getMinimalKey(first);
+ sstable.last = getMinimalKey(last);
+
+ if (finishType.isFinal)
+ {
+ iwriter.bf.close();
+ iwriter.summary.close();
+ // try to save the summaries to disk
+ sstable.saveSummary(iwriter.builder, dbuilder);
+ iwriter = null;
+ dbuilder = null;
+ }
+ return sstable;
+ }
+
+ // Close the writer and return the descriptor to the new sstable and it's metadata
+ public Pair<Descriptor, StatsMetadata> close()
+ {
+ return close(FinishType.CLOSE, this.repairedAt);
+ }
+
+ private Pair<Descriptor, StatsMetadata> close(FinishType type, long repairedAt)
+ {
+ switch (type)
+ {
+ case EARLY: case CLOSE: case NORMAL:
+ iwriter.close();
+ dataFile.close();
+ if (type == FinishType.CLOSE)
+ iwriter.bf.close();
+ }
+
+ // write sstable statistics
+ Map<MetadataType, MetadataComponent> metadataComponents;
+ metadataComponents = metadataCollector
+ .finalizeMetadata(partitioner.getClass().getCanonicalName(),
+ metadata.getBloomFilterFpChance(),repairedAt);
+
+ // remove the 'tmp' marker from all components
+ Descriptor descriptor = this.descriptor;
+ if (type.isFinal)
+ {
+ dataFile.writeFullChecksum(descriptor);
+ writeMetadata(descriptor, metadataComponents);
+ // save the table of components
+ SSTable.appendTOC(descriptor, components);
+ descriptor = rename(descriptor, components);
+ }
+
+ return Pair.create(descriptor, (StatsMetadata) metadataComponents.get(MetadataType.STATS));
+ }
+
+ private static void writeMetadata(Descriptor desc, Map<MetadataType, MetadataComponent> components)
+ {
+ SequentialWriter out = SequentialWriter.open(new File(desc.filenameFor(Component.STATS)));
+ try
+ {
+ desc.getMetadataSerializer().serialize(components, out.stream);
+ }
+ catch (IOException e)
+ {
+ throw new FSWriteError(e, out.getPath());
+ }
+ finally
+ {
+ out.close();
+ }
+ }
+
+ public long getFilePointer()
+ {
+ return dataFile.getFilePointer();
+ }
+
+ public long getOnDiskFilePointer()
+ {
+ return dataFile.getOnDiskFilePointer();
+ }
+
+ /**
+ * Encapsulates writing the index and filter for an SSTable. The state of this object is not valid until it has been closed.
+ */
+ class IndexWriter
+ {
+ private final SequentialWriter indexFile;
+ public final SegmentedFile.Builder builder;
+ public final IndexSummaryBuilder summary;
+ public final IFilter bf;
+ private FileMark mark;
+
+ IndexWriter(long keyCount, final SequentialWriter dataFile)
+ {
+ indexFile = SequentialWriter.open(new File(descriptor.filenameFor(Component.PRIMARY_INDEX)));
+ builder = SegmentedFile.getBuilder(DatabaseDescriptor.getIndexAccessMode());
+ summary = new IndexSummaryBuilder(keyCount, metadata.getMinIndexInterval(), Downsampling.BASE_SAMPLING_LEVEL);
+ bf = FilterFactory.getFilter(keyCount, metadata.getBloomFilterFpChance(), true);
+ // register listeners to be alerted when the data files are flushed
+ indexFile.setPostFlushListener(new Runnable()
+ {
+ public void run()
+ {
+ summary.markIndexSynced(indexFile.getLastFlushOffset());
+ }
+ });
+ dataFile.setPostFlushListener(new Runnable()
+ {
+ public void run()
+ {
+ summary.markDataSynced(dataFile.getLastFlushOffset());
+ }
+ });
+ }
+
+ // finds the last (-offset) decorated key that can be guaranteed to occur fully in the flushed portion of the index file
+ IndexSummaryBuilder.ReadableBoundary getMaxReadable()
+ {
+ return summary.getLastReadableBoundary();
+ }
+
+ public void append(DecoratedKey key, RowIndexEntry indexEntry, long dataEnd)
+ {
+ bf.add(key);
+ long indexStart = indexFile.getFilePointer();
+ try
+ {
+ ByteBufferUtil.writeWithShortLength(key.getKey(), indexFile.stream);
+ rowIndexEntrySerializer.serialize(indexEntry, indexFile.stream);
+ }
+ catch (IOException e)
+ {
+ throw new FSWriteError(e, indexFile.getPath());
+ }
+ long indexEnd = indexFile.getFilePointer();
+
+ if (logger.isTraceEnabled())
+ logger.trace("wrote index entry: {} at {}", indexEntry, indexStart);
+
+ summary.maybeAddEntry(key, indexStart, indexEnd, dataEnd);
+ builder.addPotentialBoundary(indexStart);
+ }
+
+ public void abort()
+ {
+ summary.close();
+ indexFile.abort();
+ bf.close();
+ }
+
+ /**
+ * Closes the index and bloomfilter, making the public state of this writer valid for consumption.
+ */
+ public void close()
+ {
+ if (components.contains(Component.FILTER))
+ {
+ String path = descriptor.filenameFor(Component.FILTER);
+ try
+ {
+ // bloom filter
+ FileOutputStream fos = new FileOutputStream(path);
+ DataOutputStreamAndChannel stream = new DataOutputStreamAndChannel(fos);
+ FilterFactory.serialize(bf, stream);
+ stream.flush();
+ fos.getFD().sync();
+ stream.close();
+ }
+ catch (IOException e)
+ {
+ throw new FSWriteError(e, path);
+ }
+ }
+
+ // index
+ long position = indexFile.getFilePointer();
+ indexFile.close(); // calls force
+ FileUtils.truncate(indexFile.getPath(), position);
+ }
+
+ public void mark()
+ {
+ mark = indexFile.mark();
+ }
+
+ public void resetAndTruncate()
+ {
+ // we can't un-set the bloom filter addition, but extra keys in there are harmless.
+ // we can't reset dbuilder either, but that is the last thing called in afterappend so
+ // we assume that if that worked then we won't be trying to reset.
+ indexFile.resetAndTruncate(mark);
+ }
+ }
+}