org.apache.flink.runtime.state.heap.CopyOnWriteStateTableSnapshot Maven / Gradle / Ivy
/*
* 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.flink.runtime.state.heap;
import org.apache.flink.annotation.Internal;
import org.apache.flink.api.common.typeutils.TypeSerializer;
import org.apache.flink.core.memory.DataOutputView;
import org.apache.flink.runtime.state.KeyGroupRange;
import org.apache.flink.runtime.state.KeyGroupRangeAssignment;
import java.io.IOException;
/**
* This class represents the snapshot of a {@link CopyOnWriteStateTable} and has a role in operator state checkpointing. Besides
* holding the {@link CopyOnWriteStateTable}s internal entries at the time of the snapshot, this class is also responsible for
* preparing and writing the state in the process of checkpointing.
*
* IMPORTANT: Please notice that snapshot integrity of entries in this class rely on proper copy-on-write semantics
* through the {@link CopyOnWriteStateTable} that created the snapshot object, but all objects in this snapshot must be considered
* as READ-ONLY!. The reason is that the objects held by this class may or may not be deep copies of original objects
* that may still used in the {@link CopyOnWriteStateTable}. This depends for each entry on whether or not it was subject to
* copy-on-write operations by the {@link CopyOnWriteStateTable}. Phrased differently: the {@link CopyOnWriteStateTable} provides
* copy-on-write isolation for this snapshot, but this snapshot does not isolate modifications from the
* {@link CopyOnWriteStateTable}!
*
* @param type of key
* @param type of namespace
* @param type of state
*/
@Internal
public class CopyOnWriteStateTableSnapshot
extends AbstractStateTableSnapshot> {
/**
* Version of the {@link CopyOnWriteStateTable} when this snapshot was created. This can be used to release the snapshot.
*/
private final int snapshotVersion;
/**
* The number of entries in the {@link CopyOnWriteStateTable} at the time of creating this snapshot.
*/
private final int stateTableSize;
/**
* The state table entries, as by the time this snapshot was created. Objects in this array may or may not be deep
* copies of the current entries in the {@link CopyOnWriteStateTable} that created this snapshot. This depends for each entry
* on whether or not it was subject to copy-on-write operations by the {@link CopyOnWriteStateTable}.
*/
private final CopyOnWriteStateTable.StateTableEntry[] snapshotData;
/**
* Offsets for the individual key-groups. This is lazily created when the snapshot is grouped by key-group during
* the process of writing this snapshot to an output as part of checkpointing.
*/
private int[] keyGroupOffsets;
/**
* A local duplicate of the table's key serializer.
*/
private final TypeSerializer localKeySerializer;
/**
* A local duplicate of the table's namespace serializer.
*/
private final TypeSerializer localNamespaceSerializer;
/**
* A local duplicate of the table's state serializer.
*/
private final TypeSerializer localStateSerializer;
/**
* Creates a new {@link CopyOnWriteStateTableSnapshot}.
*
* @param owningStateTable the {@link CopyOnWriteStateTable} for which this object represents a snapshot.
*/
CopyOnWriteStateTableSnapshot(CopyOnWriteStateTable owningStateTable) {
super(owningStateTable);
this.snapshotData = owningStateTable.snapshotTableArrays();
this.snapshotVersion = owningStateTable.getStateTableVersion();
this.stateTableSize = owningStateTable.size();
// We create duplicates of the serializers for the async snapshot, because TypeSerializer
// might be stateful and shared with the event processing thread.
this.localKeySerializer = owningStateTable.keyContext.getKeySerializer().duplicate();
this.localNamespaceSerializer = owningStateTable.metaInfo.getNamespaceSerializer().duplicate();
this.localStateSerializer = owningStateTable.metaInfo.getStateSerializer().duplicate();
this.keyGroupOffsets = null;
}
/**
* Returns the internal version of the {@link CopyOnWriteStateTable} when this snapshot was created. This value must be used to
* tell the {@link CopyOnWriteStateTable} when to release this snapshot.
*/
int getSnapshotVersion() {
return snapshotVersion;
}
/**
* Partitions the snapshot data by key-group. The algorithm first builds a histogram for the distribution of keys
* into key-groups. Then, the histogram is accumulated to obtain the boundaries of each key-group in an array.
* Last, we use the accumulated counts as write position pointers for the key-group's bins when reordering the
* entries by key-group. This operation is lazily performed before the first writing of a key-group.
*
* As a possible future optimization, we could perform the repartitioning in-place, using a scheme similar to the
* cuckoo cycles in cuckoo hashing. This can trade some performance for a smaller memory footprint.
*/
@SuppressWarnings("unchecked")
private void partitionEntriesByKeyGroup() {
// We only have to perform this step once before the first key-group is written
if (null != keyGroupOffsets) {
return;
}
final KeyGroupRange keyGroupRange = owningStateTable.keyContext.getKeyGroupRange();
final int totalKeyGroups = owningStateTable.keyContext.getNumberOfKeyGroups();
final int baseKgIdx = keyGroupRange.getStartKeyGroup();
final int[] histogram = new int[keyGroupRange.getNumberOfKeyGroups() + 1];
CopyOnWriteStateTable.StateTableEntry[] unfold = new CopyOnWriteStateTable.StateTableEntry[stateTableSize];
// 1) In this step we i) 'unfold' the linked list of entries to a flat array and ii) build a histogram for key-groups
int unfoldIndex = 0;
for (CopyOnWriteStateTable.StateTableEntry entry : snapshotData) {
while (null != entry) {
int effectiveKgIdx =
KeyGroupRangeAssignment.computeKeyGroupForKeyHash(entry.key.hashCode(), totalKeyGroups) - baseKgIdx + 1;
++histogram[effectiveKgIdx];
unfold[unfoldIndex++] = entry;
entry = entry.next;
}
}
// 2) We accumulate the histogram bins to obtain key-group ranges in the final array
for (int i = 1; i < histogram.length; ++i) {
histogram[i] += histogram[i - 1];
}
// 3) We repartition the entries by key-group, using the histogram values as write indexes
for (CopyOnWriteStateTable.StateTableEntry t : unfold) {
int effectiveKgIdx =
KeyGroupRangeAssignment.computeKeyGroupForKeyHash(t.key.hashCode(), totalKeyGroups) - baseKgIdx;
snapshotData[histogram[effectiveKgIdx]++] = t;
}
// 4) As byproduct, we also created the key-group offsets
this.keyGroupOffsets = histogram;
}
@Override
public void release() {
owningStateTable.releaseSnapshot(this);
}
@Override
public void writeMappingsInKeyGroup(DataOutputView dov, int keyGroupId) throws IOException {
if (null == keyGroupOffsets) {
partitionEntriesByKeyGroup();
}
final CopyOnWriteStateTable.StateTableEntry[] groupedOut = snapshotData;
KeyGroupRange keyGroupRange = owningStateTable.keyContext.getKeyGroupRange();
int keyGroupOffsetIdx = keyGroupId - keyGroupRange.getStartKeyGroup() - 1;
int startOffset = keyGroupOffsetIdx < 0 ? 0 : keyGroupOffsets[keyGroupOffsetIdx];
int endOffset = keyGroupOffsets[keyGroupOffsetIdx + 1];
// write number of mappings in key-group
dov.writeInt(endOffset - startOffset);
// write mappings
for (int i = startOffset; i < endOffset; ++i) {
CopyOnWriteStateTable.StateTableEntry toWrite = groupedOut[i];
groupedOut[i] = null; // free asap for GC
localNamespaceSerializer.serialize(toWrite.namespace, dov);
localKeySerializer.serialize(toWrite.key, dov);
localStateSerializer.serialize(toWrite.state, dov);
}
}
/**
* Returns true iff the given state table is the owner of this snapshot object.
*/
boolean isOwner(CopyOnWriteStateTable stateTable) {
return stateTable == owningStateTable;
}
}