com.gemstone.gemfire.internal.cache.HDFSRegionMapDelegate Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of gemfire-core Show documentation
Show all versions of gemfire-core Show documentation
SnappyData store based off Pivotal GemFireXD
/*
* Copyright (c) 2010-2015 Pivotal Software, Inc. All rights reserved.
*
* Licensed 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. See accompanying
* LICENSE file.
*/
package com.gemstone.gemfire.internal.cache;
import java.io.IOException;
import java.lang.ref.Reference;
import java.lang.ref.ReferenceQueue;
import java.util.Collection;
import java.util.Collections;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;
import java.util.concurrent.ConcurrentMap;
import com.gemstone.gemfire.cache.EvictionAction;
import com.gemstone.gemfire.cache.CustomEvictionAttributes;
import com.gemstone.gemfire.cache.asyncqueue.internal.AsyncEventQueueImpl;
import com.gemstone.gemfire.cache.hdfs.HDFSIOException;
import com.gemstone.gemfire.cache.hdfs.internal.HDFSBucketRegionQueue;
import com.gemstone.gemfire.cache.hdfs.internal.HDFSEntriesSet;
import com.gemstone.gemfire.cache.hdfs.internal.HDFSEntriesSet.HDFSIterator;
import com.gemstone.gemfire.cache.hdfs.internal.HDFSGatewayEventImpl;
import com.gemstone.gemfire.cache.hdfs.internal.HDFSParallelGatewaySenderQueue;
import com.gemstone.gemfire.cache.hdfs.internal.PersistedEventImpl;
import com.gemstone.gemfire.cache.hdfs.internal.SortedHoplogPersistedEvent;
import com.gemstone.gemfire.distributed.internal.membership.InternalDistributedMember;
import com.gemstone.gemfire.i18n.LogWriterI18n;
import com.gemstone.gemfire.internal.cache.LocalRegion.IteratorType;
import com.gemstone.gemfire.internal.cache.RegionMap.Attributes;
import com.gemstone.gemfire.internal.cache.lru.EnableLRU;
import com.gemstone.gemfire.internal.cache.lru.LRUEntry;
import com.gemstone.gemfire.internal.cache.versions.RegionVersionVector;
import com.gemstone.gemfire.internal.cache.versions.VersionStamp;
import com.gemstone.gemfire.internal.cache.versions.VersionTag;
import com.gemstone.gemfire.internal.cache.wan.AbstractGatewaySenderEventProcessor;
import com.gemstone.gemfire.internal.cache.wan.parallel.ConcurrentParallelGatewaySenderQueue;
import com.gemstone.gemfire.internal.cache.wan.parallel.ParallelGatewaySenderImpl;
import com.gemstone.gemfire.internal.i18n.LocalizedStrings;
import com.gemstone.gemfire.internal.util.concurrent.FutureResult;
/**
* This class encapsulates all the functionality of HDFSRegionMap, so
* that it can be provided to HDFSLRURegionMap.
*
* @author sbawaska
*/
public class HDFSRegionMapDelegate {
private final LogWriterI18n logger;
private final BucketRegion owner;
private ConcurrentParallelGatewaySenderQueue hdfsQueue;
private final RegionMap backingRM;
/** queue of dead iterators */
private final ReferenceQueue refs;
private static final boolean DEBUG = Boolean.getBoolean("hdfsRegionMap.DEBUG");
/**
* used for serializing fetches from HDFS
*/
private ConcurrentMap futures = new ConcurrentHashMap();
public HDFSRegionMapDelegate(LocalRegion owner, Attributes attrs,
InternalRegionArguments internalRegionArgs, RegionMap backingRM) {
assert owner instanceof BucketRegion;
this.owner = (BucketRegion) owner;
this.logger = owner.getLogWriterI18n();
this.backingRM = backingRM;
refs = new ReferenceQueue();
}
public RegionEntry getEntry(Object key, EntryEventImpl event) {
RegionEntry re = getEntry(key, event, true);
// get from tx should put the entry back in map
// it should be evicted once tx completes
if (re != null && getTXState(event) != null) {
// put the region entry in backing CHM of AbstractRegionMap so that
// it can be locked in basicPut/destroy
RegionEntry oldRe = backingRM.putEntryIfAbsent(key, re);
if (oldRe != null) {
if (re instanceof OffHeapRegionEntry && !oldRe.equals(re)) {
((OffHeapRegionEntry)re).release();
}
return oldRe;
}
re.setMarkedForEviction();
owner.updateSizeOnCreate(key,
owner.calculateRegionEntryValueSize(re));
((AbstractRegionMap)backingRM).incEntryCount(1);
((AbstractRegionMap)backingRM).lruEntryCreate(re);
}
return re;
}
private TXStateInterface getTXState(EntryEventImpl event) {
return event != null ? event.getTXState(this.owner) : this.owner
.getTXState();
}
/**
*
* @param key
* @param event
* @param forceOnHeap if true will return heap version of off-heap region entries
*/
private RegionEntry getEntry(Object key, EntryEventImpl event, boolean forceOnHeap) {
closeDeadIterators();
RegionEntry re = backingRM.getEntryInVM(key);
if (logger.finerEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "HDFS: Found the key in CHM: " + key
+ " ,value=" + (re == null? "null" : "[" + re._getValue() + " or (" + re.getValueAsToken() + ")]"));
}
if ((re == null || (re.isRemoved() && !re.isTombstone()))
&& owner.getBucketAdvisor().isPrimary() && allowReadFromHDFS()) {
if (logger.finerEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "HDFS: fetching from hdfs key:" + key);
}
try {
this.owner.getPartitionedRegion().hdfsCalled(key);
re = getEntryFromFuture(key);
if (re != null) {
return re;
}
assert this.owner.getPartitionedRegion().getDataPolicy()
.withHDFS();
byte[] k = EntryEventImpl.serialize(key);
// for destroy ops we will retain the entry in the region map so
// tombstones can be tracked
//final boolean forceOnHeap = (event==null || !event.getOperation().isDestroy());
// get from queue
re = getFromHDFSQueue(key, k, forceOnHeap);
if (re == null) {
// get from HDFS
re = getFromHDFS(key, k, forceOnHeap);
}
if (re != null && re.isTombstone()) {
RegionVersionVector vector = this.owner.getVersionVector();
// if (vector == null) {
// this.owner.getLogWriterI18n().info(LocalizedStrings.DEBUG,
// "found a tombstone in a region w/o a version vector: " + re + "; region: " + this.owner);
// }
if (vector == null
|| vector.isTombstoneTooOld(re.getVersionStamp().getMemberID(),
re.getVersionStamp().getRegionVersion())) {
re = null;
}
}
if (logger.finerEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "HDFS: returning from hdfs re:" + re);
}
} catch (ForceReattemptException e) {
throw new PrimaryBucketException(e.getLocalizedMessage(), e);
} catch (IOException e) {
throw new HDFSIOException("Error reading from HDFS", e);
} finally {
notifyFuture(key, re);
// If we mark it here, the table scan may miss it causing updates/delete using table
// scan to fail.
// if (re != null) {
// re.setMarkedForEviction();
// }
if(re != null && event != null && !re.isTombstone()) {
if (logger.finerEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "HDFS: loaded from hdfs re:" + re);
}
BucketRegion br = (BucketRegion)owner;
//CustomEvictionAttributes csAttr = br.getCustomEvictionAttributes();
//if(csAttr!=null)
event.setLoadedFromHDFS(true);
}
}
}
if(re!=null && re.isMarkedForEviction() && !re.isTombstone()) {
if(event!=null) {
event.setLoadedFromHDFS(true);
}
}
return re;
}
/**
* This method returns true if the RegionEntry should be read from HDFS.
* fixes #49101 by not allowing reads from HDFS for persistent regions
* that do not define an eviction criteria.
*
* @return true if RegionEntry should be read from HDFS
*/
private boolean allowReadFromHDFS() {
if (!owner.getDataPolicy().withPersistence()
|| owner.getCustomEvictionAttributes() != null
|| isEvictionActionLocalDestroy()
|| owner.isUsedForIndex()) {
// when region does not have persistence, we have to read from HDFS (even
// though there is no eviction criteria) for constraint checks
return true;
}
return false;
}
private boolean isEvictionActionLocalDestroy() {
PartitionedRegion pr = owner.getPartitionedRegion();
if (pr.getEvictionAttributes() != null) {
return pr.getEvictionAttributes().getAction() == EvictionAction.LOCAL_DESTROY;
}
return false;
}
protected RegionEntry getEntry(EntryEventImpl event) {
RegionEntry re = getEntry(event.getKey(), event, false);
if (re != null && event.isLoadedFromHDFS()) {
// put the region entry in backing CHM of AbstractRegionMap so that
// it can be locked in basicPut/destroy
RegionEntry oldRe = backingRM.putEntryIfAbsent(event.getKey(), re);
if (oldRe != null) {
if (re instanceof OffHeapRegionEntry && !oldRe.equals(re)) {
((OffHeapRegionEntry) re).release();
}
return oldRe;
}
// since the entry is faulted in from HDFS, it must have
// satisfied the eviction criteria in the past, so mark it for eviction
re.setMarkedForEviction();
owner.updateSizeOnCreate(event.getKey(), owner.calculateRegionEntryValueSize(re));
((AbstractRegionMap) backingRM).incEntryCount(1);
((AbstractRegionMap) backingRM).lruEntryCreate(re);
}
return re;
}
@SuppressWarnings("unchecked")
public Collection regionEntries() {
closeDeadIterators();
if (!owner.getPartitionedRegion().includeHDFSResults()) {
if (logger.fineEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "Ignoring HDFS results for #regionEntries");
}
return backingRM.regionEntriesInVM();
}
try {
return createEntriesSet(IteratorType.ENTRIES);
} catch (ForceReattemptException e) {
throw new PrimaryBucketException(e.getLocalizedMessage(), e);
}
}
public int size() {
closeDeadIterators();
if (!owner.getPartitionedRegion().includeHDFSResults()) {
if (logger.fineEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "Ignoring HDFS results for #size");
}
return backingRM.sizeInVM();
}
try {
return createEntriesSet(IteratorType.KEYS).size();
} catch (ForceReattemptException e) {
throw new PrimaryBucketException(e.getLocalizedMessage(), e);
}
}
public boolean isEmpty() {
closeDeadIterators();
if (!owner.getPartitionedRegion().includeHDFSResults()) {
if (logger.fineEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "Ignoring HDFS results for #isEmpty");
}
return backingRM.sizeInVM() == 0;
}
try {
return createEntriesSet(IteratorType.KEYS).isEmpty();
} catch (ForceReattemptException e) {
throw new PrimaryBucketException(e.getLocalizedMessage(), e);
}
}
private void notifyFuture(Object key, RegionEntry re) {
FutureResult future = this.futures.remove(key);
if (future != null) {
future.set(re);
}
}
private RegionEntry getEntryFromFuture(Object key) {
FutureResult future = new FutureResult(this.owner.getCancelCriterion());
FutureResult old = this.futures.putIfAbsent(key, future);
if (old != null) {
if (logger.finerEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "HDFS: waiting for concurrent fetch to complete for key:" + key);
}
try {
return (RegionEntry) old.get();
} catch (InterruptedException e) {
Thread.currentThread().interrupt();
this.owner.getCache().getCancelCriterion().checkCancelInProgress(null);
}
}
return null;
}
private RegionEntry getFromHDFS(Object key, byte[] k, boolean forceOnHeap) throws IOException, ForceReattemptException {
SortedHoplogPersistedEvent ev;
try {
ev = (SortedHoplogPersistedEvent) owner.getHoplogOrganizer().read(k);
} catch (IOException e) {
owner.checkForPrimary();
throw e;
}
if (ev != null) {
if (logger.finerEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "HDFS: got from hdfs ev:" + ev);
}
return getEntryFromEvent(key, ev, forceOnHeap, false);
}
return null;
}
/**
* set the versionTag on the newly faulted-in entry
*/
private void setVersionTag(RegionEntry re, VersionTag versionTag) {
if (owner.concurrencyChecksEnabled) {
versionTag.setMemberID(
owner.getVersionVector().getCanonicalId(versionTag.getMemberID()));
VersionStamp versionedRe = (VersionStamp) re;
versionedRe.setVersions(versionTag);
}
}
private RegionEntry getFromHDFSQueue(Object key, byte[] k, boolean forceOnHeap) throws ForceReattemptException {
ConcurrentParallelGatewaySenderQueue q = getHDFSQueue();
if (q == null) return null;
HDFSGatewayEventImpl hdfsGatewayEvent = (HDFSGatewayEventImpl) q.get(owner.getPartitionedRegion(), k, owner.getId());
if (hdfsGatewayEvent != null) {
if (logger.finerEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "HDFS: got from hdfs queue: " + hdfsGatewayEvent);
}
return getEntryFromEvent(key, hdfsGatewayEvent, forceOnHeap, false);
}
return null;
}
private ConcurrentParallelGatewaySenderQueue getHDFSQueue()
throws ForceReattemptException {
if (this.hdfsQueue == null) {
String asyncQId = this.owner.getPartitionedRegion().getHDFSEventQueueName();
final AsyncEventQueueImpl asyncQ = (AsyncEventQueueImpl)this.owner.getCache().getAsyncEventQueue(asyncQId);
final ParallelGatewaySenderImpl gatewaySender = (ParallelGatewaySenderImpl)asyncQ.getSender();
AbstractGatewaySenderEventProcessor ep = gatewaySender.getEventProcessor();
if (ep == null) return null;
hdfsQueue = (ConcurrentParallelGatewaySenderQueue)ep.getQueue();
}
// Check whether the queue has become primary here.
// There could be some time between bucket becoming a primary
// and underlying queue becoming a primary, so isPrimaryWithWait()
// waits for some time for the queue to become a primary on this member
final HDFSBucketRegionQueue brq = hdfsQueue.getBucketRegionQueue(
this.owner.getPartitionedRegion(), this.owner.getId());
if (brq != null) {
if (owner.getBucketAdvisor().isPrimary()
&& !brq.getBucketAdvisor().isPrimaryWithWait()) {
InternalDistributedMember primaryHolder = brq.getBucketAdvisor()
.basicGetPrimaryMember();
throw new PrimaryBucketException("Bucket " + brq.getName()
+ " is not primary. Current primary holder is " + primaryHolder);
}
}
return hdfsQueue;
}
public RegionEntry getEntryFromEvent(Object key, HDFSGatewayEventImpl event, boolean forceOnHeap, boolean forUpdate) {
Object val;
if (event.getOperation().isDestroy()) {
val = Token.TOMBSTONE;
} else if (event.getOperation().isInvalidate()) {
val = Token.INVALID;
} else {
val = event.getValue();
}
RegionEntry re = null;
final TXStateInterface tx = owner.getTXState();
if (tx == null) {
re = createRegionEntry(key, val, event.getVersionTag(), forceOnHeap);
return re;
}
else
if (val != null) {
if (((re = this.backingRM.getEntryInVM(key)) == null)
|| (re.isRemoved() && !re.isTombstone())) {
boolean shouldCreateOnHeapEntry = !(owner.getEnableOffHeapMemory() && forUpdate);
re = createRegionEntry(key, val, event.getVersionTag(), shouldCreateOnHeapEntry);
if (forUpdate) {
if (re != null && tx != null) {
// put the region entry in backing CHM of AbstractRegionMap so that
// it can be locked in basicPut/destroy
RegionEntry oldRe = backingRM.putEntryIfAbsent(key, re);
if (oldRe != null) {
if (re instanceof OffHeapRegionEntry && !oldRe.equals(re)) {
((OffHeapRegionEntry)re).release();
}
return oldRe;
}
re.setMarkedForEviction();
owner.updateSizeOnCreate(key,
owner.calculateRegionEntryValueSize(re));
((AbstractRegionMap)backingRM).incEntryCount(1);
((AbstractRegionMap)backingRM).lruEntryCreate(re);
}
}
}
}
return re;
}
public RegionEntry getEntryFromEvent(Object key, SortedHoplogPersistedEvent event, boolean forceOnHeap, boolean forUpdate) {
Object val = getValueFromEvent(event);
RegionEntry re = null;
final TXStateInterface tx = owner.getTXState();
if (tx == null) {
re = createRegionEntry(key, val, event.getVersionTag(), forceOnHeap);
return re;
}
else // FOR TX case, we need to create off heap entry if required
if (val != null) {
if (((re = this.backingRM.getEntryInVM(key)) == null)
|| (re.isRemoved() && !re.isTombstone())) {
boolean shouldCreateOnHeapEntry = !(owner.getEnableOffHeapMemory() && forUpdate);
re = createRegionEntry(key, val, event.getVersionTag(), shouldCreateOnHeapEntry);
if(forUpdate) {
if (re != null && tx != null) {
// put the region entry in backing CHM of AbstractRegionMap so that
// it can be locked in basicPut/destroy
RegionEntry oldRe = backingRM.putEntryIfAbsent(key, re);
if (oldRe != null) {
if (re instanceof OffHeapRegionEntry && !oldRe.equals(re)) {
((OffHeapRegionEntry)re).release();
}
return oldRe;
}
re.setMarkedForEviction();
owner.updateSizeOnCreate(key,
owner.calculateRegionEntryValueSize(re));
((AbstractRegionMap)backingRM).incEntryCount(1);
((AbstractRegionMap)backingRM).lruEntryCreate(re);
}
}
}
}
return re;
}
private RegionEntry createRegionEntry(Object key, Object value, VersionTag tag, boolean forceOnHeap) {
RegionEntryFactory ref = backingRM.getEntryFactory();
if (forceOnHeap) {
ref = ref.makeOnHeap();
}
value = getValueDuringGII(key, value);
RegionEntry re = ref.createEntry(this.owner, key, value);
setVersionTag(re, tag);
if (re instanceof LRUEntry) {
assert backingRM instanceof AbstractLRURegionMap;
EnableLRU ccHelper = ((AbstractLRURegionMap)backingRM)._getCCHelper();
((LRUEntry)re).updateEntrySize(ccHelper);
}
return re;
}
private Object getValueDuringGII(Object key, Object value) {
if (owner.getIndexUpdater() != null && !owner.isInitialized()) {
return AbstractRegionMap.listOfDeltasCreator.newValue(key, owner, value,
null);
}
return value;
}
private Set createEntriesSet(IteratorType type)
throws ForceReattemptException {
ConcurrentParallelGatewaySenderQueue q = getHDFSQueue();
if (q == null) return Collections.emptySet();
HDFSBucketRegionQueue brq = q.getBucketRegionQueue(this.owner.getPartitionedRegion(), owner.getId());
return new HDFSEntriesSet(owner, brq, owner.getHoplogOrganizer(), type, refs);
}
private void closeDeadIterators() {
Reference weak;
while ((weak = refs.poll()) != null) {
if (logger.fineEnabled() || DEBUG) {
logger.info(LocalizedStrings.DEBUG, "Closing weak ref for iterator "
+ weak.get());
}
weak.get().close();
}
}
/**
* gets the value from event, deserializing if necessary.
*/
private Object getValueFromEvent(PersistedEventImpl ev) {
if (ev.getOperation().isDestroy()) {
return Token.TOMBSTONE;
} else if (ev.getOperation().isInvalidate()) {
return Token.INVALID;
}
return ev.getValue();
}
}