Search in sources :

Example 1 with ReplicationResponse

use of org.elasticsearch.action.support.replication.ReplicationResponse in project elasticsearch by elastic.

the class TransportShardFlushAction method shardOperationOnPrimary.

@Override
protected PrimaryResult shardOperationOnPrimary(ShardFlushRequest shardRequest, IndexShard primary) {
    primary.flush(shardRequest.getRequest());
    logger.trace("{} flush request executed on primary", primary.shardId());
    return new PrimaryResult(shardRequest, new ReplicationResponse());
}
Also used : ReplicationResponse(org.elasticsearch.action.support.replication.ReplicationResponse)

Example 2 with ReplicationResponse

use of org.elasticsearch.action.support.replication.ReplicationResponse in project elasticsearch by elastic.

the class TransportShardRefreshAction method shardOperationOnPrimary.

@Override
protected PrimaryResult shardOperationOnPrimary(BasicReplicationRequest shardRequest, IndexShard primary) {
    primary.refresh("api");
    logger.trace("{} refresh request executed on primary", primary.shardId());
    return new PrimaryResult(shardRequest, new ReplicationResponse());
}
Also used : ReplicationResponse(org.elasticsearch.action.support.replication.ReplicationResponse)

Example 3 with ReplicationResponse

use of org.elasticsearch.action.support.replication.ReplicationResponse in project crate by crate.

the class TransportShardFlushAction method shardOperationOnPrimary.

protected void shardOperationOnPrimary(ShardFlushRequest shardRequest, IndexShard primary, ActionListener<PrimaryResult<ShardFlushRequest, ReplicationResponse>> listener) {
    ActionListener.completeWith(listener, () -> {
        primary.flush(shardRequest.getRequest());
        logger.trace("{} flush request executed on primary", primary.shardId());
        return new PrimaryResult<>(shardRequest, new ReplicationResponse());
    });
}
Also used : ReplicationResponse(org.elasticsearch.action.support.replication.ReplicationResponse)

Example 4 with ReplicationResponse

use of org.elasticsearch.action.support.replication.ReplicationResponse in project crate by crate.

the class RecoverySourceHandler method recoverToTarget.

/**
 * performs the recovery from the local engine to the target
 */
public void recoverToTarget(ActionListener<RecoveryResponse> listener) {
    final Closeable releaseResources = () -> IOUtils.close(resources);
    final ActionListener<RecoveryResponse> wrappedListener = ActionListener.notifyOnce(listener);
    try {
        cancellableThreads.setOnCancel((reason, beforeCancelEx) -> {
            final RuntimeException e;
            if (shard.state() == IndexShardState.CLOSED) {
                // check if the shard got closed on us
                e = new IndexShardClosedException(shard.shardId(), "shard is closed and recovery was canceled reason [" + reason + "]");
            } else {
                e = new CancellableThreads.ExecutionCancelledException("recovery was canceled reason [" + reason + "]");
            }
            if (beforeCancelEx != null) {
                e.addSuppressed(beforeCancelEx);
            }
            IOUtils.closeWhileHandlingException(releaseResources, () -> wrappedListener.onFailure(e));
            throw e;
        });
        final Consumer<Exception> onFailure = e -> {
            assert Transports.assertNotTransportThread(RecoverySourceHandler.this + "[onFailure]");
            IOUtils.closeWhileHandlingException(releaseResources, () -> wrappedListener.onFailure(e));
        };
        final boolean softDeletesEnabled = shard.indexSettings().isSoftDeleteEnabled();
        final SetOnce<RetentionLease> retentionLeaseRef = new SetOnce<>();
        runUnderPrimaryPermit(() -> {
            final IndexShardRoutingTable routingTable = shard.getReplicationGroup().getRoutingTable();
            ShardRouting targetShardRouting = routingTable.getByAllocationId(request.targetAllocationId());
            if (targetShardRouting == null) {
                logger.debug("delaying recovery of {} as it is not listed as assigned to target node {}", request.shardId(), request.targetNode());
                throw new DelayRecoveryException("source node does not have the shard listed in its state as allocated on the node");
            }
            assert targetShardRouting.initializing() : "expected recovery target to be initializing but was " + targetShardRouting;
            retentionLeaseRef.set(shard.getRetentionLeases().get(ReplicationTracker.getPeerRecoveryRetentionLeaseId(targetShardRouting)));
        }, shardId + " validating recovery target [" + request.targetAllocationId() + "] registered ", shard, cancellableThreads, logger);
        final Engine.HistorySource historySource;
        if (softDeletesEnabled && (shard.useRetentionLeasesInPeerRecovery() || retentionLeaseRef.get() != null)) {
            historySource = Engine.HistorySource.INDEX;
        } else {
            historySource = Engine.HistorySource.TRANSLOG;
        }
        final Closeable retentionLock = shard.acquireHistoryRetentionLock(historySource);
        resources.add(retentionLock);
        final long startingSeqNo;
        final boolean isSequenceNumberBasedRecovery = request.startingSeqNo() != SequenceNumbers.UNASSIGNED_SEQ_NO && isTargetSameHistory() && shard.hasCompleteHistoryOperations("peer-recovery", historySource, request.startingSeqNo()) && (historySource == Engine.HistorySource.TRANSLOG || (retentionLeaseRef.get() != null && retentionLeaseRef.get().retainingSequenceNumber() <= request.startingSeqNo()));
        if (isSequenceNumberBasedRecovery && softDeletesEnabled && retentionLeaseRef.get() != null) {
            // all the history we need is retained by an existing retention lease, so we do not need a separate retention lock
            retentionLock.close();
            logger.trace("history is retained by {}", retentionLeaseRef.get());
        } else {
            // all the history we need is retained by the retention lock, obtained before calling shard.hasCompleteHistoryOperations()
            // and before acquiring the safe commit we'll be using, so we can be certain that all operations after the safe commit's
            // local checkpoint will be retained for the duration of this recovery.
            logger.trace("history is retained by retention lock");
        }
        final StepListener<SendFileResult> sendFileStep = new StepListener<>();
        final StepListener<TimeValue> prepareEngineStep = new StepListener<>();
        final StepListener<SendSnapshotResult> sendSnapshotStep = new StepListener<>();
        final StepListener<Void> finalizeStep = new StepListener<>();
        if (isSequenceNumberBasedRecovery) {
            logger.trace("performing sequence numbers based recovery. starting at [{}]", request.startingSeqNo());
            startingSeqNo = request.startingSeqNo();
            if (retentionLeaseRef.get() == null) {
                createRetentionLease(startingSeqNo, ActionListener.map(sendFileStep, ignored -> SendFileResult.EMPTY));
            } else {
                sendFileStep.onResponse(SendFileResult.EMPTY);
            }
        } else {
            final Engine.IndexCommitRef safeCommitRef;
            try {
                safeCommitRef = shard.acquireSafeIndexCommit();
                resources.add(safeCommitRef);
            } catch (final Exception e) {
                throw new RecoveryEngineException(shard.shardId(), 1, "snapshot failed", e);
            }
            // Try and copy enough operations to the recovering peer so that if it is promoted to primary then it has a chance of being
            // able to recover other replicas using operations-based recoveries. If we are not using retention leases then we
            // conservatively copy all available operations. If we are using retention leases then "enough operations" is just the
            // operations from the local checkpoint of the safe commit onwards, because when using soft deletes the safe commit retains
            // at least as much history as anything else. The safe commit will often contain all the history retained by the current set
            // of retention leases, but this is not guaranteed: an earlier peer recovery from a different primary might have created a
            // retention lease for some history that this primary already discarded, since we discard history when the global checkpoint
            // advances and not when creating a new safe commit. In any case this is a best-effort thing since future recoveries can
            // always fall back to file-based ones, and only really presents a problem if this primary fails before things have settled
            // down.
            startingSeqNo = softDeletesEnabled ? Long.parseLong(safeCommitRef.getIndexCommit().getUserData().get(SequenceNumbers.LOCAL_CHECKPOINT_KEY)) + 1L : 0;
            logger.trace("performing file-based recovery followed by history replay starting at [{}]", startingSeqNo);
            try {
                final int estimateNumOps = shard.estimateNumberOfHistoryOperations("peer-recovery", historySource, startingSeqNo);
                final Releasable releaseStore = acquireStore(shard.store());
                resources.add(releaseStore);
                sendFileStep.whenComplete(r -> IOUtils.close(safeCommitRef, releaseStore), e -> {
                    try {
                        IOUtils.close(safeCommitRef, releaseStore);
                    } catch (final IOException ex) {
                        logger.warn("releasing snapshot caused exception", ex);
                    }
                });
                final StepListener<ReplicationResponse> deleteRetentionLeaseStep = new StepListener<>();
                runUnderPrimaryPermit(() -> {
                    try {
                        // If the target previously had a copy of this shard then a file-based recovery might move its global
                        // checkpoint backwards. We must therefore remove any existing retention lease so that we can create a
                        // new one later on in the recovery.
                        shard.removePeerRecoveryRetentionLease(request.targetNode().getId(), new ThreadedActionListener<>(logger, shard.getThreadPool(), ThreadPool.Names.GENERIC, deleteRetentionLeaseStep, false));
                    } catch (RetentionLeaseNotFoundException e) {
                        logger.debug("no peer-recovery retention lease for " + request.targetAllocationId());
                        deleteRetentionLeaseStep.onResponse(null);
                    }
                }, shardId + " removing retention lease for [" + request.targetAllocationId() + "]", shard, cancellableThreads, logger);
                deleteRetentionLeaseStep.whenComplete(ignored -> {
                    assert Transports.assertNotTransportThread(RecoverySourceHandler.this + "[phase1]");
                    phase1(safeCommitRef.getIndexCommit(), startingSeqNo, () -> estimateNumOps, sendFileStep);
                }, onFailure);
            } catch (final Exception e) {
                throw new RecoveryEngineException(shard.shardId(), 1, "sendFileStep failed", e);
            }
        }
        assert startingSeqNo >= 0 : "startingSeqNo must be non negative. got: " + startingSeqNo;
        sendFileStep.whenComplete(r -> {
            assert Transports.assertNotTransportThread(RecoverySourceHandler.this + "[prepareTargetForTranslog]");
            // For a sequence based recovery, the target can keep its local translog
            prepareTargetForTranslog(shard.estimateNumberOfHistoryOperations("peer-recovery", historySource, startingSeqNo), prepareEngineStep);
        }, onFailure);
        prepareEngineStep.whenComplete(prepareEngineTime -> {
            assert Transports.assertNotTransportThread(RecoverySourceHandler.this + "[phase2]");
            /*
                 * add shard to replication group (shard will receive replication requests from this point on)
                 * now that engine is open. This means that any document indexed into the primary after
                 * this will be replicated to this replica as well make sure to do this before sampling
                 * the max sequence number in the next step, to ensure that we send all documents up to
                 * maxSeqNo in phase2.
                 */
            runUnderPrimaryPermit(() -> shard.initiateTracking(request.targetAllocationId()), shardId + " initiating tracking of " + request.targetAllocationId(), shard, cancellableThreads, logger);
            final long endingSeqNo = shard.seqNoStats().getMaxSeqNo();
            // CRATE_PATCH
            try {
                blobRecoveryHook();
            } catch (Exception e) {
                throw new RecoveryEngineException(shard.shardId(), 1, "blobRecoveryHook failed", e);
            }
            if (logger.isTraceEnabled()) {
                logger.trace("snapshot translog for recovery; current size is [{}]", shard.estimateNumberOfHistoryOperations("peer-recovery", historySource, startingSeqNo));
            }
            final Translog.Snapshot phase2Snapshot = shard.getHistoryOperations("peer-recovery", historySource, startingSeqNo);
            resources.add(phase2Snapshot);
            retentionLock.close();
            // we have to capture the max_seen_auto_id_timestamp and the max_seq_no_of_updates to make sure that these values
            // are at least as high as the corresponding values on the primary when any of these operations were executed on it.
            final long maxSeenAutoIdTimestamp = shard.getMaxSeenAutoIdTimestamp();
            final long maxSeqNoOfUpdatesOrDeletes = shard.getMaxSeqNoOfUpdatesOrDeletes();
            final RetentionLeases retentionLeases = shard.getRetentionLeases();
            final long mappingVersionOnPrimary = shard.indexSettings().getIndexMetadata().getMappingVersion();
            phase2(startingSeqNo, endingSeqNo, phase2Snapshot, maxSeenAutoIdTimestamp, maxSeqNoOfUpdatesOrDeletes, retentionLeases, mappingVersionOnPrimary, sendSnapshotStep);
        }, onFailure);
        // Recovery target can trim all operations >= startingSeqNo as we have sent all these operations in the phase 2
        final long trimAboveSeqNo = startingSeqNo - 1;
        sendSnapshotStep.whenComplete(r -> finalizeRecovery(r.targetLocalCheckpoint, trimAboveSeqNo, finalizeStep), onFailure);
        finalizeStep.whenComplete(r -> {
            // TODO: return the actual throttle time
            final long phase1ThrottlingWaitTime = 0L;
            final SendSnapshotResult sendSnapshotResult = sendSnapshotStep.result();
            final SendFileResult sendFileResult = sendFileStep.result();
            final RecoveryResponse response = new RecoveryResponse(sendFileResult.phase1FileNames, sendFileResult.phase1FileSizes, sendFileResult.phase1ExistingFileNames, sendFileResult.phase1ExistingFileSizes, sendFileResult.totalSize, sendFileResult.existingTotalSize, sendFileResult.took.millis(), phase1ThrottlingWaitTime, prepareEngineStep.result().millis(), sendSnapshotResult.sentOperations, sendSnapshotResult.tookTime.millis());
            try {
                wrappedListener.onResponse(response);
            } finally {
                IOUtils.close(resources);
            }
        }, onFailure);
    } catch (Exception e) {
        IOUtils.closeWhileHandlingException(releaseResources, () -> wrappedListener.onFailure(e));
    }
}
Also used : CancellableThreads(org.elasticsearch.common.util.CancellableThreads) Arrays(java.util.Arrays) IndexFormatTooNewException(org.apache.lucene.index.IndexFormatTooNewException) Releasables(org.elasticsearch.common.lease.Releasables) RecoveryEngineException(org.elasticsearch.index.engine.RecoveryEngineException) RetentionLeaseNotFoundException(org.elasticsearch.index.seqno.RetentionLeaseNotFoundException) CorruptIndexException(org.apache.lucene.index.CorruptIndexException) StoreFileMetadata(org.elasticsearch.index.store.StoreFileMetadata) Transports(org.elasticsearch.transport.Transports) AtomicInteger(java.util.concurrent.atomic.AtomicInteger) Locale(java.util.Locale) ThreadPool(org.elasticsearch.threadpool.ThreadPool) ActionRunnable(org.elasticsearch.action.ActionRunnable) IOContext(org.apache.lucene.store.IOContext) StepListener(org.elasticsearch.action.StepListener) Releasable(org.elasticsearch.common.lease.Releasable) ByteSizeValue(org.elasticsearch.common.unit.ByteSizeValue) PlainActionFuture(org.elasticsearch.action.support.PlainActionFuture) IndexShardRoutingTable(org.elasticsearch.cluster.routing.IndexShardRoutingTable) BytesReference(org.elasticsearch.common.bytes.BytesReference) Engine(org.elasticsearch.index.engine.Engine) RemoteTransportException(org.elasticsearch.transport.RemoteTransportException) List(java.util.List) Logger(org.apache.logging.log4j.Logger) Version(org.elasticsearch.Version) InputStreamIndexInput(org.elasticsearch.common.lucene.store.InputStreamIndexInput) TimeValue(io.crate.common.unit.TimeValue) ReplicationTracker(org.elasticsearch.index.seqno.ReplicationTracker) CopyOnWriteArrayList(java.util.concurrent.CopyOnWriteArrayList) IndexCommit(org.apache.lucene.index.IndexCommit) ShardRouting(org.elasticsearch.cluster.routing.ShardRouting) IndexShardClosedException(org.elasticsearch.index.shard.IndexShardClosedException) IndexShardRelocatedException(org.elasticsearch.index.shard.IndexShardRelocatedException) CompletableFuture(java.util.concurrent.CompletableFuture) Deque(java.util.Deque) ParameterizedMessage(org.apache.logging.log4j.message.ParameterizedMessage) ArrayList(java.util.ArrayList) BytesArray(org.elasticsearch.common.bytes.BytesArray) RetentionLease(org.elasticsearch.index.seqno.RetentionLease) RetentionLeases(org.elasticsearch.index.seqno.RetentionLeases) Store(org.elasticsearch.index.store.Store) StreamSupport(java.util.stream.StreamSupport) IntSupplier(java.util.function.IntSupplier) Loggers(org.elasticsearch.common.logging.Loggers) ArrayUtil(org.apache.lucene.util.ArrayUtil) FutureUtils(org.elasticsearch.common.util.concurrent.FutureUtils) SequenceNumbers(org.elasticsearch.index.seqno.SequenceNumbers) IndexShardState(org.elasticsearch.index.shard.IndexShardState) IndexInput(org.apache.lucene.store.IndexInput) SetOnce(org.apache.lucene.util.SetOnce) IOUtils(io.crate.common.io.IOUtils) IndexShard(org.elasticsearch.index.shard.IndexShard) IOException(java.io.IOException) StopWatch(org.elasticsearch.common.StopWatch) IndexFormatTooOldException(org.apache.lucene.index.IndexFormatTooOldException) ConcurrentLinkedDeque(java.util.concurrent.ConcurrentLinkedDeque) Consumer(java.util.function.Consumer) ExceptionsHelper(org.elasticsearch.ExceptionsHelper) AtomicLong(java.util.concurrent.atomic.AtomicLong) ReplicationResponse(org.elasticsearch.action.support.replication.ReplicationResponse) Closeable(java.io.Closeable) Translog(org.elasticsearch.index.translog.Translog) ThreadedActionListener(org.elasticsearch.action.support.ThreadedActionListener) Comparator(java.util.Comparator) Collections(java.util.Collections) RateLimiter(org.apache.lucene.store.RateLimiter) ActionListener(org.elasticsearch.action.ActionListener) IndexShardRoutingTable(org.elasticsearch.cluster.routing.IndexShardRoutingTable) Closeable(java.io.Closeable) ReplicationResponse(org.elasticsearch.action.support.replication.ReplicationResponse) Translog(org.elasticsearch.index.translog.Translog) RecoveryEngineException(org.elasticsearch.index.engine.RecoveryEngineException) Engine(org.elasticsearch.index.engine.Engine) TimeValue(io.crate.common.unit.TimeValue) CancellableThreads(org.elasticsearch.common.util.CancellableThreads) SetOnce(org.apache.lucene.util.SetOnce) IOException(java.io.IOException) IndexFormatTooNewException(org.apache.lucene.index.IndexFormatTooNewException) RecoveryEngineException(org.elasticsearch.index.engine.RecoveryEngineException) RetentionLeaseNotFoundException(org.elasticsearch.index.seqno.RetentionLeaseNotFoundException) CorruptIndexException(org.apache.lucene.index.CorruptIndexException) RemoteTransportException(org.elasticsearch.transport.RemoteTransportException) IndexShardClosedException(org.elasticsearch.index.shard.IndexShardClosedException) IndexShardRelocatedException(org.elasticsearch.index.shard.IndexShardRelocatedException) IOException(java.io.IOException) IndexFormatTooOldException(org.apache.lucene.index.IndexFormatTooOldException) RetentionLeases(org.elasticsearch.index.seqno.RetentionLeases) RetentionLeaseNotFoundException(org.elasticsearch.index.seqno.RetentionLeaseNotFoundException) IndexShardClosedException(org.elasticsearch.index.shard.IndexShardClosedException) RetentionLease(org.elasticsearch.index.seqno.RetentionLease) StepListener(org.elasticsearch.action.StepListener) Releasable(org.elasticsearch.common.lease.Releasable) ShardRouting(org.elasticsearch.cluster.routing.ShardRouting)

Example 5 with ReplicationResponse

use of org.elasticsearch.action.support.replication.ReplicationResponse in project crate by crate.

the class RetentionLeaseSyncAction method sync.

final void sync(ShardId shardId, String primaryAllocationId, long primaryTerm, RetentionLeases retentionLeases, ActionListener<ReplicationResponse> listener) {
    final Request request = new Request(shardId, retentionLeases);
    transportService.sendChildRequest(clusterService.localNode(), transportPrimaryAction, new ConcreteShardRequest<>(request, primaryAllocationId, primaryTerm), transportOptions, new TransportResponseHandler<ReplicationResponse>() {

        @Override
        public ReplicationResponse read(StreamInput in) throws IOException {
            return newResponseInstance(in);
        }

        @Override
        public String executor() {
            return ThreadPool.Names.SAME;
        }

        @Override
        public void handleResponse(ReplicationResponse response) {
            listener.onResponse(response);
        }

        @Override
        public void handleException(TransportException e) {
            if (ExceptionsHelper.unwrap(e, AlreadyClosedException.class, IndexShardClosedException.class) == null) {
                getLogger().warn(new ParameterizedMessage("{} retention lease sync failed", shardId), e);
            }
            listener.onFailure(e);
        }
    });
}
Also used : ReplicationRequest(org.elasticsearch.action.support.replication.ReplicationRequest) StreamInput(org.elasticsearch.common.io.stream.StreamInput) ParameterizedMessage(org.apache.logging.log4j.message.ParameterizedMessage) IOException(java.io.IOException) TransportException(org.elasticsearch.transport.TransportException) ReplicationResponse(org.elasticsearch.action.support.replication.ReplicationResponse)

Aggregations

ReplicationResponse (org.elasticsearch.action.support.replication.ReplicationResponse)24 Test (org.junit.Test)10 CountDownLatch (java.util.concurrent.CountDownLatch)9 IndexShard (org.elasticsearch.index.shard.IndexShard)9 ShardId (org.elasticsearch.index.shard.ShardId)8 ShardRouting (org.elasticsearch.cluster.routing.ShardRouting)7 Collections (java.util.Collections)6 List (java.util.List)6 AtomicBoolean (java.util.concurrent.atomic.AtomicBoolean)6 PlainActionFuture (org.elasticsearch.action.support.PlainActionFuture)6 IndexShardRoutingTable (org.elasticsearch.cluster.routing.IndexShardRoutingTable)6 IndicesService (org.elasticsearch.indices.IndicesService)6 IOException (java.io.IOException)5 ArrayList (java.util.ArrayList)5 Arrays (java.util.Arrays)5 BytesArray (org.elasticsearch.common.bytes.BytesArray)5 ByteSizeValue (org.elasticsearch.common.unit.ByteSizeValue)5 Engine (org.elasticsearch.index.engine.Engine)5 SequenceNumbers (org.elasticsearch.index.seqno.SequenceNumbers)5 Translog (org.elasticsearch.index.translog.Translog)5