use of org.elasticsearch.common.util.concurrent.ReleasableLock in project elasticsearch by elastic.
the class InternalEngine method delete.
@Override
public DeleteResult delete(Delete delete) throws IOException {
DeleteResult result;
try (ReleasableLock ignored = readLock.acquire()) {
assert assertVersionType(delete);
ensureOpen();
// NOTE: we don't throttle this when merges fall behind because delete-by-id does not create new segments:
result = innerDelete(delete);
} catch (RuntimeException | IOException e) {
try {
maybeFailEngine("index", e);
} catch (Exception inner) {
e.addSuppressed(inner);
}
throw e;
}
maybePruneDeletedTombstones();
return result;
}
use of org.elasticsearch.common.util.concurrent.ReleasableLock in project crate by crate.
the class InternalEngine method rollTranslogGeneration.
@Override
public void rollTranslogGeneration() throws EngineException {
try (ReleasableLock ignored = readLock.acquire()) {
ensureOpen();
translog.rollGeneration();
translog.trimUnreferencedReaders();
} catch (AlreadyClosedException e) {
failOnTragicEvent(e);
throw e;
} catch (Exception e) {
try {
failEngine("translog trimming failed", e);
} catch (Exception inner) {
e.addSuppressed(inner);
}
throw new EngineException(shardId, "failed to roll translog", e);
}
}
use of org.elasticsearch.common.util.concurrent.ReleasableLock in project crate by crate.
the class InternalEngine method trimOperationsFromTranslog.
@Override
public void trimOperationsFromTranslog(long belowTerm, long aboveSeqNo) throws EngineException {
try (ReleasableLock lock = readLock.acquire()) {
ensureOpen();
translog.trimOperations(belowTerm, aboveSeqNo);
} catch (AlreadyClosedException e) {
failOnTragicEvent(e);
throw e;
} catch (Exception e) {
try {
failEngine("translog operations trimming failed", e);
} catch (Exception inner) {
e.addSuppressed(inner);
}
throw new EngineException(shardId, "failed to trim translog operations", e);
}
}
use of org.elasticsearch.common.util.concurrent.ReleasableLock in project crate by crate.
the class InternalEngine method flush.
@Override
public CommitId flush(boolean force, boolean waitIfOngoing) throws EngineException {
ensureOpen();
if (force && waitIfOngoing == false) {
assert false : "wait_if_ongoing must be true for a force flush: force=" + force + " wait_if_ongoing=" + waitIfOngoing;
throw new IllegalArgumentException("wait_if_ongoing must be true for a force flush: force=" + force + " wait_if_ongoing=" + waitIfOngoing);
}
final byte[] newCommitId;
try (ReleasableLock lock = readLock.acquire()) {
ensureOpen();
if (flushLock.tryLock() == false) {
// if we can't get the lock right away we block if needed otherwise barf
if (waitIfOngoing) {
logger.trace("waiting for in-flight flush to finish");
flushLock.lock();
logger.trace("acquired flush lock after blocking");
} else {
return new CommitId(lastCommittedSegmentInfos.getId());
}
} else {
logger.trace("acquired flush lock immediately");
}
try {
// Only flush if (1) Lucene has uncommitted docs, or (2) forced by caller, or (3) the
// newly created commit points to a different translog generation (can free translog),
// or (4) the local checkpoint information in the last commit is stale, which slows down future recoveries.
boolean hasUncommittedChanges = indexWriter.hasUncommittedChanges();
boolean shouldPeriodicallyFlush = shouldPeriodicallyFlush();
if (hasUncommittedChanges || force || shouldPeriodicallyFlush || getProcessedLocalCheckpoint() > Long.parseLong(lastCommittedSegmentInfos.userData.get(SequenceNumbers.LOCAL_CHECKPOINT_KEY))) {
ensureCanFlush();
try {
translog.rollGeneration();
logger.trace("starting commit for flush; commitTranslog=true");
commitIndexWriter(indexWriter, translog, null);
logger.trace("finished commit for flush");
// we need to refresh in order to clear older version values
refresh("version_table_flush", SearcherScope.INTERNAL, true);
translog.trimUnreferencedReaders();
} catch (AlreadyClosedException e) {
failOnTragicEvent(e);
throw e;
} catch (Exception e) {
throw new FlushFailedEngineException(shardId, e);
}
refreshLastCommittedSegmentInfos();
}
newCommitId = lastCommittedSegmentInfos.getId();
} catch (FlushFailedEngineException ex) {
maybeFailEngine("flush", ex);
throw ex;
} finally {
flushLock.unlock();
}
}
// (e.g., moves backwards) we will at least still sometimes prune deleted tombstones:
if (engineConfig.isEnableGcDeletes()) {
pruneDeletedTombstones();
}
return new CommitId(newCommitId);
}
use of org.elasticsearch.common.util.concurrent.ReleasableLock in project crate by crate.
the class InternalEngine method index.
@Override
public IndexResult index(Index index) throws IOException {
assert Objects.equals(index.uid().field(), IdFieldMapper.NAME) : index.uid().field();
final boolean doThrottle = index.origin().isRecovery() == false;
try (ReleasableLock releasableLock = readLock.acquire()) {
ensureOpen();
assert assertIncomingSequenceNumber(index.origin(), index.seqNo());
try (Releasable ignored = versionMap.acquireLock(index.uid().bytes());
Releasable indexThrottle = doThrottle ? () -> {
} : throttle.acquireThrottle()) {
lastWriteNanos = index.startTime();
/* A NOTE ABOUT APPEND ONLY OPTIMIZATIONS:
* if we have an autoGeneratedID that comes into the engine we can potentially optimize
* and just use addDocument instead of updateDocument and skip the entire version and index lookupVersion across the board.
* Yet, we have to deal with multiple document delivery, for this we use a property of the document that is added
* to detect if it has potentially been added before. We use the documents timestamp for this since it's something
* that:
* - doesn't change per document
* - is preserved in the transaction log
* - and is assigned before we start to index / replicate
* NOTE: it's not important for this timestamp to be consistent across nodes etc. it's just a number that is in the common
* case increasing and can be used in the failure case when we retry and resent documents to establish a happens before
* relationship. For instance:
* - doc A has autoGeneratedIdTimestamp = 10, isRetry = false
* - doc B has autoGeneratedIdTimestamp = 9, isRetry = false
*
* while both docs are in in flight, we disconnect on one node, reconnect and send doc A again
* - now doc A' has autoGeneratedIdTimestamp = 10, isRetry = true
*
* if A' arrives on the shard first we update maxUnsafeAutoIdTimestamp to 10 and use update document. All subsequent
* documents that arrive (A and B) will also use updateDocument since their timestamps are less than
* maxUnsafeAutoIdTimestamp. While this is not strictly needed for doc B it is just much simpler to implement since it
* will just de-optimize some doc in the worst case.
*
* if A arrives on the shard first we use addDocument since maxUnsafeAutoIdTimestamp is < 10. A` will then just be skipped
* or calls updateDocument.
*/
final IndexingStrategy plan = indexingStrategyForOperation(index);
final IndexResult indexResult;
if (plan.earlyResultOnPreFlightError.isPresent()) {
indexResult = plan.earlyResultOnPreFlightError.get();
assert indexResult.getResultType() == Result.Type.FAILURE : indexResult.getResultType();
} else {
// generate or register sequence number
if (index.origin() == Operation.Origin.PRIMARY) {
index = new Index(index.uid(), index.parsedDoc(), generateSeqNoForOperationOnPrimary(index), index.primaryTerm(), index.version(), index.versionType(), index.origin(), index.startTime(), index.getAutoGeneratedIdTimestamp(), index.isRetry(), index.getIfSeqNo(), index.getIfPrimaryTerm());
final boolean toAppend = plan.indexIntoLucene && plan.useLuceneUpdateDocument == false;
if (toAppend == false) {
advanceMaxSeqNoOfUpdatesOrDeletesOnPrimary(index.seqNo());
}
} else {
markSeqNoAsSeen(index.seqNo());
}
assert index.seqNo() >= 0 : "ops should have an assigned seq no.; origin: " + index.origin();
if (plan.indexIntoLucene || plan.addStaleOpToLucene) {
indexResult = indexIntoLucene(index, plan);
} else {
indexResult = new IndexResult(plan.versionForIndexing, index.primaryTerm(), index.seqNo(), plan.currentNotFoundOrDeleted);
}
}
if (index.origin().isFromTranslog() == false) {
final Translog.Location location;
if (indexResult.getResultType() == Result.Type.SUCCESS) {
location = translog.add(new Translog.Index(index, indexResult));
} else if (indexResult.getSeqNo() != SequenceNumbers.UNASSIGNED_SEQ_NO) {
// if we have document failure, record it as a no-op in the translog and Lucene with the generated seq_no
final NoOp noOp = new NoOp(indexResult.getSeqNo(), index.primaryTerm(), index.origin(), index.startTime(), indexResult.getFailure().toString());
location = innerNoOp(noOp).getTranslogLocation();
} else {
location = null;
}
indexResult.setTranslogLocation(location);
}
if (plan.indexIntoLucene && indexResult.getResultType() == Result.Type.SUCCESS) {
final Translog.Location translogLocation = trackTranslogLocation.get() ? indexResult.getTranslogLocation() : null;
versionMap.maybePutIndexUnderLock(index.uid().bytes(), new IndexVersionValue(translogLocation, plan.versionForIndexing, index.seqNo(), index.primaryTerm()));
}
localCheckpointTracker.markSeqNoAsProcessed(indexResult.getSeqNo());
if (indexResult.getTranslogLocation() == null) {
// the op is coming from the translog (and is hence persisted already) or it does not have a sequence number
assert index.origin().isFromTranslog() || indexResult.getSeqNo() == SequenceNumbers.UNASSIGNED_SEQ_NO : "version conflict: index operation not coming from translog should not have seqNo, but found [" + indexResult.getSeqNo() + "]";
localCheckpointTracker.markSeqNoAsPersisted(indexResult.getSeqNo());
}
indexResult.setTook(System.nanoTime() - index.startTime());
indexResult.freeze();
return indexResult;
}
} catch (RuntimeException | IOException e) {
try {
if (e instanceof AlreadyClosedException == false && treatDocumentFailureAsTragicError(index)) {
failEngine("index id[" + index.id() + "] origin[" + index.origin() + "] seq#[" + index.seqNo() + "]", e);
} else {
maybeFailEngine("index id[" + index.id() + "] origin[" + index.origin() + "] seq#[" + index.seqNo() + "]", e);
}
} catch (Exception inner) {
e.addSuppressed(inner);
}
throw e;
}
}
Aggregations