private void retryRecovery(
final RecoveryTarget recoveryTarget,
TimeValue retryAfter,
final StartRecoveryRequest currentRequest) {
try {
onGoingRecoveries.resetRecovery(recoveryTarget.recoveryId(), recoveryTarget.shardId());
} catch (Exception e) {
onGoingRecoveries.failRecovery(
recoveryTarget.recoveryId(), new RecoveryFailedException(currentRequest, e), true);
}
threadPool.schedule(
retryAfter, ThreadPool.Names.GENERIC, new RecoveryRunner(recoveryTarget.recoveryId()));
}
private void doRecovery(final RecoveryTarget recoveryTarget) {
assert recoveryTarget.sourceNode() != null : "can't do a recovery without a source node";
logger.trace("collecting local files for {}", recoveryTarget);
Store.MetadataSnapshot metadataSnapshot = null;
try {
if (recoveryTarget.indexShard().indexSettings().isOnSharedFilesystem()) {
// we are not going to copy any files, so don't bother listing files, potentially running
// into concurrency issues with the primary changing files underneath us.
metadataSnapshot = Store.MetadataSnapshot.EMPTY;
} else {
metadataSnapshot = recoveryTarget.indexShard().snapshotStoreMetadata();
}
} catch (org.apache.lucene.index.IndexNotFoundException e) {
// happens on an empty folder. no need to log
metadataSnapshot = Store.MetadataSnapshot.EMPTY;
} catch (IOException e) {
logger.warn("error while listing local files, recover as if there are none", e);
metadataSnapshot = Store.MetadataSnapshot.EMPTY;
} catch (Exception e) {
// this will be logged as warning later on...
logger.trace("unexpected error while listing local files, failing recovery", e);
onGoingRecoveries.failRecovery(
recoveryTarget.recoveryId(),
new RecoveryFailedException(recoveryTarget.state(), "failed to list local files", e),
true);
return;
}
logger.trace("{} local file count: [{}]", recoveryTarget, metadataSnapshot.size());
final StartRecoveryRequest request =
new StartRecoveryRequest(
recoveryTarget.shardId(),
recoveryTarget.sourceNode(),
clusterService.localNode(),
metadataSnapshot,
recoveryTarget.state().getPrimary(),
recoveryTarget.recoveryId());
final AtomicReference<RecoveryResponse> responseHolder = new AtomicReference<>();
try {
logger.trace(
"[{}][{}] starting recovery from {}",
request.shardId().getIndex().getName(),
request.shardId().id(),
request.sourceNode());
recoveryTarget.indexShard().prepareForIndexRecovery();
recoveryTarget
.CancellableThreads()
.execute(
() ->
responseHolder.set(
transportService
.submitRequest(
request.sourceNode(),
PeerRecoverySourceService.Actions.START_RECOVERY,
request,
new FutureTransportResponseHandler<RecoveryResponse>() {
@Override
public RecoveryResponse newInstance() {
return new RecoveryResponse();
}
})
.txGet()));
final RecoveryResponse recoveryResponse = responseHolder.get();
assert responseHolder != null;
final TimeValue recoveryTime = new TimeValue(recoveryTarget.state().getTimer().time());
// do this through ongoing recoveries to remove it from the collection
onGoingRecoveries.markRecoveryAsDone(recoveryTarget.recoveryId());
if (logger.isTraceEnabled()) {
StringBuilder sb = new StringBuilder();
sb.append('[')
.append(request.shardId().getIndex().getName())
.append(']')
.append('[')
.append(request.shardId().id())
.append("] ");
sb.append("recovery completed from ")
.append(request.sourceNode())
.append(", took[")
.append(recoveryTime)
.append("]\n");
sb.append(" phase1: recovered_files [")
.append(recoveryResponse.phase1FileNames.size())
.append("]")
.append(" with " + "total_size of [")
.append(new ByteSizeValue(recoveryResponse.phase1TotalSize))
.append("]")
.append(", took [")
.append(timeValueMillis(recoveryResponse.phase1Time))
.append("], throttling_wait [")
.append(timeValueMillis(recoveryResponse.phase1ThrottlingWaitTime))
.append(']')
.append("\n");
sb.append(" : reusing_files [")
.append(recoveryResponse.phase1ExistingFileNames.size())
.append("] with " + "total_size of [")
.append(new ByteSizeValue(recoveryResponse.phase1ExistingTotalSize))
.append("]\n");
sb.append(" phase2: start took [")
.append(timeValueMillis(recoveryResponse.startTime))
.append("]\n");
sb.append(" : recovered [")
.append(recoveryResponse.phase2Operations)
.append("]")
.append(" transaction log " + "operations")
.append(", took [")
.append(timeValueMillis(recoveryResponse.phase2Time))
.append("]")
.append("\n");
logger.trace("{}", sb);
} else {
logger.debug(
"{} recovery done from [{}], took [{}]",
request.shardId(),
recoveryTarget.sourceNode(),
recoveryTime);
}
} catch (CancellableThreads.ExecutionCancelledException e) {
logger.trace("recovery cancelled", e);
} catch (Exception e) {
if (logger.isTraceEnabled()) {
logger.trace(
(Supplier<?>)
() ->
new ParameterizedMessage(
"[{}][{}] Got exception on recovery",
request.shardId().getIndex().getName(),
request.shardId().id()),
e);
}
Throwable cause = ExceptionsHelper.unwrapCause(e);
if (cause instanceof CancellableThreads.ExecutionCancelledException) {
// this can also come from the source wrapped in a RemoteTransportException
onGoingRecoveries.failRecovery(
recoveryTarget.recoveryId(),
new RecoveryFailedException(request, "source has canceled the" + " recovery", cause),
false);
return;
}
if (cause instanceof RecoveryEngineException) {
// unwrap an exception that was thrown as part of the recovery
cause = cause.getCause();
}
// do it twice, in case we have double transport exception
cause = ExceptionsHelper.unwrapCause(cause);
if (cause instanceof RecoveryEngineException) {
// unwrap an exception that was thrown as part of the recovery
cause = cause.getCause();
}
// here, we would add checks against exception that need to be retried (and not removeAndClean
// in this case)
if (cause instanceof IllegalIndexShardStateException
|| cause instanceof IndexNotFoundException
|| cause instanceof ShardNotFoundException) {
// if the target is not ready yet, retry
retryRecovery(
recoveryTarget,
"remote shard not ready",
recoverySettings.retryDelayStateSync(),
request);
return;
}
if (cause instanceof DelayRecoveryException) {
retryRecovery(recoveryTarget, cause, recoverySettings.retryDelayStateSync(), request);
return;
}
if (cause instanceof ConnectTransportException) {
logger.debug(
"delaying recovery of {} for [{}] due to networking error [{}]",
recoveryTarget.shardId(),
recoverySettings.retryDelayNetwork(),
cause.getMessage());
retryRecovery(
recoveryTarget, cause.getMessage(), recoverySettings.retryDelayNetwork(), request);
return;
}
if (cause instanceof AlreadyClosedException) {
onGoingRecoveries.failRecovery(
recoveryTarget.recoveryId(),
new RecoveryFailedException(request, "source shard is " + "closed", cause),
false);
return;
}
onGoingRecoveries.failRecovery(
recoveryTarget.recoveryId(), new RecoveryFailedException(request, e), true);
}
}