honeymoose
/
OpenSearch
mirror of https://github.com/honeymoose/OpenSearch.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Failure to recover a shard might cause loosing translog data (especially with no replicas), closes #869.

This commit is contained in:
kimchy 2011-04-20 04:52:47 +03:00
parent 3c233347b8
commit f5dbcb2fa4
2 changed files with 17 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
modules/elasticsearch/src/main/java/org/elasticsearch/index/gateway/local/LocalIndexShardGateway.java
View File

@ -144,15 +144,24 @@ public class LocalIndexShardGateway extends AbstractIndexShardComponent implemen
try {
InputStreamStreamInput si = new InputStreamStreamInput(new FileInputStream(recoveringTranslogFile));
while (true) {
int opSize = si.readInt();
Translog.Operation operation = TranslogStreams.readTranslogOperation(si);
Translog.Operation operation;
try {
int opSize = si.readInt();
operation = TranslogStreams.readTranslogOperation(si);
} catch (EOFException e) {
// ignore, not properly written the last op
break;
} catch (IOException e) {
// ignore, not properly written last op
break;
}
recoveryStatus.translog().addTranslogOperations(1);
indexShard.performRecoveryOperation(operation);
}
} catch (EOFException e) {
// ignore this exception, its fine
} catch (IOException e) {
// ignore this as well
} catch (Throwable e) {
// we failed to recovery, make sure to delete the translog file (and keep the recovering one)
indexShard.translog().close(true);
throw new IndexShardGatewayRecoveryException(shardId, "failed to recover shard", e);
}
indexShard.performRecoveryFinalization(true);

4
modules/elasticsearch/src/main/java/org/elasticsearch/indices/cluster/IndicesClusterStateService.java
View File

@ -539,7 +539,7 @@ public class IndicesClusterStateService extends AbstractLifecycleComponent<Indic
synchronized (mutex) {
if (indexService.hasShard(shardRouting.shardId().id())) {
try {
indexService.cleanShard(shardRouting.shardId().id(), "recovery failure [" + ExceptionsHelper.detailedMessage(failure) + "]");
indexService.removeShard(shardRouting.shardId().id(), "recovery failure [" + ExceptionsHelper.detailedMessage(failure) + "]");
} catch (IndexShardMissingException e) {
// the node got closed on us, ignore it
} catch (Exception e1) {
@ -576,7 +576,7 @@ public class IndicesClusterStateService extends AbstractLifecycleComponent<Indic
synchronized (mutex) {
if (indexService.hasShard(shardId.id())) {
try {
indexService.cleanShard(shardId.id(), "engine failure [" + ExceptionsHelper.detailedMessage(failure) + "]");
indexService.removeShard(shardId.id(), "engine failure [" + ExceptionsHelper.detailedMessage(failure) + "]");
} catch (IndexShardMissingException e) {
// the node got closed on us, ignore it
} catch (Exception e1) {