Skip to content

Master should wait on cluster state publication when failing a shard #15468

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Dec 16, 2015
Merged

Master should wait on cluster state publication when failing a shard #15468

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
73 changes: 51 additions & 22 deletions core/src/main/java/org/elasticsearch/cluster/action/shard/ShardStateAction.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@
import org.elasticsearch.cluster.ClusterStateTaskConfig;
import org.elasticsearch.cluster.ClusterStateTaskExecutor;
import org.elasticsearch.cluster.ClusterStateTaskListener;
import org.elasticsearch.cluster.NotMasterException;
import org.elasticsearch.cluster.metadata.IndexMetaData;
import org.elasticsearch.cluster.node.DiscoveryNode;
import org.elasticsearch.cluster.routing.RoutingService;
Expand Down Expand Up @@ -53,6 +54,7 @@
import java.io.IOException;
import java.util.ArrayList;
import java.util.List;
import java.util.Locale;

import static org.elasticsearch.cluster.routing.ShardRouting.readShardRoutingEntry;

Expand Down Expand Up @@ -113,7 +115,7 @@ public void handleResponse(TransportResponse.Empty response) {

@Override
public void handleException(TransportException exp) {
logger.warn("failed to send failed shard to {}", exp, masterNode);
logger.warn("unexpected failure while sending request to [{}] to fail shard [{}]", exp, masterNode, shardRoutingEntry);
listener.onShardFailedFailure(masterNode, exp);
}
});
Expand All @@ -122,22 +124,62 @@ public void handleException(TransportException exp) {
private class ShardFailedTransportHandler implements TransportRequestHandler<ShardRoutingEntry> {
@Override
public void messageReceived(ShardRoutingEntry request, TransportChannel channel) throws Exception {
handleShardFailureOnMaster(request);
channel.sendResponse(TransportResponse.Empty.INSTANCE);
handleShardFailureOnMaster(request, new ClusterStateTaskListener() {
@Override
public void onFailure(String source, Throwable t) {
logger.error("unexpected failure while failing shard [{}]", t, request.shardRouting);
try {
channel.sendResponse(t);
} catch (Throwable channelThrowable) {
logger.warn("failed to send failure [{}] while failing shard [{}]", channelThrowable, t, request.shardRouting);
}
}

@Override
public void onNoLongerMaster(String source) {
logger.error("no longer master while failing shard [{}]", request.shardRouting);
try {
channel.sendResponse(new NotMasterException(source));
} catch (Throwable channelThrowable) {
logger.warn("failed to send no longer master while failing shard [{}]", channelThrowable, request.shardRouting);
}
}

@Override
public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this keeps bugging me :) we should something on the executor as well....

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@bleskes What keeps bugging you?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We discussed this through another channel; the issue is the possible reroutes are being issued for every task rather than once per batch. Modifying this behavior will require a little extra machinery in the cluster state task execution framework. I opened #15482 to address.

try {
int numberOfUnassignedShards = newState.getRoutingNodes().unassigned().size();
if (oldState != newState && numberOfUnassignedShards > 0) {
String reason = String.format(Locale.ROOT, "[%d] unassigned shards after failing shard [%s]", numberOfUnassignedShards, request.shardRouting);
if (logger.isTraceEnabled()) {
logger.trace(reason + ", scheduling a reroute");
}
routingService.reroute(reason);
}
} finally {
try {
channel.sendResponse(TransportResponse.Empty.INSTANCE);
} catch (Throwable channelThrowable) {
logger.warn("failed to send response while failing shard [{}]", channelThrowable, request.shardRouting);
}
}
}
}
);
}
}

class ShardFailedClusterStateHandler implements ClusterStateTaskExecutor<ShardRoutingEntry>, ClusterStateTaskListener {
class ShardFailedClusterStateHandler implements ClusterStateTaskExecutor<ShardRoutingEntry> {
@Override
public BatchResult<ShardRoutingEntry> execute(ClusterState currentState, List<ShardRoutingEntry> tasks) throws Exception {
BatchResult.Builder<ShardRoutingEntry> batchResultBuilder = BatchResult.builder();
List<FailedRerouteAllocation.FailedShard> shardRoutingsToBeApplied = new ArrayList<>(tasks.size());
List<FailedRerouteAllocation.FailedShard> failedShards = new ArrayList<>(tasks.size());
for (ShardRoutingEntry task : tasks) {
shardRoutingsToBeApplied.add(new FailedRerouteAllocation.FailedShard(task.shardRouting, task.message, task.failure));
failedShards.add(new FailedRerouteAllocation.FailedShard(task.shardRouting, task.message, task.failure));
}
ClusterState maybeUpdatedState = currentState;
try {
RoutingAllocation.Result result = allocationService.applyFailedShards(currentState, shardRoutingsToBeApplied);
RoutingAllocation.Result result = allocationService.applyFailedShards(currentState, failedShards);
if (result.changed()) {
maybeUpdatedState = ClusterState.builder(currentState).routingResult(result).build();
}
Expand All @@ -147,31 +189,18 @@ public BatchResult<ShardRoutingEntry> execute(ClusterState currentState, List<Sh
}
return batchResultBuilder.build(maybeUpdatedState);
}

@Override
public void clusterStateProcessed(String source, ClusterState oldState, ClusterState newState) {
if (oldState != newState && newState.getRoutingNodes().unassigned().size() > 0) {
logger.trace("unassigned shards after shard failures. scheduling a reroute.");
routingService.reroute("unassigned shards after shard failures, scheduling a reroute");
}
}

@Override
public void onFailure(String source, Throwable t) {
logger.error("unexpected failure during [{}]", t, source);
}
}

private final ShardFailedClusterStateHandler shardFailedClusterStateHandler = new ShardFailedClusterStateHandler();

private void handleShardFailureOnMaster(final ShardRoutingEntry shardRoutingEntry) {
private void handleShardFailureOnMaster(final ShardRoutingEntry shardRoutingEntry, ClusterStateTaskListener listener) {
logger.warn("{} received shard failed for {}", shardRoutingEntry.failure, shardRoutingEntry.shardRouting.shardId(), shardRoutingEntry);
clusterService.submitStateUpdateTask(
"shard-failed (" + shardRoutingEntry.shardRouting + "), message [" + shardRoutingEntry.message + "]",
shardRoutingEntry,
ClusterStateTaskConfig.build(Priority.HIGH),
shardFailedClusterStateHandler,
shardFailedClusterStateHandler);
listener);
}

public void shardStarted(final ShardRouting shardRouting, String indexUUID, final String reason) {
Expand Down