Module org.elasticsearch.server

Package org.elasticsearch.action.support.replication

Class BroadcastRequestSplitHelper

java.lang.Object

org.elasticsearch.action.support.replication.BroadcastRequestSplitHelper

public final class BroadcastRequestSplitHelper
extends Object

This class implements the logic to "split", rather forward, Broadcast requests like refresh and flush to the target nodes after a Resharding operation. It also implements the logic to combine responses from source and target shards. The logic to split indexing requests is different because each document id has to be inspected to route it to the correct shard post Resharding. That logic is implemented in ShardBulkSplitHelper

Method Summary

Modifier and Type	Method	Description
static <T extends ReplicationRequest<T>> Tuple<ReplicationResponse,Exception>	combineSplitResponses(T originalRequest, Map<ShardId,T> splitRequests, Map<ShardId,Tuple<ReplicationResponse,Exception>> responses)
static <T extends ReplicationRequest<T>> Map<ShardId,T>	splitRequest(T request, ProjectMetadata project, BiFunction<ShardId,SplitShardCountSummary,T> targetRequestFactory)	Given a stale Replication Request, like flush or refresh, split it into multiple requests, one for the source shard and one for the target shard.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

splitRequest

public static <T extends ReplicationRequest<T>>
Map<ShardId,T> splitRequest(T request,
 ProjectMetadata project,
 BiFunction<ShardId,SplitShardCountSummary,T> targetRequestFactory)

Given a stale Replication Request, like flush or refresh, split it into multiple requests, one for the source shard and one for the target shard. See ShardBulkSplitHelper for how we split BulkShardRequest We are here because there was a mismatch between the SplitShardCountSummary in the request and that on the primary shard node. Note that IndexReshardingMetadata cannot be NULL here because it is evaluated in ReplicationSplitHelper.needsSplitCoordination(Logger, ReplicationRequest, IndexMetadata) with the same metadata before arriving here. The request would have been rejected there if there was no Resharding metadata found. TODO: We assume here that the request is exactly 1 reshard split behind the current state. We might either revise this assumption or enforce it in a follow up

combineSplitResponses

public static <T extends ReplicationRequest<T>>
Tuple<ReplicationResponse,Exception> combineSplitResponses(T originalRequest,
 Map<ShardId,T> splitRequests,
 Map<ShardId,Tuple<ReplicationResponse,Exception>> responses)