Resharding: Route requests to source shard while target is pre-handoff #131306
Conversation
…outeIndexingRequest Refresh branch
…outeIndexingRequest refresh
ankikuma
added
the
:Distributed Indexing/Distributed
elasticsearchmachine
added
the
serverless-linked
…outeIndexingRequest Refresh branch
ankikuma
added
Team:Distributed Indexing
|
Pinging @elastic/es-distributed-indexing (Team:Distributed Indexing) |
| * @return Updated shardId | ||
| */ | ||
| protected final int rerouteIfResharding(int shardId) { | ||
| if (indexReshardingMetadata != null && indexReshardingMetadata.getSplit().isTargetShard(shardId)) { |
There was a problem hiding this comment.
nit: should we check for indexReshardingMetadata.isSplit()? It's technically possible.
There was a problem hiding this comment.
possibly we should assert it is split, because I don't know right now what the correct behaviour here would be if it isn't. Good chance that we'll also need some logic here for any other resharding type.
| return (indexReshardingMetadata.getSplit().sourceShard(shardId)); | ||
| } | ||
| } | ||
| return (shardId); |
| * @return Updated shardId | ||
| */ | ||
| protected final int rerouteIfResharding(int shardId) { | ||
| if (indexReshardingMetadata != null && indexReshardingMetadata.getSplit().isTargetShard(shardId)) { |
There was a problem hiding this comment.
possibly we should assert it is split, because I don't know right now what the correct behaviour here would be if it isn't. Good chance that we'll also need some logic here for any other resharding type.
…outeIndexingRequest Refresh
…outeIndexingRequest Refresh branch
…outeIndexingRequest Refresh branch
…outeIndexingRequest Refresh
…outeIndexingRequest Refresh
| */ | ||
| protected final int rerouteIfResharding(int shardId) { | ||
| if (indexReshardingMetadata != null && indexReshardingMetadata.getSplit().isTargetShard(shardId)) { | ||
| assert indexReshardingMetadata.isSplit() : "Index resharding state is not a split"; |
There was a problem hiding this comment.
Shouldn't this be before the line with getSplit()?
This PR routes
DocWriteRequests(index/delete/update requests) that are meant for the target shards, to source shards, when the target shard is not yet ready.The changes here should also handle bulk requests, because
BulkOperation#groupRequestsByShardscallsint shardId = docWriteRequest.route(indexRouting);to obtain the shardId. Theroute()codepath has been modified in this PR to handle resharding scenario.This PR does not address the following cases: