Safely deleting indices

[amitavmohanty01 (Amitav Mohanty)]

I wanted to clean up old indices so I deleted them. Now, I have started seeing the following exceptions.

[2018-12-31T07:22:02,260][WARN ][o.e.i.s.GlobalCheckpointSyncAction] [elkelastic01.host1] unexpected error during the primary phase for action [indices:admin/seq_no/global_checkpoint_sync], request [GlobalCheckpointSyncAction.Request{shardId=[log_type1-2018.12.23][20], timeout=1m, index='log_type1-2018.12.23', waitForActiveShards=1}]

org.elasticsearch.index.IndexNotFoundException: no such index

[2018-12-31T07:22:02,261][INFO ][o.e.i.s.GlobalCheckpointSyncAction] [elkelastic01.host1] [log_type1-2018.12.23][20] global checkpoint sync failed

Is the DELETE call not a distributed commit? Will these exceptions stop after some attempts ?

[DavidTurner]

@amitavmohanty01

Is the DELETE call not a distributed commit?

It is, but assuming this is the same cluster as before it has so many shards that cluster state publishing might be happening rather slowly.

@amitavmohanty01

Will these exceptions stop after some attempts?

They should do, yes.

[amitavmohanty01 (Amitav Mohanty)]

@DavidTurner

It is, but assuming this is the same cluster as before it has so many shards that cluster state publishing might be happening rather slowly.

Slow is fine. Will the HTTP call return before the distributed commit is completed?

I was thinking that if the HTTP call has completed then these logs should not be showing up.

[DavidTurner]

@amitavmohanty01

Will the HTTP call return before the distributed commit is completed?

Possibly, yes. If it returns 200 OK then the deletion has been committed and will eventually reach all nodes.