UNASSIGNED state after REPLICA_ADDED

@DavidTurner should I ask it (last comment) as a separate question or do you think that it is might be related to the original question?

Sorry I'm not sure I understand the question. Relocating shards takes time, especially if you're actively indexing into them during the relocation.

sorry, I'll try to clarify.
My flow is separated to tasks:

  1. Create index in dedicated nodes set (data-wrk nodes).
  2. Index data (bulks)
  3. Relocate shards to other nodes (awareness).

I am talking task 2 - during the bulks operations shards start moving from one node to another (within the same node sets) without any explicit command.
I guess that Elasticsearch try to balance shards within that node sets, and if so - it's just a waste of time IMO. The screenshots above show the reason for that move (it that's help). I would like to avoid that relocation during the indexing task/phase..

I see. To answer that I think we'd need to understand why ES is moving those shards. Is it balancing them out to 3-per-node? If so, why were they not 3-per-node to start with?

Because there are more indices being baked/indexed at the same time, all of them are located in this data-wrk set.
I guess that Elasticsearch is trying to balance shards because other indices might finish their indexing task and start their relocation to the data-srv node-set.
If that's true, is there a way to disable the balancing for a specific node.attr? (my data-wrk nodes)

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.