Tribe node to ignore cluster events

Elastic Stack Elasticsearch
1

Hi,

I have tribe nodes that look at multiple clusters, but every time there's a index rotation or any sort of maintenance, the tribe node goes down until things are allocated and back to normal which could take over an hour.

I heard there was a patch that was given to blizzard for the tribe node to ignore cluster events. I was wondering if I could get a similar patch or tell me how to patch it?

Thanks!
Krystle

2

Completely ignoring cluster "events" is inherently dangerous yet there are aspects of cluster state updates that are expensive yet can be ignored on tribe nodes.

Upgrade to a version that includes Skip shard management code when updating cluster state on client/tribe nodes by ywelsch · Pull Request #20731 · elastic/elasticsearch · GitHub (first released in 2.4.2, and also included in 5.0.0).

3

I'm running version 2.4.4 . wouldn't that have the change if 2.4.2 has it?

4

What else could I do to for more stability for the tribe nodes?

5

Yes.

6

I don't know because I don't have enough detail to assess what your problem is.

7

basically when the indices rotate (create new ones and delete old ones), the tribe node goes in a red state saying it can't connect to elasticsearch. This happens for an hour sometimes as the clusters are trying to stabilize. We have users trying to access the tribe nodes but get errors. We are looking to reduce the downtime of the tribe node. And when I was listening to blizzard (which looks like we have a similar setup) they mentioned there was a patch that was applied that helped with the downtime of the tribe node.

8

Please provide logs.

9

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