How to stop auto-relocation?


(Eugene Gwon) #1

Hello all,
I'm using 3 nodes of ES 1.7.4 with Graylog2-1.3.
currently, one of ES node is a master, and 2 nodes are data node. and these nodes have 8GB of ram on each server.

and here's the problem : If I querying too big data, then one of data node crashes with OOM.
after than, master node automatically relocate shards and replicas.
but in most time, I can restart elasticsearch in very short time, so It seems doesn't need to relocate shard.

Is there anyway to stop auto-relocation some amount of time?
If I can stop it about 10~20 mins, It would be great.

Thanks.


(Mark Walkom) #2

You can adopt the process here - https://www.elastic.co/guide/en/elasticsearch/reference/1.7/setup-upgrade.html#rolling-upgrades

But you can't do that as a node goes down, it has to be before or as soon as you realise it's gone.


(Yannick Welsch) #3

Delayed allocation (see here) will help you.


(Eugene Gwon) #4

After some test and 'little' frustrating situation, I decided to respect the nature of ElasticSearch and take peace of mind.
Thanks for all your response :slight_smile:


(system) #5