Converting from a single jvm per host to multi

(BS) #1

In a cluster with 20 servers @ 256GB RAM hosting a single JVM at 30G, what would be the procedure for converting to 4 JVMs per node ?

If we take one host out, reconfigure it w/ the appropriate shard awareness stuff, and start up the 4 new instances, will Elasticsearch allocate 4x as many shards to this server, causing a hotspot until we can repeat for the rest of the servers ?

(Mark Walkom) #2

It shouldn't immediately do that. Of course as you add more data it will.

We don't really have a process for this, but what you've mentioned is ok. Just watch out if you change the data path, cause you may find your data "missing" if you change this/

(system) #3

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

(Mark Walkom) #4

(BS) #5

Chiming in on this old thread with a summary of our findings:

  • when we added one multi-jvm server to the cluster, Elasticsearch did tread each JVM as a "regular" node and we ended up with 4x the data on this new server.
  • to combat this, we used the watermark setting to prevent too many shards from being allocated to the new node

(system) #6