Elasticsearch cluster with mixed OS


We would like to perform an OS migration from UBunto to RHEL.
Our Cluster is 2 nodes, ES Version of 1.3.7 and Logstash 1.4.2

We would like to go to ES 1.7.1 and Logstash 1.5.4
(*) ES 1.7.1 because we need the use of http-basic plugin.

I am very limited with resources, And would like to perform the OS migration and ES upgrade on the actual nodes of the cluster.

Can someone consult on the best and safer way to do so ?
Is it possible to have a two nodes cluster with different OS on each ? (One with UBunto and second with RHEL)



It doesn't really matter too much about the OS. But it may matter with ES.

What's your plan of attack here?


Cluster with 2 nodes- node1 and node2.
Each Index has 2 shards with replication factor of 2.

The plan is to take node2 down, then node1 is with primary shards.
Install RHEL on node2, install ES 1.7.1 and LogStash 1.5.4.
Join node 2 to the cluster, and the replication process should be able to create the replication shard in node2.
Cluster should be Green.

Then, take down node1, node2 become the primary.
Repeat the same Install steps for node1 and connect it to the cluster.

According to DOC's, replication will work from lower release into higher release, but not vice versa, since there might be infrastructure changes for indexes.


That will work, but you'd be better off installing the latest 1.7.3, not 1.7.1.

Replication of 2 with a 2 node cluster is pointless, you won't be able to assign the last replica.

You are correct, I was mistaking, it is a Replication factor of 1.