Node Replicas suggestion on production environment

Hi All,

I have four node cluster established for production i have 5 shards and 1 replica is there for every index. Four nodes are master and data enabled.Now can I increase replica to 2 to increase search performance on cluster please suggest me how can we get more search performance.

Thanks
phani

Are you having performance problems now? How are you measuring this?

What does your config look like? What specs are your nodes?

Hi Mark,

we have designed search applications based on elastic search.Now performance of search is fine and we are getting results faster.on my node search apps goes on increasing day by day.so I read in es guide increasing replicas for index will increase search performance i thought like if search apps increasing we will hit more index request so es should server all those requests.as of now performance wise we are good.
I am measuring speed from search apps based on elastic search.nodes are establish on different servers on RHEL enterprise 6.5 operating system and i have 3 core processor on each node and 24 GB of RAM for each node.configured 12GB heapsapce for each node.all es configurations are defined in /etc/sysconfig/elasticsearch
file like directory paths and heapspace etc...

Thanks
phani

Replicas add concurrency, which can definitely improve performance.
This is something you would need to test and measure though.

Okay Mark Thank you and please suggest me with above 4 node architecture is it safe to add 2 replicas for each index ?

Thanks
phani

It's safe but may be overkill.

You also really want an uneven number of nodes to protect against the possibility of a split brain.

But as long as discovery.zen.minimum_master_nodes is set to N/2+1 an even N should be fine, right?

True.