Dedicated master node and client node

Hi,

Currently in our ElasticSearch cluster we have one dedicated master node
(master: true, data: false) and it's also the nodes that receives all
requests. This is bad and we want to change it.
Here's some questions:

  1. Is it a good idea to merge dedicated master nodes with client nodes?
    Say in the cluster, we have 3 machines at dedicated master nodes and we do
    some kind of round robin DNS on these 3 for incoming requests.
  2. If we opt for separated, dedicated master, could we use smaller
    instances? Say 3 machines of m3.medium.
  3. In general, what's the general suggestion for a cluster setup? We are
    currently talking about: 5 data nodes, 3 dedicated master node and 2 client
    nodes.

Thanks!

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/4a6ad3d7-b8dd-4e7d-8395-a583b3b70df4%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

  1. That is ok, but just make sure you size the heap to account for large
    queries (ie aggs) or your master could still OOM (which is bad). You may
    find as your cluster grows it'll make sense to split the masters and
    clients.
  2. Should be ok, the master doesn't need much heap. But you don't want
    to go smaller than that.
  3. Depends. Unless those nodes are massive and you are doing a lot of
    heavy queries then I (personally) wouldn't bother with masters or clients,
    and just round robin with your existing nodes.

If you are combining master+data+client nodes, then make sure you monitor
them for heap use :slight_smile:

On 24 December 2014 at 02:28, Han JU ju.han.felix@gmail.com wrote:

Hi,

Currently in our Elasticsearch cluster we have one dedicated master node
(master: true, data: false) and it's also the nodes that receives all
requests. This is bad and we want to change it.
Here's some questions:

  1. Is it a good idea to merge dedicated master nodes with client nodes?
    Say in the cluster, we have 3 machines at dedicated master nodes and we do
    some kind of round robin DNS on these 3 for incoming requests.
  2. If we opt for separated, dedicated master, could we use smaller
    instances? Say 3 machines of m3.medium.
  3. In general, what's the general suggestion for a cluster setup? We are
    currently talking about: 5 data nodes, 3 dedicated master node and 2 client
    nodes.

Thanks!

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/4a6ad3d7-b8dd-4e7d-8395-a583b3b70df4%40googlegroups.com
https://groups.google.com/d/msgid/elasticsearch/4a6ad3d7-b8dd-4e7d-8395-a583b3b70df4%40googlegroups.com?utm_medium=email&utm_source=footer
.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAEYi1X_8%3DX79-NeqDfoOc-OH5_8YG6eK_6O-XvZT2QbLdak1uA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.