Master Data Node Architecture Help


(Phani Nadiminti) #1

Hi All,

I have planned a production cluster with the following configurations. Is this correct to form the master data architecture of elasticsearch..

I have 3 master nodes only with 8 GB configuration of each node.;
I have 2 data nodes with 30 GB of memory to handle the data.

I have planned all write requests and cluster operations will hit the master and all the read operations will hit the data nodes is this correct?

so I have disabled multicast and i have enabled unicast pinging to all my nodes master and as well as data nodes is this configurations.......... is fine or we don't have to include the data nodes in ping unicast host list.is it correct configurations please help me in this.

Thanks
phani


(Ed) #2

MMm, if your only having two data nodes having a master really does not make sense, you would be better off having 3 data nodes at that point

Try it this way instead If you want "Caching" use 2 client nodes (IE DATA and master turned off) , let the data nodes be both master and data

This way all the client nodes will receive all client searches, cache and when you go to add data nodes you don't have to tell your "Kibana/Other users things have changed"

Separating the master node out is useful only if your having a large cluster ( I wold say >5 data nodes) As the masters main job is to keep track of the Cluster Health, and you really don't want users breaking that with large queries and what not.


(system) #3

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