Elasticsearch cluster architecture with 6 nodes

Hi everyone,

Please I need your recommendations regarding the following architectures for a 6-node Elasticsearch cluster, based on the following characteristics:

  1. I have 2 zones (zn1, zn2)
  2. I have 3 nodes or servers in each zone
  3. All my nodes have the same hardware characteristics (RAM, CPU, Storage)
  4. I need 6 of the 4 nodes to store data (indexes)
  5. I am working with free licensing of version 7.14
  6. For each primary index 1 replica will be configured
  7. The shard allocation awareness will be configured so that the information is in both areas.
  • Architecture1
Zone Node Rol
zn1 Master
zn1 Master – Data
zn1 Data
zn2 Master
zn2 Data
zn2 Data
  • Architecture2
Zone Node Rol
zn1 Master
zn1 Master – Data
zn1 Master - Data
zn2 Master
zn2 Master - Data
zn2 Data

Questions:
Which of the 2 architectures would work better in the event of a zn1 crash?
Which of the 2 architectures would work better in the event of a zn2 crash?

Both architectures have a majority of master nodes in zn1 so will behave the same way. They will Both be able to handle zn2 going down but will result in a red cluster if zn1 fails. You can not logically build a highly available cluster that can handle the loss of either zone across only 2 zones. You need 3 distinct zones.

I would recommend reading this section of the docs and the sections around different cluster sizes it links to.

1 Like