With the default Elasticsearch helm charts configuration all elasticsearch pods plays all roles(master,data, client). Also there is example configuration for setting different roles.
Wanted to understand which is the best solution in production, is it to deploy with all pods playing all roles or having pods with specific roles ?
If we use the solution of all pods performing all roles, can you please help to know how to we scale the pods when there is more requests or want to store more data. By scaling the pods, it would also indirectly have the same number of masters as the number of pod replicas. But have read some blogs which suggests to have odd number of masters to prevent brain split issues in cluster.
Please help with the above queries.
Thanks in advance.