I have gone through Node types in Elastic 6.1 version & read one link in Elastic Blog regarding optimizing the elastic ( https://www.elastic.co/elasticon/tour/2018/washington-dc/optimizing-elastic-for-search-at-mcqueen-solutions ). I like the architecture proposed in 40th slide in above link.
Based on it, I was planning my architecture as like same.
As picture shows,
- 4 elastic data nodes available & marked 3 out of them as master nodes also to support master electing process with principle of (available nodes/2-1)
- One Dedicated Elastic Coordinate Node created to connect for Elastic cluster from Kibana or Java Web applications (The clustered Java Application)
Here are my questions,
- To support cluster, we need to make few nodes as master so I selected 3 nodes as master. Is that correct number for 4 nodes or is it there any better approach?
- Dedicated coordinate node should be part of Elastic clustering or how dedicate coordinate node sends requests to Elastic clustered nodes?
- Sending requests from Java clustered web application to coordinate node will be good instead to Elastic cluster directly? This web application will do all operations (PUSH, PUT, DELETE, GET)
AFAIK, The coordinate node should have more memory to serve the requests passing to Elastic cluster & response to Java applications or Kibana.
Your inputs are valuable to me.