-
Notifications
You must be signed in to change notification settings - Fork 25.6k
Description
Describe the feature:
I have setup my ES cluster to write to ingest only nodes, and read from coordinator nodes that's not master nor data nodes.
The problem is if I add more coordinator "client" nodes (not master, not ingest, not data) to server query spikes, there's no way for the client to auto discover those nodes without including either master or data nodes. The workaround I have now is to make those new nodes to be ingest nodes as well.
So I just auto discover ingest nodes for read, but then I won't be able to split write & read nodes.
The only way to discover nodes via REST afaik is to use /_nodes/xyz, where xyz is a comma separated "node roles". Maybe there's another way to do this? Or maybe splitting read from write like what I'm doing is overkill?