We have our cluster behind a generic HTTP load balancer, so can't readily determine which node a request is hitting initially. Is there any straightforward way to identify from an API response, which node actually serviced a request (acted as the coordinating node)? (My fallback option here would be to persuade our networks people to add it to the response headers at the LB, but that will only be IP, not node name or ID)
Adding a "profile": "true" option will give plenty of detail about the nodes to which shard queries were sent, but there seems to be nothing in there which states the coordinating node or anything about activity on it. The profile docs point out that profiling doesn’t measure network latency, time spent in the search fetch phase, time spent while the requests spends in queues or while merging shard responses on the coordinating node) - is that a deliberate choice not to include (considered not significant/useful, or that the information is already adequately covered as metrics like thread pool stats) or is it hard to evaluate?
Are there other "tweaks" that can be added per-request to get metadata about the request handling?