Albert

Unfortunately, I have very little practical prod experience with the ingester so I can't promise quality help here. Having said that, taking a cursory look at the metrics provided shows...

> we found currently, the bottleneck is Ingester, not elasticsearch. How did you come to this conclusion? > When we use pstack to check ingester, find ingester is waiting all...

@judyQD thanks for correcting me :) my assumption was incorrect and indeed jaeger does use the ES bulk API. What I suggest to do as next steps is to: 1....

@judyQD did you manage to improve the ingester's consumption rate?

@judyQD Did you manage to get some measurements from the ingester and try modifying some of its parameters https://github.com/jaegertracing/jaeger/issues/2927#issuecomment-830422922? For example, perhaps adding more workers or influencing the flush rate...

@judyQD If I'm not mistaken, the "Kafka consumer goroutine" (after consuming the span from the topic) eventually attempts to write/add the span to a buffer that is eventually bulk written...

That would be appreciated, @lonewolf3739.

I believe this is fixed by https://github.com/jaegertracing/jaeger/pull/3854, however it looks like GH Actions runs against the tag, so it wouldn't pickup that fix even if we could re-run the release...

I'll close this issue for now in favour of waiting for the next major version release in 2 weeks, which should contain the the windows release. If an urgent release...

> TBH, it's not clear to me if querying OS for aggregate data is preferable over the Monitoring tab approach of using a metrics system for the same aggregates. It's...