Hi All,
I'm not that experienced in Elastic Search, but I have a problem and I'm out of ideas to try.
In a test setup, I have a 3-node cluster running ES6.8.22 on windows.
Until a windows reboot (for updates) 2 days ago, all queries were fast (<5ms).
But now we get very regularly (~50%) that queries take 1sec or even 2sec. (and it's always a full sec, not 700ms, but always 1000ms +-20ms.)
We have this on search queries, but also on the _doc endpoint on basically all indexes. the queries are not logged in the slowLog, I think because ES 'thinks' they are fast? (the profiler always says they are fast, but the "took" field in the response is then always "1000".
CPU or Memory do not seem to be the issue. I already tried setting the refresh_interval to 30s or to -1, but no difference.
Enabling ARS, does make more queries return fast, but not to an acceptable level.
Any input is highly appreciated.