High search latency after upgrading from es 6.8 to 7.12.1

[frank_thyes]

Since we upgraded our elasticsearch cluster from version 6.8.0 to 7.12.1, consisting of 6 data nodes and 3 master nodes, the search latency is three times higher than on the previous version Are there any known issues?

Any help would be appreciated.

Regards
Frank

[Mark_Harwood]

There was an optimisation to date aggregations that went in that was good for most use cases but turned out to be slower in some edge cases. This is discussed here. If that's your problem there's a switch added in 7.13 that disables this feature.

[frank_thyes]

Unfortunately we don't use any date aggregations.

[twilson]

We had a similar issue after upgrading to 7.12 and we also noticed CPU usage on most our elasticsearch nodes was very high after the upgrade. I found an issue in github that looked promising and it did, in fact, solve our problem. It has to do with searches in non-default Kibana spaces not being deleted properly. There is a workaround given in the github issue page.

https://github.com/elastic/kibana/issues/96124