We updated our elastic (5 servers, two nodes each) from 7.2.0 to 7.8.0 and noticed that from this moment some new errors of type CircuitBreakingException started to occure and heap size can touch now 90-100% of available memory which was not the case on 7.2.0. Our routine didn't change.
Additional deatails and observations
- Update was made without reindexing documents and just uses 7.2.0 data (can it affect something?)
- We use default config for es in both cases
- Docker container in both cases
- Heap size is near 32Gb in both cases
- After update we have permanent zero gc stat
"old": {
* "collection_count": 0,
* "collection_time_in_millis": 0
}
which I think is strange
6) We use bundled java in both cases
7) I noticed that jdk changed from Oracle to AdoptOpenJdk, can it be connected with my problem?