Indeed the problem looks very similar to yours (100% cpu always, data inter node about 10 TB! And we ingested like only 8GB in the interval). I'm not in my working computer right now but I will provide a view of what happened:
After nobody from elastic mod team could help here, I had to "pay to play". As I am the only one managing the cluster I upgraded my license to platinum and filled an urgent ticket. My cluster was green but useless in this meantime, and I couldn't wait three business days, the SLA for normal things.
The engineers were very kind, provided two fixes during one day, and sent me a bunch of commands to run in dev tools a few times (4 hours interval, run the command).
And then the elastic dev team solved the issue and fixed the cluster.
I think you should do a rollback if you can (I couldn't find this option in elastic cloud). Probably elastic will provide a wider fix in the next days/weeks for the public. Hope the best for your cluster.
Workaround did not resolve the performance issue. 7.11.2 Was fast. 7.12.0 is not.
When data is cached it is sometimes fast.
Even some Markdown text which does not require any further query result takes long to render.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.