Increased CPU usage on data nodes after Elasticsearch upgrade from 8.8.2 to 8.15.3

Hi!

We are observing 1.5-2x increase in CPU usage after upgrade of our ES cluster from version 8.8.2 to 8.15.3.

Our cluster is hosted on AWS EC2 instances, it consists of 6 nodes: 3 master nodes (c6i.large) and 3 data nodes (c7i.8xlarge).

Load on the cluster is mostly "search" requests from our search & recommendations APIs (no significant indexing is going on).

This is how CPU consumption on data nodes looks before/after upgrade:

On a " stack monitoring" tab, some weird things I see is a spike on "cgroup" graph:

and increase in "request time" for our most used index:

Please advise what is the most convenient way to debug this.

Thanks in advance, Alex.