Noticed a new OOM in prod on JMAP pods

[chibenwa]

The Cassandra liveness check was enough to shutdown and reboot the pod.

No idea what caused the OOM - a log audit of nearby events did not lead to anything convincing...

I will:

• [ ] Audit GC logs (heap infos, gc infos)
• [ ] Hunt memory leak if any
• [x] Strictly limit count of /set and /get as per JMAP specification

[chibenwa]

Over 3 days without OOM.

GC logs show a healthy app.

Heap analysis shows nothing.

I did review the log activity on that time period, which was moderate to low.

My theory is still a rogue request alone pulling alone too much resources and crashing the POD.

I hope https://github.com/apache/james-project/pull/2096 to be a good mitigation...

[chibenwa]

No longer happening...

Closing for now.