2022-09-05 15:13:32 UTC - Ben Carver: I have two difficult issues right now. First, my Java functions have been getting terminated by Linux cgroups for OOM errors. I've configured the JVM (`-Xmx` and `-Xms`) heap size to be about half of the container size, and I'm still running into problems. This is presumbaly just due to my application itself rather than anything related to OpenWhisk though. https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1662390812917769 ---- 2022-09-05 15:14:36 UTC - Ben Carver: But secondarily, I have action-level concurrency enabled such that a single function may serve many different clients. When these functions get terminated due to OOM, the requests all need to be resubmitted. There could be _many_ requests resubmitted, which causes a massive over-provisioning of functions that results in decreased performance and decreased resource availability. I'm not sure how to configure OpenWhisk to avoid this issue. https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1662390876475549?thread_ts=1662390876.475549&cid=C3TPCAQG1 ---- 2022-09-05 15:43:06 UTC - Ben Carver: Interestingly, when I profile the image used by my OpenWhisk function in a local Docker container, I often don't experience the same OOM crashes as when the function actually gets deployed. https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1662392586128049 ----
