[Bug] Pulsar logs are truncated when logging to files and Pulsar runs out of direct memory when -Dpulsar.allocator.exit_on_oom=true

[lhotari]

Search before asking

• I searched in the issues and found nothing similar.

Version

2.10.x

Minimal reproduce step

It's possible to reproduce the issue when running Pulsar in a configuration where Pulsar logs to log files. (a VM based deployment). To reproduce, configure the direct memory to a low value and stress the broker with sufficient load so that the direct memory runs out.
This problem doesn't appear in Kubernetes environments since logging uses System.out / stdout which flushes after each line by default.

What did you expect to see?

the Pulsar logs should be writted to disk before Pulsar exits the Java process

What did you see instead?

Pulsar log files are truncated and don't contain the last logged lines

Anything else?

No response

Are you willing to submit a PR?

• I'm willing to submit a PR!

[lhotari]

It seems that LogManager.shutdown(); should be called before Runtime.getRuntime().halt is called in the Pulsar code base.