The java application server in the pod was killed due to excessive memory usage

I can see that about 2.2gb is retained in the heap dump through a single instance of co.elastic.apm.agent.collections.SpanConcurrentHashMap, so that definitely looks like a memory leak.

However, as this class isn't present in the code anymore it could be something that is already fixed. The next thing you should do here is to upgrade the agent to the latest version (1.49.0 as of today) and see if this issue happens again.

If it still happens, then you should capture a new heap dump so we can investigate it.

1 Like