The service is crashing with OutOfMemoryError in production. How to localize the problem?

For localizing OutOfMemoryError in production:

1. Collect a heap dump at the moment of failure or immediately after the error occurs. This can be done using JVM parameters (-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/path/to/dump) or tools like jmap.

2. Analyze the heap dump using profilers (e.g., Eclipse MAT, VisualVM). Look for:

   • Objects occupying a lot of memory.
   • Memory leaks — objects that should not live that long.

3. Check logs and metrics:

   • Use monitoring (e.g., Prometheus, Grafana) to track memory consumption over time.
   • See if memory usage gradually increases (memory leak) or suddenly.

4. Check JVM settings:

   • Heap size (-Xmx, -Xms), garbage collector parameters.
   • Consider increasing memory or optimizing GC.

5. Review the code:

   • Look for large collections that are not cleared.
   • Pay special attention to caches, singletons, static fields.

6. Test locally with load to reproduce the problem and verify hypotheses.

Example command to manually create a heap dump:

jmap -dump:format=b,file=heapdump.hprof <pid>

Analyzing the dump in Eclipse MAT can help identify "Leak Suspects" — objects that retain memory.