Details
-
Improvement
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
1.1.0
-
None
-
None
Description
The default JVM options for the workers specify the path to use for the heap dumps, see
https://github.com/apache/storm/blob/1.x-branch/conf/defaults.yaml#L171, however when a memory error happens more than once for the same worker, only the first dump is kept as the file can't be overridden. Instead, would it make sense to use something like
"-XX:HeapDumpPath=artifacts/heapdump<pid>.hprof"_ so that a different dump is generated for each JVM instance? Or is the current pattern used on purpose to avoid too much disk space being used?