CODEWITHSUNDEEP
out-of-memoryapache-flinkmatgrpc-javametaspace
ilya82
ilya82

Reputation: 1

Apache flink metaspace exception while running job several times

I'm working on fixing the issue "java.lang.OutOfMemoryError: Metaspace" in the time of running an flink job via "Submit new job" web interface several times.

My flink job gets data from gRpc service usig grpc-java. After press submit job button several times I got OutOfMemoryError: Metaspace.

enter image description here

After research I've found out that every org.apache.flink.util.ChildFirstClassLoader wasn't cleared by GC. My job was launched 4 times and MAT shows me 4 ChildFirstClassLoader in Dominator tree.

enter image description here

Using MAT(Path to GC Roots exclude all phantom/weak/soft/etc) I tried to found out who keeps ClassLoaders alive. But 0 paths was found on every ChildFirstClassLoader. It looks like nobody keeps ClassLoaders and they are ready to be collected by GC but It didn't happen.

enter image description here

Any ideas how to find a reason?

Upvotes: 0

Views: 240

Answers (0)

Related Questions