[Bug]: Flink Job manager metaspace classloader leak

[gli-marc-hurabielle]

What happened?

Some similar issues has been reported and fixed in beam: #29890 #25510 and in flink https://issues.apache.org/jira/browse/FLINK-28248

We are using Apache Beam with the Python SDK to submit some batch jobs with the Flink REST API.

• beam: apache-beam==2.58.0
• flink: 1.18.1

The metaspace memory is only growing in flink (after each submit):

At some point, it is not possible to submit new jobs (the Flink api hang).

I am not really use to debug JVM memory leak. We tried to run jmap -clstats 1 there is a lot of duplicate class like: org.apache.beam.vendor.grpc.v1p60p1.com.google.protobuf.DescriptorProtos$DescriptorProto
org.apache.beam.vendor.grpc.v1p60p1.com.google.protobuf.Descriptors$Descriptor

but those size are really small, so I am not sure it is a problem.

Issue Priority

Priority: 2 (default / most bugs should be filed as P2)

Issue Components

• Component: Python SDK
• Component: Java SDK
• Component: Go SDK
• Component: Typescript SDK
• Component: IO connector
• Component: Beam YAML
• Component: Beam examples
• Component: Beam playground
• Component: Beam katas
• Component: Website
• Component: Infrastructure
• Component: Spark Runner
• Component: Flink Runner
• Component: Samza Runner
• Component: Twister2 Runner
• Component: Hazelcast Jet Runner
• Component: Google Cloud Dataflow Runner