Has anyone experienced a problem when using multiple interpreters where the whole JVM just exits?


We have been encountering this issue for some time now. Initially, it was due to the interpreter.exec() method creating a new class everytime it was called. This would eventually cause the classloader to fail. To solve that problem we cached the PyCode objects for each interpreter. We now monitor the number of PyCode objects that are being created, however, under heavy load when there are 50-60 Interpreters running at once the JVM still exists for seemingly no reason. We thought the issue may still be the number of PyCode objects being created so we forced all interpreters to share the PyCode objects through a global cache. The problem still exists.


Any ideas would be appreciated,




-- Cameron