As mentioned in the above discussion, a user is reporting that the aggregate PGA limit is being reached in their environment. They provided their number of log switches and it seems their switch count fell drastically over the recent days before the error. It could be related to the fact that lower log switches mean a longer running ad-hoc session which might be why they experienced the issue.
As proposed in the discussion, the following would be considered:
- Add a JMX metric where we capture the PGA memory consumption for the session
- Periodically log the PGA memory consumption in the logs
If determined that the PGA memory consumption is growing under this situation, then it would be advisable to introduce a PGA threshold configurable option that users can set to specify a point in which the connector should terminate the current session & re-start if the memory consumption grows beyond the configured value.