I am experiencing some strange connection pool behavior with Glassfish 9.1_01 (build b09d-fcs) running on top of JDK 1.6.0_05 on RHEL 5.1 (x64).
I configured a pool for a PostgreSQL database that uses the following parameters (please check attached image):
- Initial/maximum pool size: 8
- Maximum pool size: 32
- Pool resize quantity: 8
- Idle timeout: 3600s
In the resource monitoring page I get the following (please check attached image):
- NumConnUsed: -199
- NumConnFree: 207
Can this indicate a race condition in the connection pool code? Glassfish seems to be working fine but I'm bothered about this since I cannot properly monitor the connection pool anymore. Since NumConnUsed + NumConnFree = 8 (pool size) this might be only a display issue, but the last time I dug inside Glassfish source code I remember seeing these variables in the pool code being directly exposed via JMX, so I am more inclined to believe that those are the actual values stored used by the connection pool.
Has anyone experienced this before? Any fix/workaround?
[Message sent by forum member 'lbschenkel' (lbschenkel)]
http://forums.java.net/jive/thread.jspa?messageID=272238