Odd PostgreSQL problem

[crazeinc]

I talked to the sysadmin and he said the network has a short TCP timeout so there's 2 C3P0 properties I'm looking at to get around this:

maxIdleTime & idleConnectionTestPeriod. I could use the maxIdleTime to just say if the connection is idle for 15 minutes, then discard it. I think that would eliminate the problem, but I'd rather not get rid of them in case of a surge of users after some dead time and there's no connections available. So that's why I was looking at the idleConnectionTestPeriod, to test the connections every 10 minutes or so, so they won't break and the Exception won't be thrown. This is all theoretical until I see if the exception happens again or not.

[Costin Leau]

In this case adjust idleConnectionTestPeriod to be shorter then the TCP timeout and if it works (it should) start from there increasing the value. After all if you say the scenario is going to be changed I wouldn't worry about optimizing the network configuration at this point.

[crazeinc]

That's basically what I'm doing now. I set the idleConnectionTestPeriod to 10 minutes and haven't gotten the exception since then so I'm happy to have this solved for the interim until the server gets moved.

[wgillett]

We were getting mysterious connection closed exceptions from PostgreSQL with our application. Switching from DBCP to C3P0 eliminated them. Thanks for the info in this thread, very helpful.

-Walter Gillett