JDBMaxPoolSizeException error messages

Hello;

We have been running production since March '06. In that time we have had this recurring problem. When you hit the limit on the connection pool all of the sessions running on that webserver come to a complete stop. The default for this setting is 50. First time we ran into this problem support suggested setting it to 100 which I did. Second time, suggested 200. Third, 500. For the last week or so it has started happening again. Can run a webserver instance for almost 1 day before the problem crops up. Just increased the setting to 1000 but obviously there is something else wrong here. We only have about 60 users concurrent.
Details: Websphere 5.0.2, E1 8.11sp1, Tools 8.86a1. One 11,000 CPW AS/400 enterprise server and one Intel-based enterprise server. 2 Webserver machines. Each is Dell 3.0 Ghz XEON quad-processor with 12 GB RAM running Windows 2000 server Ent. Edition. Restarting the webserver or starting a different one solves problem for one day so I can keep us alive by restarting every night but need a better solution.

Thanks all.

Just a couple suggestions.

First I wouldn't suggest running any tools release until you hit C1. I think you'd have been better off with 8.95_N1 instead of 8.96_A1.

Second I see that you're running WAS 5.02. Is this base? If you're running 5.0.2 on Intel then you really should be running 5.0.2.12 with the JDK_SR9. You can check by running X:\WebSphere\AppServer\bin\versionInfo.bat and X:\WebSphere\AppServer\bin\java\bin\java -version

Third you web servers are massivly overpowered. I could run 1500 users concurrently on your configuration. Not that this a bad thing but I don't think you've got it properly configured. First of all how much memory have you allocated to the JVM's? Did you install WAS ND (Network Deployment)? Are you using clustering? (you should be). How have you load balanced the 2 servers (horizontal, WAS Edge Server, DNS Round Robin, Hardware Appliance?

Can you post your JAS.INI and the JDBJ.INI? I really think your issue is elsewhere. It might also help to post the Enterprise Server JDE.INI.

Also why do you need an Intel Enterprise Server if you have 11,000 CPW and only 60 users? Unless there is a real need unnecessarily complicating your infrastructure may actually slow it down.

Finally, it's really rare to run out of JDENET connections. This is set by the following parameter in the JAS.INI

[JDENET]
maxPoolSize=50

I'd check in SAW to see if the number of connections is really running out or not.

I've got 150 concurrent users on a 2 CPU Dell machine with only 6 GB or RAM (WAS is clustered) with stellar performance (8.11 with 8.94).

Colin

Are you making the change directly to your jdbj.ini file and restarting the JAS instance? If not this solution documents might apply:

Solution ID 200992501 on the Customer Connection states that changing the parameter through SAW "is not as reliable as making the change directly in the jdbj.ini"

Typically I set this value to the number of expected concurrent user sessions. I have not found any documentation to support that formula but it seems to be the general number applied on various "official" pre-golive technical audits done by Oracle consultants that I have reviewed.

Could you post your jdbj.ini?