users@glassfish.java.net

Re: Connection Pool monitoring data - spooky math?

From: Shalini <shalini.muthukrishnan_at_oracle.com>
Date: Mon, 05 Nov 2012 10:00:08 +0530

Please file an issue with all the steps and the connection pool
configuration.

On Thursday 01 November 2012 07:41 PM, Bryan Varner wrote:
> I'm in the process of migrating an existing system to glassfish 3.1.2.2
>
> At this point, we're trying to get the bean pool's & connection pools tuned appropriately for our apps.
>
> I've enabled high monitoring server-wide, and started running some normal usage pattern test cases on our servers.
>
> I'm accustomed to C3P0 connection pools, which are really easy to instrument and gauge your pool resource usage.
> With C3P0 the most pertinent data points we monitor when tuning the app are:
> numConnections (current pool size)
> numBusyConnections (number of connections currently in use)
> maxPoolSize (maximum size of pool)
> minPoolSize (min / steady pool size)
>
> It looks like the analogous values in GF monitoring would be:
> NumConnCreated - NumConnDestroyed = current pool size
> NumConnUsed = number of connections currently in use
>
> And then I could read the min / max off the pool's config.
>
> The problem that I have, is that the numbers I'm seeing in the GF monitoring data don't make sense!
>
> For instance, right now I have NumConnFree = -3989, and NumConnUsed = 4000, on a pool that's configured with a maximum connection size of 75.
>
> How can the number of free connections in the pool be a _negative_ number, and how can NumConnUsed be so high?
>
> Am I reading the intent of these metrics incorrectly because I'm expecting to be able to see the current state of the pool rather than just hit counters?
>
> Help!?
>
> Regards,
> - Bryan Varner