Connections being lost, similar to a memory leak

[@fujisawa.com Pete Byer]

I've just run into the strangest transaction pooling/connection problem and
I hope somebody has heard of a fix.

My scenario:
I have a very simple webpb application where the user must log in to the
system. When they enter their user id, the id is used to as the database
userid connection parameter. This application has been running for a couple
years (along with others), but we hadn't realized the problem. Our
transaction pooling is set at a min 2/max 10 connections. Its not a high
volume system. I've found that over time we lose the ability to make
connections and finally found the reason. Every time a database connection
is refused for ANY reason (like invalid userid), we lose one connection from
our max connections. After 10 refused connections (the transpool max), the
webpb application server will no longer make a connection to the database.
We then have no alternative but to restart the application server to restore
our connections. This problem is exactly like a slow memory leak.

This application is no big deal to restart, it's just a hassle. But in the
bigger picture, we have another webpb application server that actually hosts
4 different applications (2 applications use database1 and the other 2 use
database2) on it. As soon as one of those applications "loses" its
connections, that server has to be restarted. Even if application1 using
database1 causes the problem, application2 using database2 will no longer
connect.

In case you think I'm nuts, here is an easy test for you to conduct on your
own webpb app. Change your database connection parameters to use an invalid
userid. If your transpool has a max of 10 and you try to connect 10 times,
you will get 10 errors stating an invalid userid. On the 11th attempt, your
connect will return a 0 (success), but you can not retrieve any data from
the db. You will usually get something like a settrans error. Once your
connect returns the 0 return code, you will never connect again, until you
restart the server (to clear the trans pool).

Anybody run into this and/or heard of a fix (or at least a bug report)? I
couldn't find any specific mention of this type of error anywhere, but I did
read several threads in this newsgroup that seemed to have the same
symptoms.

We are using PB 6.5.1 (also occurs on pb 7.0), running on windows NT. The
results have been reproduced connecting to (or should I say not connecting
to) Sybase 11, Sybase 12 and Oracle.

TIA,
Pete Byer