Forms memory leak....(?)

From: Jonathan Lemon <jlemon_at_netcom.com>
Date: Sat, 26 Mar 1994 03:14:52 GMT
Message-ID: <jlemonCn950s.1KA_at_netcom.com>


I'm seeing some annoying (!) behavior with Oracle/Forms, and was wondering if anyone else has had a similar problem.

We have a large forms hierarchy, where the users can start up a single runform instance, and then execute many different forms. What we're seeing is that after a user has been working for a few hours and visited quite a few forms, the processes' DATA+BSS segment has gotten quite huge. Something on the order of 7-9MB per runform instance is typical, with 12MB not all that uncommon. Now, multiply 9MB * ~200 instances, and we're looking at runforms taking up more memory than all of the oracle shadow processes altogether, since at least the shadow processes share memory.

Worse yet, the DATA/BSS segments never appear to shrink.

Examining the processes, it shows that its memory's working set is sitting near its configured maximum (4M) and anything over that is swapped out. To me, this means that for some reason, all of it's large data set is being constantly accessed; otherwise it would be just swapped out and simply chew up swap, not real memory.

The end result is that Oracle has turned into a Turtle, and the machine is executing roughly 300 paging ops per CPU, per second. This is on a Sequent Symmetry, running Oracle v7.0.15.mumble, and whatever version of runform (v3) comes with it. (I'm at home now, can't remember the exact version.)

This has only started occurring in the last month or so, before that, the runforms were only taking up about 2M per instance, which was acceptable. And yes, this problem is reproducible. My favorite was when I answered 'Y' to those umbiquitous "Do you want to quit: (Y)es, (N)o or (C)ancel" messages, and watched DATA+BSS shoot up 2MB after Oracle accepted my answer.

So, has anyone happened across something similar to this, or have any hints/suggestions that we should try?

--
Jonathan
Received on Sat Mar 26 1994 - 04:14:52 CET

Original text of this message