Chris, Well, it's nice to know I'm not alone :) I have yet to find a way to fix it: http://atlas.gc.ca/cgi-bin/zopegraph Those are 1 hour samples. I applied the cAccessControl patch, removed my usage of REQUEST.set as much as possible, and even unset all my PythonScript variables as one had suggested. This has changed nothing. I'm also 99% sure I don't have any circling/cyclic refrences (I'll try to go back again and check even more thoroughly, but I doubt it, I don't write Python that fancy just yet :) I have one Pyton/C external module that is "non standard", and checked it for leaks, nothing came up. But, if we're suffering from the same issue, then this obviously wouldn't be it (You don't use the Python interface to PROJ.4 do you ? :) I definitely know that the leakage is proportional to usage (The leaking slows down, or even stops, overnight), but I'm not closer to figuring out where it comes from. I've kept an eye on refcounts, but I have yet to see a single refcount that grows out of control (Even the AccessControl ones, but applied the patch anyway, just in case), which suggests the leak is NOT related to object instances, object caching, or anything like that. I'm getting really desperate here ... The funny thing is when the memory grows too large, it seems one of the processes/threads dies with a SIG11, and Zope restarts itself, clean memory! I'm currently using Zope 2.5.1 with CMF 1.1 (2 instances) and the latest Psycopg 1.0.13, and I also allow the importing of some extra modules into Zope (urllib is one, my custom Python/C module is the other). Maybe if we find what we have in common, we can try tracking down what's going on here? I've been thinking fo filing a bug, but I wouldn't even be sure what to say in it!!! PS: I copied to the list, the more eyes the better :) Let me know what you find out, if anything! Cheers, J.F. -----Original Message----- From: Chris Muldrow [mailto:cmuldrow@fredericksburg.com] Sent: Friday, January 17, 2003 10:50 AM To: Jean-Francois.Doyon@ccrs.nrcan.gc.ca Subject: Zope memory leak Did you ever locate the source of your memory leak? We seem to be having similar problems, including very high deltas on DateTime (very very high!) and on BTrees._IOBTree.IOBucket. Did anything you did help out? Thanks! Chris muldrow@mac.com