[Zope] Zope and memory leaks

[Toby Dickenson]

On Monday 30 Sep 2002 8:30 pm, Howard Hansen wrote:

> Here's my question, though.  On my old machine, the Zope process leaked
> memory like all get-out.  For instance, in a mere 14 hours, it went fro=
m
> 75MB to 285MB.  I had to restart the Zope process daily.  Here's the ou=
tput
> of the debug page:

Thanks for posting this, it is very interesting. It looks like you captur=
ed=20
5530 cycles. Each cycle involves a folder, a response, two conflict=20
exceptions, and three request objects.

I wonder what makes a conflict-retried transaction create an unbreakable=20
cycle, when non-retried transaction does not?

Some questions:

Had you turned off python's cyclic garbage collector in your old=20
configuration?

Is it the same version of python?

Are you seeing many conflict exceptions in your new configuration?

Do you have any idea why you had so many conflicts?


> Class                                           00:28     14:37     Del=
ta
> -----------------------------------------------------------------------=
--
> Acquisition.ImplicitAcquirerWrapper            16,253   107,895   +91,6=
42
> ZPublisher.HTTPRequest.HTTPRequest              3,267    19,872   +16,6=
05
> ZODB.POSException.ConflictError                 2,171    13,067   +10,8=
96
> OFS.Folder.Folder                               1,807     7,362   + 5,5=
55
> ZServer.HTTPResponse.ZServerHTTPResponse        1,094     6,635   + 5,5=
41
> ZPublisher.BaseRequest.RequestContainer         1,091     6,632   + 5,5=
41
> ZPublisher.BaseResponse.BaseResponse            1,089     6,621   + 5,5=
32
> ZServer.HTTPResponse.ChannelPipe                1,087     6,619   + 5,5=
32
> AccessControl.DTML.TemplateDict                    42       121   +    =
79
> Products.CMFCore.FSDTMLMethod.FSDTMLMethod          6        82   +    =
76
> OFS.Image.Image                                   286       333   +    =
47
>
> Memory Usage                                     75MB     285MB     210=
MB