[ZODB-Dev] ZEO client leaking memory?

[Toby Dickenson]

On Fri, 05 Oct 2001 08:51:30 -0700, sean.upton@uniontrib.com wrote:

>but I'm still interested in finding out what sort
>of problem I'm actually having.

I have some further questions......


> week or two ago, Zope was using about 1200MB on each of my 2 =
load-balanced
>ZEO client nodes, and each box, which has 512MB and about 1GB of swap =
was
>unable to fork new processes.  Well, the site is a relatively high =
traffic
>site, and I do three intensive tasks automated daily in the middle of =
the
>night:

Are you sure that the problem comes from adding objects, not the work
during the reast of the day?

What is the memory usage like after a days work, before adding the new
documents?

> first I add a few thousand objects to a bunch of folders containing
>an aggregate of about 10-15k objects.

A Zope document (the thing you can see in the management interface)
can be made up of multiple ZODB objects. How many ZODB objects make up
your 'document'.

Are you using regular Zope folders or BTreeFolders? How many documents
in each folder?

>  I then do a bulk ZCatalog reindex,
>then I pack anything older than 1 day.  I've added to this mix a cron =
job to
>restart Zope on both my nodes after all this is done, which has seemed =
to
>solve my problem for now,=20

Also:

What are your ZODB cache settings (size and time)

How much memory is used by a 'document'

During the automatic adding, how fast are documents being added
(number per minute)

Have you checked the size of the ZODB caches while memory usage is
growing?

Have you checked the reference count debugging page while memory usage
is growing?




Toby Dickenson
tdickenson@geminidataloggers.com