[Zope] Zope threads dying continuosly

Dennis Allison allison at sumeru.stanford.EDU
Fri Nov 21 01:40:23 EST 2003


Tim--

I've been running Zope and MySQL on RH9 but the system has been very
unstable even after we upgraded to the bleeding edge (kernel and glibc), 
set all the magic parameters, etc.  I had been using Pound as a load
distributing reverse proxy under RH7.3, but it fails under RH9.  I wrote 
my own reverse proxy modifying Pydirector but it has had its problems,
possibly also related to RH9 thread problems.  Still, we appear to have 
had an easy time of it relative to the disasterous problems you see.

I run on a Athlon dual processor system, 4GB of memory, and a RAID disk.
We use Zope 2.6.2b3 with Python 2.1.3.  

I'm curious what kind of processor you are using.  Is it one of the
hyperthreaded Intel P4 machines?

Anyhow, the bad experiences with RH9 and the changes Redhat is making to
their business model have motivated me to explore other distributions.
Over the next few days, I'm planning to move our Zopes onto a new OS, one
based on the Gentoo distribution.  I trolled for comments a week or three
ago and Gentoo got raves from several folks I trust.

Moving to a different operating system might be the right thing.  If you 
remain with RH9, you should be running the latest kernel and the latest
glibc.

You may find it useful to export into the environment 

LD_ASSUME_KERNEL=2.4.1

MySQL and Zope to turn off the new posix threading library which seems to
be the culprit (or one of the culprits).



On Fri, 21 Nov 2003, Tim Edwards wrote:

> Hmmm... I'll try and explain a bit more about our problems. We found Zope to
> be generally unstable in that, after a few mins up to an hour or so, Zope
> would sometimes freeze. Also when multiple people were accessing a Zope
> application that retreives or modifies data in our MS SQL Server database
> the Zope server would freeze and have to be restarted - always, not just
> some of the time.
> 
> After installing a glibc update for Red Hat 9 that fixes a problem with
> Python programs not being able to create threads and specifying the -t
> option in the Zope start script (yes with 500 threads!) it seemed a lot more
> stable. Now multiple people can use our database access application without
> it crashing. As for the 500 threads I have now put it down to 9, however it
> doesnt seem to make that much of a difference in resources - a ps listing
> tells us we only ever have the normal 2 z2.py processes open at a time.
> 
> I am now running the Zope server in debug mode on a terminal and can see
> that it occaisionally crashes and automatically restarts iteself. Also if I
> go into the management interface and click on a ZSQL method it crashes - and
> then restarts itself. Applications that use those methods still seem to work
> for some reason. However I cannot get to the management interface page for
> any ZSQL method in the whole Zope installation.
> 
> Thanks
> 
> 
> -----Original Message-----
> From: Dylan Reinhardt [mailto:zope at dylanreinhardt.com]
> Sent: Friday, 21 November 2003 4:16 PM
> To: Tim Edwards
> Cc: 'Zope at zope.org'
> Subject: RE: [Zope] Zope threads dying continuosly
> 
> 
> On Thu, 2003-11-20 at 18:15, Tim Edwards wrote:
> > Thanks for your replies, the original reason I asked is that we are having
> > stability issues with Zope.
> 
> Such as...?
> 
> >  I have tried it without the demon process, but
> > that doesn't appear to help. I then reenabled the demon process and
> changed
> > the thread count to 500 
> 
> 500 is a huge number of threads.  You should probably stick with single
> digits until you have a well-formed case for doing otherwise.
> 
> HTH,
> 
> Dylan
> 
> _______________________________________________
> Zope maillist  -  Zope at zope.org
> http://mail.zope.org/mailman/listinfo/zope
> **   No cross posts or HTML encoding!  **
> (Related lists - 
>  http://mail.zope.org/mailman/listinfo/zope-announce
>  http://mail.zope.org/mailman/listinfo/zope-dev )
> 




More information about the Zope mailing list