[Zope] Troubleshooting Zope 100% CPU utilization

Matthew X. Economou xenophon at irtnog.org
Wed Mar 9 16:22:39 EST 2005


After running properly for about an hour or two, my Zope instance stops
responding to HTTP requests and begins consuming all available CPU time.
How should I go about troubleshooting this problem?

Specs:

Digital AlphaServer 2000 5/250
FreeBSD/alpha 5.3-RELEASE
Python 2.3.4
Zope 2.7.4
Plone 2.0.5

I have already looked at the following:

1. I recently imported a number of objects from a Zope instance running
on a completely different architecture (i386 to alpha).  i386 is
big-endian, alpha is little-endian.  Could this be a problem when
exporting and importing objects?

2. I examined the Zope web access and event log files, but I can find no
common event prior to when the Zope instance stops responding.  There
are no error messages in the Z2.log file, neither are there any errors
in the event.log file (in fact, the last line is "Zope Ready to handle
requests").  Running Zope in debug mode does not result in additional
logging beyond this.

3. I traced the Zope process.  There are a number of calls to poll(2)
and the SIGPROF signal handler (which resets itself, apparently, as I
see follow-up calls to gettimeofday(2) and sigprocmask(2)).  These calls
repeat.

4. I ran lsof on the Zope process.  Oddly enough, there are still a few
established TCP connections between Zope and the Apache web server in
front of Zope.  There are several other TCP connections in the state "no
PCB, CANTSENDMORE, CANTRCVMORE".

What else should I check?

Best wishes,
Matthew

-- 
"The challenge of a moral life is to do nothing that requires
forgiveness."  - Roger Ebert in his review of _The Woodsman_


More information about the Zope mailing list