So you've probably tried this already, but when your site is in the "failed" state can you get to Zope directly instead of through squid (i.e., at port 8080 instead of 80)?

That would give some indication of whether the problem is with Zope/Plone or with Squid.

Another thing to try is running your Zope instance in the foreground, since that gives more on screen logging than the event.log provides.

I have many virtualized Plone instances (mostly on Ubuntu 8.04LTS rather than BSD), and I have had zero problems.

Another thing to check (again, which you've probably already done) is resource utilization on the virtual machine--are you out of disk space or memory? When a disk fills, Zope tends to just stop responding.

I've recently switched from Squid over to Varnish, and I have not regretted it for a second. Squid is an unwieldy beast.

I actually wrote up a howto at Plone.org (but it has been "Waiting for reviewer" for about a month now):

http://plone.org/documentation/how-to/plone-behind-varnish-using-pound-for-ssl

J.

On Sep 8, 2009, at 10:20 AM, Ang, Trish wrote:

Helloo Plonistas,
 
 We’ve been having a problem with our main Plone server. So here’s the gist of it:
 
 We have a Plone 3.2.3 instance with Zope 2.10.7-final running on a virtualized open bsd machine. Everything was fine and dandy for a few days, then a plague of errors struck. It has been crashing the instance with the following error message at least once a day (see attached file) for the past three weeks, not super regularly but very often around 8 or 9 in the morning, and then again sometimes around 11pm. When the errors start showing up, the instance is still running although you can’t access the ZMI or any websites. The VM host is also still running at that time.
 
 We’ve set up a chron job that should check to see if the instance is down and then restart it, however this no longer works since the instance is seemingly still running. Nothing is logged in the error_log in the ZMI, in the squid-cache.log we have ( http://pastie.org/609799 ), and finally in the event.log we have ( http://pastie.org/609819 ).
 
 Does this look like anything y’all have come up against before..? We’re pretty much shooting blind on this one, as we haven’t been able to locate anything that might be the root problem. Most of our other Plone instances are also on virtualized machines and run smoothly. This is our first virtualized Plone instance that is in production though. Any help would be appreciated, or if you can recommend a good Plone consultant who could work with this. Thanks everyone!
 
 Trish
  <2009-09-08.png>_______________________________________________
ZUGOD-Discuss mailing list
ZUGOD-Discuss <at> velocipede.dcn.davis.ca.us
http://www2.dcn.org/mailman/listinfo/zugod-discuss

Hi Trish,

Does restarting Zope fix it? Or does the virtual box need restarting?

It would be interesting to ssh when Zope's unresponsive and check to see if it's spinning (consuming a lot of CPU) or inactive. You could also check storage, including file handles available at that moment. vmstat output could also be interesting.

Steve

BTW, if the Zope client appears running, even when not responding, you could have your cron job try to connect to zope rather than just check it's status.