[Zope] Zope woes continue - server going down regularly.

Stuart 'Zen' Bishop zen@cs.rmit.edu.au
Wed, 8 Dec 1999 13:27:02 +1100 (EST)


On Tue, 7 Dec 1999, chas wrote:

> >Sorry, you're right, it was 'spinning' - 100% CPU resources.
> >Does that suggest anything in particular ?
> 
> Ok, I feel like the Cold-Fusion-Man today, I have a window open
> on my desktop showing the 'top' output of my server. Whenever
> the CPU usage goes up to over 90%, I know that the PCGI process
> is no longer responding and I restart Zope. Restarting the 
> server does not help any more than restarting Zope itself.
> Repeat ad nauseum every 30-45 minutes if I'm lucky.

My guess is you have some code that is going into an infinite loop
or deadlocking. And since you are the only one reporting it a further
guess would be it is code that is specific to your site, or a 3rd
party product being used in an unusual fashion, or a combination of
products.

I believe DTML has checks in it to stop this sort of thing, so it would
likely be some python code somewhere that is blocking.

For a test, start up your server in single thread mode. When the problem
next occurs, the last page accessed should be the cause of your problem.
If the standard log doesn't do the trick (medusa might log incomming requests
before they are passed off to Zope handlers), you can probably put your own 
logging information in zhttp_channel.work in ZServer/HTTPServer.py unless
you work out how to turn on the built in debug logging.

-- 
 ___
   //     Zen (alias Stuart Bishop)     Work: zen@cs.rmit.edu.au
  // E N  Senior Systems Alchemist      Play: zen@shangri-la.dropbear.id.au
 //__     Computer Science, RMIT 	 WWW: http://www.cs.rmit.edu.au/~zen