[ZODB-Dev] ZEO non-communication problem

[Jeremy Hylton]

On Thu, 2003-09-11 at 07:18, Chris Withers wrote:
> Having problems with ZEO communication on Redhat Linux 7.2.
> Both the storage server and client are running on the same machine.
> The event logs for both event they've started up okay and connected to each 
> other fine.
> 
> However, when I actually do a request to the client, I don't get any response. 
> There's no CPU activity on the machine either.
> 
> Now, the finger points to ZEO as when I remove DBTab's custom_zodb.py, which 
> mounts up the client storages, requests are handled fine.
> 
> Any ideas on what could be happening here? And, given that I'm not certain ZEO 
> is the problem, how on earth I'd go about debugging it?

I think there have been problems in the past with interactions between
Python's import lock and dbtab.  You always run the top-level module or
package __init__ code on import.  If that causes a chain of imports
where the same module is imported twice, you will get a deadlock.

You may be able to check this by running python -v, but I don't recall
how exactly it prints import status.  You could also attach to the
dormant Zope process with gdb and see where it's threads are sitting.

Jeremy