[cvsnt] CVS HEAD stuck?!?

[bwhicks at aep.com]

cvsnt-bounces at cvsnt.org wrote on 06/29/2007 05:36:45 PM:

> SSPI.  I can go on to any server at any given time, stop the service, 
> and there will be zombies.  I'm not sure how long it takes for them to 
> reappear after restarting the services since I hate stopping CVS; there 
> are usually 10-20 users actively doing something on any of our CVS 
> servers at any given time.  Maybe this is one of those "big user" 
> problems.

FYI, we also have a zombie cvs process problem. We are running cvs 
2.5.02-2099 on linux (RedhatES3) and use SSPI connecting to an Active 
Directory. We see ntlm_auth getting hung up, which in turn hangs cvs. It 
looks like ntlm_auth is waiting for a response from the AD server, but 
from the AD server, it looks like all the necessary responses have already 
been made. Our solution was to run a watchdog script that checks if a 
particular cvs process has been alive past a certain amount of time (i.e. 
2 hours) and sends an email to an admin who can determine if it's ok to 
kill it. 

We are getting ready to upgrade to RH ES4 and cvs 2.5.03.2382. I'll post 
back to the group if it makes any difference...

Brian