Re: Jamming up with mutex_lock

[Andrew Daviel]

On Wed, 20 Jun 2007, Andrew Daviel wrote:

> But on the production machine, I see about 20 threads. Several of them
> have been around for an hour, far in excess of the sendmail timeouts

I realise that the sendmail timeouts on SMTP DATA were indeed set to an 
hour, compared to the 5 minutes for milter timeout. We clearly have 
various ratware losing connections when spamming which explains the 
stalled milter threads. I've dropped some of the SMTP timeouts to 20 
minutes; I doubt we have legitimate users sending us 20Mb emails over a 
dialup line anymore.

When I log PPID from the milter, it seems I may have been on the wrong 
track. When the milter jams, it seems that the parent process 
listening on the milter socket has died, along with its extant threads, 
but there may be one or more child processes in mutex_wait. So the 
problem is not that the processes are deadlocked, it's that the parent 
died. I'm still not sure why.



--
Andrew Daviel, TRIUMF, Canada
Tel. +1 (604) 222-7376  (Pacific Time)
Network Security Manager