Hi all,
We are seeing a problem with procmail under IRIX 6.2. Many of our users
run procmail from their .forward file. Since upgrading to IRIX 6.2, we
have started seeing instances where the sendmail daemon and procmail program
appear to hang when attempting to deliver messages. This problem doesn't
happen all the time, and affects different users at different times.
When it does occur for a given user, it can disable our mail server pretty
quickly because mail for the user backs up and generates a number of
sendmails on the system. You eventually hit your sendmail system load
settings which causes the system to stop accepting new mail connections.
One user that experienced the problem recently had 20+ messages in the
mail queue that were attempting to write the messages to his NFS mounted
home directory.
The receipe that he had in his .procmailrc file was as follows:
:0H
* (^Subject: *Status of SDRC PIR.*)|(^Subject: *TEST Status of SDRC PIR.*)
$HOME/nsmail/PIR_Stuff
He wasn't specifying a "local lockfile" for this receipe.
The procmail process would simply hang on the system. While this process was
hung, you could not touch the $HOME/nsmail directory with any command. If
you did an "ls -l $HOME/nsmail", your process would hang. The hung processes
could not be killed.
It appears that it is related to file locking on NFS disks.
Here is our version output from procmail:
# procmail -v
procmail v3.11pre4 1995/10/29 written and created by Stephen R. van den Berg
Locking strategies: dotlocking
Default rcfile: $HOME/.procmailrc
System mailbox: /var/mail/$LOGNAME
How does procmail handle file locks if no trailing ":" is specified for a
receipe?
What's the current release of procmail?
Has anybody else experienced problems related to procmail on IRIX 6.2?
Note: we have seen a the same problem in some cases with users running the
elm "filter" program as well.
We are trying to understand how these applications are doing file locking.
SGI has pretty much dismissed this as a procmail issue so I'm trying to
understand how locking is being done through procmail. I believe it is the
old statd/lockd NFS related locking problems commonly seen with mail related
applications, but have no way of proving it. I will use the VERBOSE setting
the next time the problem occurs, but was hoping someone could shed some
light on how file locking is done with procmail.
The problem typically clears itself up after several hours of waiting. The
procmail will eventually get the lock that it needed. All the mail will be
delivered. The windows that were hung up on the system due to accessing the
affected directory all free up. Everything operates okay until the next
instance. We have seen this 3 or 4 times since upgrading 450 systems to IRIX
6.2.
Note: we also had to rebuild numerous applications like procmail due to the
executable incompatiablity problem. So the OS was upgraded and procmail
was rebuilt.