MailScanner hanging (by Joseph Watson)

[Michael Janssen]

On Thu, 14 Aug 2003, Joseph Watson wrote:

> Hello,
> Just today I began having a problem with mailscanner which has been running
> untouched since I upgraded to 4.22-5 in July.
>
> Mandrake Linux 9.1
> Postfix
> MailScanner 4.22-5
> Command Antivirus
>
> For some reason MailScanner seems to be hanging on a batch of messages.  I see
> log entries that say it has started a new batch of 34 messages, but it never
> follows with a "message clean" notice or a delivery entry.  It just sits
> there and the messages are piling up.  There is no error or info messages
> being logged either.  It seems there may be a message that it is chocking on,
> but I don't know how to tell what is going on.
>
> Has anyone seen this before?  What can I do to trouble shot this thing.  Any
> suggestions would be welcome.  Thanks

You can easily take a look into what MailScanner is momentarily doing.
Just find the pid of the hanging process and fire up:

"strace -p pid" Most likly you will get something like: "wait4(12345, "
which means MS waits for the process with that pid. Strace reports the
system calls the process is doing. Please note, that strace might damage
processes under circumstances.

Another Ressource is the user defined output of ps and some fancy using
of top. wchan (Wait Channel (?) - the type of doing nothing) is of much
interest:

ps -C MailScanner o pid,ppid,wchan

Of course %cpu and %mem. etime (elapsed time) might be good to know
(especially if it's still increasing). Cumulative etime is nice but
AFAIK only revealed by top: button "S".

ps and top might be a bit boring, but strace is often the key to find
your answeres.

I don't know if a look into /proc/[pid]/fd might be helpful. MS workers
havn't got many file-descriptors (which can cause trubble when pointing
to deleted files) in normal operations.


Are all your MS processes hanging?

regards
Michael