[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ProgSoc] It's all Victor's fault (Was: Victor's repeated email)



On Thu, 2006-10-26 at 00:58 +1000, jedd wrote:

>  Here, let me share some depressing insight with you.
> 
>  /var/log/exim4/mainlog snippet :
> 
>  2006-10-25 11:17:00 1GcXKF-0002Ld-QR SA: Action: spamd took more than 240 secs to run, accepting message (scanned in 0/0 secs
> | Message-Id: 1GcXKF-0002Ld-QR). From <ca@xxxxxxxxxxx> (host=mail.secunia.com [213.150.41.240]) for jedd@xxxxxxxxxxx
>  2006-10-25 11:17:00 1GcXKF-0002Ld-QR SA: Debug: Writing message to /var/spool/sa-exim/SAtimeoutsave/new/1161738780_1GcXKF-0002
> Ld-QR
>  2006-10-25 11:17:00 1GcXKF-0002Ld-QR <= ca@xxxxxxxxxxx H=mail.secunia.com (secunia.com) [213.150.41.240] P=smtp S=2660 id=2006
> 1024211703.6953.qmail@xxxxxxxxxxx
>  2006-10-25 11:17:00 1GcXKF-0002Ld-QR => jedd <jedd@xxxxxxxxxxx> R=local_user_plus T=mail_spool
> 
>  There's probably some irony in a secunia message being so delayed,
>  but it's lost on me.  It's vaguely interesting the second one goes
>  through immediately.

This looks to me as though it may be just one message, in particular the
words "accepting message" which suggests that, once the timeout is hit,
the message is accepted anyway.

This would certainly explain the duplicates, if the remote MTAs were
timing out in less than 240s.

Has there been a recent surge in SA timeouts? Does it correspond,
roughly, with the appearance of the duplicates? (Does each timeout have
a successful delivery immediately afterwards with the same transaction
ID, same host, same sender and same recipient?)

>  Curiously, /var/spool/sa-exim/SAtimeoutsave   contains messages
>  dating back to January .. maybe that's worth looking at again when
>  the Le Panto has worn off.  <ponder pulp>  | wc indicates 3500
>  messages archived in there - and 47k files (presumably this maps
>  to messages) under /var/spool/sa-exim.  This probably isn't healthy.

Hmm. Perhaps the former was an early debugging tool for whoever
implemented it on ProgSoc? (The latter may include a corpus for a Bayes
filter.)

- Raz


-
You are subscribed to the progsoc mailing list. To unsubscribe, send a
message containing "unsubscribe" to progsoc-request@xxxxxxxxxxxxxxxxxxx
If you are having trouble, ask owner-progsoc@xxxxxxxxxxxxxxxxxx for help.