The mailing list for listmasters using Sympa

[Omen Wild <address@concealed>]

Quoting David Verdin <address@concealed> on Wed, Nov 10 16:19:
>
> Hi and sorry for this late answer.

No problem.  Any and all help is greatly appreciated.

> First: You should not, in Sympa 6.x use two sympa.pl. The
> "distribution" process does nothing, all the operations it performed
> being done by bulk.pl now.
> We need to remove any reference to the previous behaviour.

I set "distribution_mode fork" on the recommendation from David Verdin
on August 27th to solve a problem we were having where digest
processing was taking 2 hours and no messages were processed during
that time.  "distribution_mode fork" solved that problem.

See my original email here
<https://listes.cru.fr/sympa/arc/sympa-users/2010-08/msg00083.html>
and David Verdin's response here
<https://listes.cru.fr/sympa/arc/sympa-users/2010-08/msg00086.html>.

> Second: the crashes. Sympa should not be wanting to make any fork: it
> is running already. What was Sympa doing at the time it crashed?

Trying to load a large (90MB) message from the incoming queue
(/var/sympa/spool/msg) for processing.  All of our lists are limited
to 25MB, so Sympa was just going to dump it anyhow.

> Even if you have plenty of RAM, the System may not let you fork,
> because it would make you exceed the number of processes you can own.
> Please have a look at this post about this mechanism in Solaris:
> http://blogs.sun.com/hasham/entry/solaris_10_increasing_number_of

bash-3.00$ ulimit -a
[ snip ]
max user processes            (-u) 29995

We have maybe 100 processes running under that user, so it's not that.

We had a case where Sympa went from 300MB of RAM to almost 3GB of RAM
on Sunday.  This was trying to process a 90MB message.  I have the
permission of the sender to upload it for analysis if anyone would
like to take a look at the message.  

At 3pm Sunday Sympa started logging:
Nov  7 15:01:08 penny-1 sympa(command)[24808]: [ID 702911 local1.warning] 
tools::safefork() Can't create new process in safefork: Not enough space
Nov  7 15:01:41 penny-1 last message repeated 2 times

This message (and no other logging messages) continued from
sympa(command) (see attached log snippet) every couple minutes until the
incoming spool backed up enough that our paging system caught it and I
restarted Sympa.  Usually when this happens Sympa dies and out monitoring
system catches that.  This was the first time sympa(command) got caught
for 5 hours trying to fork.

-- 
Omen Wild
Security Administrator
(530) 752-1700

Attachment: penny-1-sympa.log-20101107.bz2
Description: Binary data

Attachment: smime.p7s
Description: S/MIME cryptographic signature