The mailing list for listmasters using Sympa

[Adam Bernstein <address@concealed>]

Thanks for the response, Olivier.

> The mhonarc process you see might the one run via 
> List::send_to_editor(), which is the only one that sympa.pl can run. The 
> goal is to generate an HTML version of the messages for later web 
> moderation. In you case, the mhonarc call is probably blocking. If you 
> can add a trace in List::send_to_editor() you might find out why. It is 
> likely that a .mhonarc.lck file is making problems in this list archives.

I'll look into that -- in fact, I meant to mention that when the mhonarc 
process hangs, it does seem to always be on a list that is moderated.  So 
there is some connection here.  However, I might need to clarify:

> If I understand you right, you have a sympa.pl hanging, probably waiting 
> for a mhonarc child process. Try killing this mhonarc process and see if 
> sympa.pl goes back to work. You probably need to remove a .mhonarc.lck 
> file to stop these problems.

I actually think the root cause is in the digest sending, and somehow that 
leads to an error condition that causes a problem with the archiving of 
moderated messages later.  Mhonarc and moderation always work right, even 
on the lists where are problem arises, except on days when there's an extra 
sympa.pl process already hung from an earlier digest send.  So the sequence is:

    1. Some days (but not all), digest sending hangs somehow, usually (but
       not always) on the same list.  An extra sympa.pl process is left
       running.
    2. All message processing and archiving works fine, until some time
       later when a message is sent to a different moderated list.  Then
       mhonarc hangs, and sympa.pl stops processing anything.  This only
       happens if #1 has happened earlier.

We'll keep investigating and either file a bug report or come back for more 
help in tracking it down.  Oh, and running in debug mode or foreground mode 
has proven to be impractical, since this problem arises at most once a day 
(early in the morning) and sometimes not for several days.  So 
troubleshooting this one is difficult.

      adam