[Mailman-Users] mailman locking up with "too many files open"

[Mark Sapiro]

Ian Johnson wrote:
>
>I've been running a mailman list (for broadcast only) lately, and 
>recently had a post that jammed up the system. Mailman would respond to 
>other list requests, but the admin pages for the list in question would 
>not appear, and the list itself was unresponsive. This is the error 
>message that I see throughout the error log:
>
>Jun 12 20:15:28 2006 qrunner(8679): libkrb5.so.3: cannot open shared 
>object file: Too many open files in system
>Jun 12 20:15:31 2006 qrunner(8769): Traceback (most recent call last):
>Jun 12 20:15:31 2006 qrunner(8769):   File "/var/mailman/bin/qrunner", 
>line 270, in ?
>Jun 12 20:15:31 2006 qrunner(8769):   File "/var/mailman/bin/qrunner", 
>line 230, in main
>Jun 12 20:15:31 2006 qrunner(8769):   File 
>"/var/mailman/Mailman/Queue/Runner.py", line 70, in run
>Jun 12 20:15:31 2006 qrunner(8769):   File 
>"/var/mailman/Mailman/Queue/Runner.py", line 94, in _oneloop
>Jun 12 20:15:31 2006 qrunner(8769):   File 
>"/var/mailman/Mailman/Queue/Switchboard.py", line 158, in files
>Jun 12 20:15:31 2006 qrunner(8769): OSError :  [Errno 23] Too many open 
>files in system: '/var/mailman/qfiles/news'


It seems like there are missing lines in the above trace. Did you grep
on date?

Anyway, Switchboard is trying to obtain the contents of the 'news'
queue when the error occurs (this time - is it always this?).
Presumably, this is NewsRunner. It seems highly unlikely that this
specific occurrence has anything to do with a specific list.


>Three changes may have contributed to this problem:
>
>1. list went from 4000 to 6800 members
>2. the email sent out had a somewhat large attachement (~300k)
>3. I recently turned on footer customization to allow user-specific 
>unsubscribe instructions at the bottom.


None of these is likely to be the immediate cause.


>It seems like I should either:
>    1. increase max-open-file limit on system


What is the limit now?


>   or
>    2. turn off footer customization (if this would, in fact, solve the 
>problem)


I doubt this would help.

Have you tried 'mailmanctl restart'? How many runners are running? Do
you need to "clean up" from prior multiple starts? See
<http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq04.068.htp>.

-- 
Mark Sapiro <msapiro at value.net>       The highway is for gamblers,
San Francisco Bay Area, California    better use your sense - B. Dylan