[Mailman-Users] Further to "Need Help with Mailman Mail Delivery"

[Mark Sapiro]

On 09/27/2016 06:55 AM, Chuck Weinstock wrote:
> Whoops. The reinstalled Mailman stopped working with the same problem
> overnight. Two of the eight qrunners crashed.
> 
> I have 3-4 lists and one of them will not open in the web admin
> interface. It times out as per the apache log:
> 
> [Tue Sep 27 09:45:53.591373 2016] [cgi:warn] [pid 2483] [client
> 128.237.211.152:49581] AH01220: Timeout waiting for output from CGI
> script /usr/lib/mailman/cgi-bin/admin, referer:
> http://www.conjel.co/mailman/admin/fttc
> [Tue Sep 27 09:45:53.592426 2016] [cgi:error] [pid 2483] [client
> 128.237.211.152:49581] Script timed out before returning headers: admin,
> referer: http://www.conjel.co/mailman/admin/fttc
> [Tue Sep 27 09:46:53.639699 2016] [cgi:warn] [pid 2483] [client
> 128.237.211.152:49581] AH01220: Timeout waiting for output from CGI
> script /usr/lib/mailman/cgi-bin/admin, referer:
> http://www.conjel.co/mailman/admin/fttc
> [Tue Sep 27 09:46:53.640524 2016] [reqtimeout:info] [pid 2483] [client
> 128.237.211.152:49581] AH01382: Request body read timeout


The CGIs are timing out. This is normally caused by a locked list.


> Here is the access log from the same time frame:
> 
> 128.237.211.152 - - [27/Sep/2016:09:44:51 -0400] "GET
> /mailman/admin/fttc HTTP/1.1" 200 2078
> 128.237.211.152 - - [27/Sep/2016:09:44:53 -0400] "POST
> /mailman/admin/fttc HTTP/1.1" 504 247
> 
> Here is the qrunner log (from earlier when the two qrunners stopped):
> 
> Sep 27 06:09:59 2016 (7136) Master qrunner detected subprocess exit
> (pid: 1194, sig: 9, sts: None, class: VirginRunner, slice: 1/1) [restarting]

sig: 9 is a SIGKILL. This seems to say that something external is
killing the runner.

This is likely the same or a similar underlying cause as the CGI
timeouts, but is different as the CGIs are independent of the qrunners.


> 
> Finally this is the only error in the Mailman error file since the
> reinstall last night.
> 
> Sep 26 20:59:51 2016 admin(8885):
> @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 
> admin(8885): [----- Mailman Version: 2.1.15 -----] 
> admin(8885): [----- Traceback ------] 
> admin(8885): Traceback (most recent call last):
> admin(8885):   File "/usr/lib/mailman/scripts/driver", line 112, in run_main
> admin(8885):     main()
> admin(8885):   File "/usr/lib/mailman/Mailman/Cgi/admindb.py", line 198,
> in main
> admin(8885):     mlist.Save()
> admin(8885):   File "/usr/lib/mailman/Mailman/MailList.py", line 578, in
> Save
> admin(8885):     self.__save(dict)
> admin(8885):   File "/usr/lib/mailman/Mailman/MailList.py", line 555, in
> __save
> admin(8885):     os.link(fname, fname_last)
> admin(8885): OSError: [Errno 1] Operation not permitted


This is a permission or security manager (SELinux, apparmor, ?) issue.

First try running Mailman's 'bin/check_perms -f` as root. If that fixes
things, it may help. Also, see <https://wiki.list.org/x/17891756>.

Note that Mailman's CGI wrappers must be group mailman and SETGID. In
particular, these files must not be on a file system mounted with 'nosuid'.

If none of this helps, try disabling SELinux.

The qrunners being SIGKILLed is still a bit mysterious, but that could
be related to a permissions or SELinux issue.

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