Mailman problems after update

Message

libertytrek · Post by **libertytrek** » Sun Jan 06, 2008 6:05 pm

Ok, I've about run out of hair to pull out, so asking (again) for help...

I recently had some issues after updating Mailman from 2.1.9-r1 to 2.1.9-r2. I did the update because minor bumps like this aren't supposed to break anything, but of course, this isn't always the case.

Anyway, I sorted out the issues caused by the splitting of the app into two different directories (moved from /usr/local/mailman to /usr/lib64/mailman and /var/lib/mailman), but I haven't been able to get it to START reliably.

After the mailman update, mailman would give me an error [ !! ] when I tried to start it using the init script, but the qrunner was running, because the lists were working.

Well, I just updated the kernel (2.6.17-r4 to 2.6.23-r3) and about 20 other packages (none of which appear to be related to mailman in any way), and now I'm still getting the same error at boot and trying to start mailman with "/etc/init.d/mailman start", but now, the qrunner wasn't running. I can get it started manually by issuing (as root) "/usr/lib64/mailman/bin/mailmanctl -s start", but it doesn't survive a reboot. Not a BIG problem, because this server ran for about a year and a half since the last reboot - but I don't like it when things aren't working right.

One other thing I noticed...

When I issue the STOP command using the init script, I get this:

myhost init.d # /etc/init.d/mailman stop
* Caching service dependencies ...
* Can't find service 'postgresql' needed by 'pg_autovacuum'; continuing... [ ok ]
* WARNING: mailman has not yet been started.

What is with this mention of postgresql? This may be somehow related, because I was having trouble getting postgresql to recompile see this thread for details of that error.

Revdep-rebuild -p shows nothing, as does "emerge -pvuDN world"...

I sure hope someone has an idea or two...

libertytrek · Post by **libertytrek** » Tue Jan 08, 2008 12:44 pm

Ok, no one has an idea? I really need to get to the bottom of this... I'd appreciate any pointers...

libertytrek · Post by **libertytrek** » Wed Jan 16, 2008 11:11 am

Wow... no response at all...

Maybe we do need Daniel back.

kashani · Post by **kashani** » Wed Feb 06, 2008 12:13 am

You need to change the homedir of the mailman user so that the init.d script will start it from the right place.

sudo usermod -d /usr/lib64/mailman mailman

I had the other issue and it took me an hour to figure out all my data was in /var/lib/mailman/

kashani

libertytrek · Post by **libertytrek** » Wed Feb 06, 2008 5:03 pm

Hi Kashani,

Ok, thanks for the reply...

I ran the following command as root:

usermod -d /usr/lib64/mailman mailman

Attempted restart of mailman:

/etc/init.d/mailman restart

Same error:

* Restarting mailman ... [ !! ]

Did I miss something?

kashani · Post by **kashani** » Wed Feb 06, 2008 6:37 pm

I think I ran into some of the same issues and it's that the dumb mailman startup script.

1. Stop mailman using su - mailman -c 'bin/mailmanctl stop'
2. delete any left over pid files, qrunner.pid IIRC
3. /etc/init.d/mailman zap to clear whatever state Gentoo thinks mailman is in which is probably wrong.
4. /etc/ini.d/mailman start because if you use restart the script will fail because it can't find a pid file. It wasn't until I ran su - mailman -c 'bin/mailmanctl restart' that I noticed that.

Hope that works for you.

kashani

richard.scott · Post by **richard.scott** » Thu Feb 07, 2008 5:31 pm

libertytrek wrote:Wow... no response at all...

Maybe we do need Daniel back.

I've just been bitten by this bug. This really flippin sucks!

Don't get me wrong, I think Gentoo is the best OS out there. However, we need some formal QA for stuff like this!

Changing where your mailman lists are stored without telling anyone is just plain rude!

I've had to downgrade back to 2.1.9 until I can figure out what changes have happened.

Thank you dev's for the hard work, but please give us notification of such a change in future.

KCSB · Post by **KCSB** » Fri Feb 08, 2008 12:31 am

richard.scott wrote: I've had to downgrade back to 2.1.9 until I can figure out what changes have happened.

Might I ask (and please forgive me if this a beyond-noob question) how you downgraded?

For me, there a whole different set of things that have changed, as I am running sendmail with mailman. Usually, when there was an upgrade, I made a small modification to the Mailman ebuild in order to run mailman under the 'daemon' group and everything ran fine. Now, it seems the format of the ebuild has changed in addition to the home directory change.

Here's what the ebuild used to look like:

Code: Select all

INSTALLDIR="/usr/local/mailman"
APACHEGID="81"
MAILGID="2"

#if use postfix; then
 #   MAILGID="280"
#elif use sendmail; then
#    MAILGID=daemon
#elif use qmail; then
#    MAILGID="280"
#elif use courier; then
#    MAILGID=mail
#elif use exim; then
#    MAILGID=mail
#elif use xmail; then
#    MAILGID=xmail
#else
#    MAILGID="280"
#fi

I would simply comment out the everything and set the MAILGID="2" (the GID for daemon). For some reason, in spite of what the code above says, it never seemed to recognize sendmail as my mail server - but the change I made always seemed to work so I didn't really care.

Here how the new 2.1.9-r3 ebuild reads:

Code: Select all

pkg_setup() {
    INSTALLDIR=${MAILMAN_PREFIX:-"/usr/$(get_libdir)/mailman"}
    VAR_PREFIX=${MAILMAN_VAR_PREFIX:-"/var/lib/mailman"}
    CGIGID=${MAILMAN_CGIGID:-81}
    MAILUSR=${MAILMAN_MAILUSR:-mailman}
    MAILUID=${MAILMAN_MAILUID:-280}
    MAILGRP=${MAILMAN_MAILGRP:-mailman}
    MAILGID=${MAILMAN_MAILGID:-280}

    # Bug #58526: switch to enew{group,user}.
    # need to add mailman here for compile process.
    # Duplicated at pkg_postinst() for binary install.
    enewgroup ${MAILGRP} ${MAILGID}
    enewuser  ${MAILUSR} ${MAILUID} /bin/bash ${INSTALLDIR} mailman -G cron -c "mailman"
}

src_compile() {
    econf --without-permcheck \
        --prefix="${INSTALLDIR}" \
        --with-mail-gid=${MAILGID} \
        --with-cgi-gid=${CGIGID} \
        --with-cgi-ext="${MAILMAN_CGIEXT}" \
        --with-var-prefix="${VAR_PREFIX}" \
        --with-username=${MAILUSR} \
        --with-groupname=${MAILGRP} \

I've tried modifying the new ebuild with various settings for sendmail, but all attempts seem to break mailman even more.

Here's the change I try to make:

Code: Select all

pkg_setup() {
    INSTALLDIR=${MAILMAN_PREFIX:-"/usr/$(get_libdir)/mailman"}
    VAR_PREFIX=${MAILMAN_VAR_PREFIX:-"/var/lib/mailman"}
    CGIGID=${MAILMAN_CGIGID:-81}
    MAILUSR=${MAILMAN_MAILUSR:-mailman}
    MAILUID=${MAILMAN_MAILUID:-280}
    MAILGRP=${MAILMAN_MAILGRP:-daemon}
    MAILGID=${MAILMAN_MAILGID:-2}

And here's the reult when I send an e-mail to the test list.

Code: Select all

  ----- The following addresses had permanent fatal errors -----
"|/usr/lib/mailman/mail/mailman post test"
   (reason: 1)

  ----- Transcript of session follows -----
Traceback (most recent call last):
 File "/usr/lib/mailman/scripts/post", line 69, in ?
   main()
 File "/usr/lib/mailman/scripts/post", line 61, in main
   inq = get_switchboard(mm_cfg.INQUEUE_DIR)
 File "/var/tmp/portage/net-mail/mailman-2.1.9-r3/image//usr/lib/mailman/Mailman/Queue/sbcache.py", line 25, in get_switchboard
 File "/var/tmp/portage/net-mail/mailman-2.1.9-r3/image//usr/lib/mailman/Mailman/Queue/Switchboard.py", line 77, in __init__
OSError: [Errno 13] Permission denied: '/var/lib/mailman/qfiles/in'
554 5.3.0 unknown mailer error 1

So that's not right.

I should mention that I've tried setting the MAILGID to daemon in both the /etc/mailman/mm_cfg.py and the /usr/lib/mailman/Mailman/Defaults.py, and I still get the common 'group mismatch' error.

Even if I fix this, my lists will still be partially broken due to the directory change. At this point, it seems downgrading is the best option.

geyser · Post by **geyser** » Fri Feb 08, 2008 8:31 am

This update bit me also. I am still trying to get mailman configured to accept the gid that sendmail uses.
As mentioned in the previous post, the old methods of fixing the problem do not work with the recent ebuilds.

Just for the record, you will find an error like this in your syslog messages file:

Group mismatch error. Mailman expected the mail wrapper script to be executed as group "daemon", but
the system's mail server executed the mail script as group "mail". Try tweaking the mail server to run the
script as group "daemon", or re-run configure, providing the command line option `--with-mail-gid=mail'.

The fix is partly documented in the new ebuild:

Default-Configuration has changed deeply in 2.1.9-r2. You can configure
mailman with the following variables:
MAILMAN_PREFIX (default: /usr/$(get_libdir)/mailman)
MAILMAN_VAR_PREFIX (default: /var/lib/mailman)
MAILMAN_CGIGID (default: 81)
MAILMAN_CGIEXT (default: empty)
MAILMAN_MAILUSR (default: mailman)
MAILMAN_MAILUID (default: 280)
MAILMAN_MAILGRP (default: mailman)
MAILMAN_MAILGID (default: 280)

Unfortunately the file you need to change was not documented.
Can anyone drop a hint?
Thanks.

So all we have to do is set the MAILMAN_MAILxxx variable.

A few extra minutes thinking about emerge and ebuilds and all that I realized that
you set these variables on the command line with your emerge command. The error
indicated I needed to configure mailman to accept a connection from sendmail running
with the mail group. My solution was:

id mail
uid=8(mail) gid=12(mail) groups=12(mail)
MAILMAN_MAILGID=12 etry mailman > /tmp/mail.out 2>&1

If you look at the /tmp/mail.out file you will see this line:

./configure --prefix=/usr --host=i686-pc-linux-gnu --mandir=/usr/share/man --infodir=/usr/share/info --datadir=/usr/share --sysconfdir=/etc --localstatedir=/var/lib --without-permcheck --prefix=/usr/lib/mailman --with-mail-gid=12 --with-cgi-gid=81 --with-cgi-ext= --with-var-prefix=/var/lib/mailman --with-username=mailman --with-groupname=mailman --build=i686-pc-linux-gnu

You will notice that the option I needed is there: --with-mail-gid=12

Cheers!

richard.scott · Post by **richard.scott** » Fri Feb 08, 2008 9:50 am

KCSB wrote:Might I ask (and please forgive me if this a beyond-noob question) how you downgraded?

That's not a bad question so please don't worry. I once didn't know that either!

You need to be specific about the package your emerging as follows:

Code: Select all

emerge -av =net-mail/mailman-2.1.9

That will select only the 2.1.9 ebuild and not use any of the newer -r? type ebuilds.

Hope you get this fixed.

Rich.

richard.scott · Post by **richard.scott** » Fri Feb 08, 2008 9:57 am

I've posted my thoughts on the end of the mailman bug for this upgrade

http://bugs.gentoo.org/show_bug.cgi?id=208789

geyser · Post by **geyser** » Fri Feb 08, 2008 10:08 am

I am glad the mailman maintainers have updated the ebuild. The original location in /usr/local was awkward.
The move to /usr/lib and /var/lib has been painful. Besides the mailman/sendmail gid problem I mentioned above
I also had to change this symbolic link:

# pwd
/usr/adm/sm.bin
# ls -l
total 8
lrwxrwxrwx 1 root root 29 Feb 7 23:03 mailman -> /usr/lib/mailman/mail/mailman

My /etc/apache2/modules.d/50_mailman file only seemed to get updated the second time I reinstalled mailman but I might have forgotten to run etc-update.
Until this was fixed and apache restarted the web interface did not work.

Access to the public archives was also a problem. I had to move my data files, as described in earlier posts, but I also had to recreate the symbolic links in the public Archives area.

# pwd
/var/lib/mailman/archives/public
# ls -l
lrwxrwxrwx 1 root mailman 40 Feb 7 01:59 list_name -> /var/lib/mailman/archives/private/list_name

KCSB · Post by **KCSB** » Fri Feb 08, 2008 5:45 pm

richard.scott wrote:
KCSB wrote:Might I ask (and please forgive me if this a beyond-noob question) how you downgraded?
That's not a bad question so please don't worry. I once didn't know that either!

Thanks very much, Rich.

I managed to downgrade and get my mailing lists up and running again. My init.d script is now broken, but a manual start with mailmanctl seems to get things going. Ugh.

One thing I should probably mention is that none of my lists were available after the upgrade. So it looked like, in addition to dealing with the group wrapper and the new directory issues, I was going to have to start over with all of my lists. That would've been.... very bad.

Was this true for anyone else?

kashani · Post by **kashani** » Fri Feb 08, 2008 5:48 pm

Lots of the same issues with Postfix. I had done the aliases integration bit and the ebuild does not warn you that they have moved /usr/local/mailman/data/aliases into /usr/lib/.. which took me a bit of time to figure out. Postfix was hardcoded to the original spot and had to be changed and restarted.

kashani

arnuschky · Post by **arnuschky** » Wed Feb 13, 2008 12:23 pm

Hey all,

you have to change the mailman homedir in /etc/passwd from

Code: Select all

mailman:x:286:286:mailman:/usr/local/mailman:/bin/bash

to

Code: Select all

mailman:x:286:286:mailman:/usr/lib/mailman:/bin/bash

Then the startup script can find the qrunner startup script.

Of course you should copy your data files from /usr/local/mailman to /var/lib/mailman too.

jugglerkris · Post by **jugglerkris** » Sun Feb 17, 2008 5:08 am

This isn't exactly how I wanted to spend my evening. I had a working mailman install using postfix. After the upgrade from 2.1.9 to 2.1.9-r3, my lists were all down. To get things working, this is what I've done so far.

Update the /etc/apache2/modules.d/50_mailman.conf file (use etc-update or dispatch-conf)

Change mailman's home directory.

Code: Select all

usermod -d /usr/lib/mailman mailman

Copy existing lists and archives from old home directory.

Code: Select all

cp -pr /uar/local/mailman/lists /var/lib/mailman
cp -pr /usr/local/mainman/archives /var/lib/mailman

Change references of /usr/local/mailman/data/aliases to /var/lib/mailman/data/aliases in /etc/postfix/main.cf.

Switch to the mailman user. (su - mailman)

Create aliases.

Code: Select all

/usr/lib/mailman/bin/genaliases

Load new crontab file.

Code: Select all

crontab /usr/lib/mailman/cron/crontab.in

Switch back to root and restart apache, postfix and mailman.

Code: Select all

/etc/init.d/apache2 restart
/etc/init.d/postfix restart
/etc/init.d/mailman restart

This seems to have gotten things back online for me, but I'm unsure if I'm missing anything else. It seems like there should have been better instructions on how to migrate an existing setup. I assume that I can eventually purge /usr/local/mailman. I'm going to wait and see if anything else is broken first though.

rek2 · Post by **rek2** » Wed Feb 20, 2008 5:41 am

What I dont understand is how a mayor distro like gentoo is screwing with people more and more..
if you see the number of my posts and the date I joined the forum you will see I been here for long and a gentoo lover and still..
but lately a lot of screw up things have happen, some of US runs gentoo on production servers and yes " don't update" but if there is a mayor
change there should not be NEVER a revision only change.. so lets say is half half fault but I learn my lesson are the gentoo developers going to learn theirs?
PLEASE DONT MAKE SUCH BIG CHANGES ON REVISION!!! or we going to have to move our production servers to Debian again.

b1f30 · Post by **b1f30** » Wed Feb 20, 2008 6:14 am

rek2 wrote:What I dont understand is how a mayor distro like gentoo is screwing with people more and more..
if you see the number of my posts and the date I joined the forum you will see I been here for long and a gentoo lover and still..
but lately a lot of screw up things have happen, some of US runs gentoo on production servers and yes " don't update" but if there is a mayor
change there should not be NEVER a revision only change.. so lets say is half half fault but I learn my lesson are the gentoo developers going to learn theirs?
PLEASE DONT MAKE SUCH BIG CHANGES ON REVISION!!! or we going to have to move our production servers to Debian again.

I'll still take a Gentoo break over a M$ break any day.

At least when a distro breaks, you can usually fix it within 24-48 hours.

rek2 · Post by **rek2** » Wed Feb 20, 2008 6:20 am

b1f30 wrote:
rek2 wrote:What I dont understand is how a mayor distro like gentoo is screwing with people more and more..
if you see the number of my posts and the date I joined the forum you will see I been here for long and a gentoo lover and still..
but lately a lot of screw up things have happen, some of US runs gentoo on production servers and yes " don't update" but if there is a mayor
change there should not be NEVER a revision only change.. so lets say is half half fault but I learn my lesson are the gentoo developers going to learn theirs?
PLEASE DONT MAKE SUCH BIG CHANGES ON REVISION!!! or we going to have to move our production servers to Debian again.
I'll still take a Gentoo break over a M$ break any day.

At least when a distro breaks, you can usually fix it within 24-48 hours.

I 100% agree! but this does not means we have to relax 100% I want gentoo to succeed this is why one have to be critic at times as well..

b1f30 · Post by **b1f30** » Wed Feb 20, 2008 6:34 am

Being critical of something you care about is perfectly natural.

Gentoo is the only distro aside of Debian, and maybe Net/Open/FreeBSD that I seem to be able to get anything done with these days.

:-p

digrouz · Post by **digrouz** » Mon Mar 03, 2008 1:33 pm

I've also add a "-D MAILMAN" option to "APACHE2_OPTS" in my /etc/conf.d/apache2

Tortue · Post by **Tortue** » Thu Apr 17, 2008 5:04 pm

Thanks jugglerkris, you solved my problem!

tam · Post by **tam** » Wed May 28, 2008 1:43 pm

geyser wrote:
MAILMAN_PREFIX (default: /usr/$(get_libdir)/mailman)
MAILMAN_VAR_PREFIX (default: /var/lib/mailman)
MAILMAN_CGIGID (default: 81)
MAILMAN_CGIEXT (default: empty)
MAILMAN_MAILUSR (default: mailman)
MAILMAN_MAILUID (default: 280)
MAILMAN_MAILGRP (default: mailman)
MAILMAN_MAILGID (default: 280)

Unfortunately the file you need to change was not documented.
Can anyone drop a hint?

I'd like to know how and where to set ther MAILMAN_* variables.

tam · Post by **tam** » Thu May 29, 2008 9:51 am

tam wrote:to know how and where to set ther MAILMAN_* variables.

It has to be done in /etc/make.conf

Timz · Post by **Timz** » Mon Jun 23, 2008 1:08 pm

jugglerkris, Thanks a lot for your post ! very usefull for my update from 2.1.9-r1 to 2.1.9-r3.

Mailman problems after update

Mailman problems after update

Re: Mailman problems after update

more mailman sendmail problems

more notes about teh mailman changes

Re: more mailman sendmail problems

Re: more mailman sendmail problems