[lucid-proposed] post-start script broken

[Jürgen Kreileder]

5.1.41-3ubuntu12.4 from lucid-proposed has a broken post-start script.

After starting mysql, /var/log/syslog says:
init: mysql post-start process (11493) terminated with status 1

Looking at the post-start script shows two problems:

1) ret = $?
Note the spaces before and after the '='. There should be none!
As it is now, it tries to run a command called "ret" instead of setting a variable.

2) /usr/bin/mysqladmin --defaults-file=$HOME/debian.cnf ping
If this returns a non-zero exit status (most likely because mysql isn't fully up yet), the post-script exits immediately!
You need a "set +e" or something else that keeps the post-start script from exiting prematurely if the command returns with something != 0.

[hyper_ch (bugs-launchpad-net-roleplayer)]

Well, since yesterday mysql won't be properly started anymore at boot-up. Grepping the syslog only returns this:

Jul 23 10:27:28 kubi kernel: [ 105.206415] type=1505 audit(1279873648.967:10): operation="profile_load" pid=914 name="/usr/sbin/mysqld"
Jul 23 10:27:29 kubi kernel: [ 105.281964] type=1505 audit(1279873649.044:11): operation="profile_load" pid=979 name="/usr/sbin/mysqld-akonadi"

I don't get the error that the thread opener gets but in effect, mysql is not being started at boot up.

[Neil Wallace (rowinggolfer)]

+1.

please do not push this up to the stable repos.

[Richard Hansen (rhansen)]

I have created a PPA for this bug and submitted a fix:

https://edge.launchpad.net/~a7x/+archive/bug608423

It'll take a while before it's done building.

[Richard Hansen (rhansen)]

Nominating for Lucid according to the instructions at https://wiki.ubuntu.com/StableReleaseUpdates#Verification

[Richard Hansen (rhansen)]

I've updated the debdiff to fix another bug introduced in 5.1.41-3ubuntu12.4: mysql wasn't starting at boot (see comment #1) because of a missing space.

[Martin Pitt (pitti)]

Chuck, can you please reupload with -v to include the previous lucid-proposed upload?

I see that you removed the "echo timeout", which was discussed in Prague. Thanks for fixing that.

I also see

@@ -30,15 +30,8 @@
     for i in `seq 1 30` ; do
- /usr/bin/mysqladmin --defaults-file=$HOME/debian.cnf ping
- ret = $?
- if [ $ret -eq 0 ] ; then
- break
- fi
+ /usr/bin/mysqladmin --defaults-file=$HOME/debian.cnf ping || ret=$?
                sleep 1

This removes the break, but I don't see why? If we don't break the loop on success, then it will always keep trying 30 times?

[Christoph (christoph-thomas)]

Hi Chuck,
the same bug is in Maverick mysql-server 5.1.48-1ubuntu3. The patchworks flawless.
Thanks, Christoph

[Richard Hansen (rhansen)]

I have updated my debdiff to not echo anything if mysqld times out. The updated package should show up in my PPA [1] in a few hours.

Chuck's latest upload [2], currently awaiting approval in the upload queue [3], still has a couple of issues:
 1. it does not fix the missing space issue in the "start on" clause, so mysql won't start on boot
 2. it doesn't sleep between pings

Is there a reason why my patch is being ignored?

[1] https://edge.launchpad.net/~a7x/+archive/bug608423
[2] debdiff: http://launchpadlibrarian.net/52600418/mysql-dfsg-5.1_5.1.41-3ubuntu12.4_5.1.41-3ubuntu12.5.diff.gz
[3] https://edge.launchpad.net/ubuntu/lucid/+queue?queue_state=1&queue_text=mysql-dfsg-5.1

[Richard Hansen (rhansen)]

There's a third (minor) issue with the latest upload:

    3. If exec fails for some reason, the script will keep attempting to exec debian-start when it should simply fail.

[Martin Pitt (pitti)]

Chuck, please fix mysql-5.1 in maverick as well, it has the same problem.

[Martin Pitt (pitti)]

Accepted mysql-dfsg-5.1 into lucid-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Martin Pitt (pitti)]

Since this regression only affected lucid-proposed, I close the lucid task.

[Martin Pitt (pitti)]

Fixed in maverick's mysql-5.1 5.1.48-1ubuntu4.

[Jürgen Kreileder (jk)]

I agree with a7x. This is not fixed at all: Neither in lucid-proposed (a7x's points 1, 2, and 3) nor in maverick (just point 2, 3 as the new maverick package does have a 'sleep 1').

For the record, I did try the lucid-proposed package nevertheless. syslog still reports 'mysql post-start process (...) terminated with status 1'.

[ideasbuenas (ideasbuenas)]

unsuscribe ideasbuenas <email address hidden>

On Wed, Jul 28, 2010 at 3:19 PM, Jürgen Kreileder <email address hidden> wrote:

> I agree with a7x. This is not fixed at all: Neither in lucid-proposed
> (a7x's points 1, 2, and 3) nor in maverick (just point 2, 3 as the new
> maverick package does have a 'sleep 1').
>
> For the record, I did try the lucid-proposed package nevertheless.
> syslog still reports 'mysql post-start process (...) terminated with
> status 1'.
>
> ** Tags added: verification-failed
> ** Tags removed: verification-needed
>
> ** Changed in: mysql-dfsg-5.1 (Ubuntu Lucid)
> Status: Fix Released => In Progress
>
> ** Changed in: mysql-5.1 (Ubuntu Maverick)
> Status: Fix Released => In Progress
>
> --
> [lucid-proposed] post-start script broken
> https://bugs.launchpad.net/bugs/608423
> You received this bug notification because you are subscribed to Lucid.
>
> Status in “mysql-5.1” package in Ubuntu: In Progress
> Status in “mysql-dfsg-5.1” package in Ubuntu: Invalid
> Status in “mysql-5.1” source package in Lucid: Invalid
> Status in “mysql-dfsg-5.1” source package in Lucid: In Progress
> Status in “mysql-5.1” source package in Maverick: In Progress
> Status in “mysql-dfsg-5.1” source package in Maverick: Invalid
>
> Bug description:
> 5.1.41-3ubuntu12.4 from lucid-proposed has a broken post-start script.
>
> After starting mysql, /var/log/syslog says:
> init: mysql post-start process (11493) terminated with status 1
>
> Looking at the post-start script shows two problems:
>
> 1) ret = $?
> Note the spaces before and after the '='. There should be none!
> As it is now, it tries to run a command called "ret" instead of setting a
> variable.
>
> 2) /usr/bin/mysqladmin --defaults-file=$HOME/debian.cnf ping
> If this returns a non-zero exit status (most likely because mysql isn't
> fully up yet), the post-script exits immediately!
> You need a "set +e" or something else that keeps the post-start script from
> exiting prematurely if the command returns with something != 0.
>
>
>

[Richard Hansen (rhansen)]

See bug #611101 for the missing "sleep 1" issue.

[Richard Hansen (rhansen)]

See bug #611102 for the missing space preventing startup at boot issue.

[Martin Pitt (pitti)]

Accepted mysql-dfsg-5.1 into lucid-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[DAP (akadap)]

Latest patch also does not work for me. mysql is not started on boot.

[DAP (akadap)]

(my last comment was nearly simultaneous with Martin Pitt's announcement, so it refers to the previous version.)

The latest version appears to work. (mysql is running after a reboot, did not check for spurious entries in the logs).

[Richard Hansen (rhansen)]

5.1.41-3ubuntu12.6 works for me.

[Martin Pitt (pitti)]

Thanks for testing! Closing lucid task, since the regression was only in -proposed.

[Christoph (christoph-thomas)]

Hi,
maverick's mysql-5.1 5.1.48-1ubuntu4 does'n work for me. Might be the missing space "runlevel[2345]" "runlevel [2345]" problem.

[Launchpad Janitor (janitor)]

This bug was fixed in the package mysql-5.1 - 5.1.48-1ubuntu5

---------------
mysql-5.1 (5.1.48-1ubuntu5) maverick; urgency=low

  * debian/mysql-server-5.1.mysql.upstart: Fix upstart init script. (LP: #608423)
 -- Chuck Short <email address hidden> Sun, 01 Aug 2010 11:23:11 -0400

[Christoph (christoph-thomas)]

Hi Chuck,
mysql-5.1 (5.1.48-1ubuntu5) maverick works for me.
Thanks for this fix,
Christoph