[Pacemaker] MySQL HA-Cluster, error on move

[Devin Reade]

--On Monday, April 11, 2011 02:14:44 PM +0200 Patric Falinder
<patric.falinder at omg.nu> wrote:

> The problem I have is when I need to migrate/move the resources to the
> other node, or unmove it, I get this error message and mysqld won't
> start/move properly:
[snip]

> primitive mysqld lsb:mysql
> colocation mysql_on_drbd inf: mysql ms_drbd_mysql:Master
> order mysql_after_drbd inf: ms_drbd_mysql:promote mysql:start

Try ocf:heartbeat:mysql instead of lsb:mysql.  Also, have a look
at the colocation and order constraints, and adjusting your start/stop
timeouts for mysqld (compare with the values below).  The following
is a configuration that has been rock-solid for me (ymmv, especially
wrt timeouts -- depends on db workload):

primitive drbd_mysql ocf:linbit:drbd \
        params drbd_resource="mysql" \
        op monitor interval="30s" \
        op start interval="0" timeout="240s" \
        op stop interval="0" timeout="100s"
primitive fs_mysql ocf:heartbeat:Filesystem \
        params device="/dev/drbd/by-res/mysql" directory="/cluster/mysql"
fstype="ext3" \
        op monitor interval="20s" timeout="40s" \
        op start interval="0" timeout="60s" \
        op stop interval="0" timeout="60s"
primitive ip_mysql ocf:heartbeat:IPaddr2 \
        params ip="a.b.c.d"
primitive mysqld ocf:heartbeat:mysql \
        params binary="/usr/bin/mysqld_safe"
pid="/var/run/mysqld/mysqld.pid" datadir="/cluster/mysql/data"
test_passwd="password" \
        op start interval="0" timeout="120" \
        op stop interval="0" timeout="120" \
        op monitor interval="60" timeout="30"
group mysql fs_mysql ip_mysql mysqld
colocation mysql_on_drbd inf: mysql:Started ms_drbd_mysql:Master
order mysql_after_drbd inf: ms_drbd_mysql:promote mysql:start
symmetrical=true

> I also have one more question. When I run "move mysql dbcluster2" for
> example, I need to run "unmove mysql" otherwise it won't migrate back
> over if that server fails, right!? Then can I see somewhere if I need to
> run "unmove" or if it will migrate over as it should?

Correct.

If you want a daily (or whatever) reminder of anything that hasn't been
un-moved, check out my cluster-migration-check script available as part of
gno-cluster-tools at <ftp://ftp.gno.org//pub/tools/cluster-tools>.
Consider it as an early release pending additions, more docs, crypto sigs,
etc, but I've had it in production for a while and it seems to be fine.
Tested only on CentOS 5.x.

Devin