[Linux-HA] Debugging a single service (mysql)

[Dominik Klein]

RaSca wrote:
> Dominik Klein ha scritto:
>>> So, what can i do ?
>> Post logs and config ;)
>> Regards
>> Dominik
> 
> Sure.
> 
> ha.cf :
> 
> crm on
> use_logd on
> udpport 694
> bcast eth1
> keepalive 1 # 1 second
> deadtime 10
> initdead 80
> node mycluster-ha-1 mycluster-ha-2
> auto_failback yes
> 
> cib.xml is attached.
> 
> ha-debug which is attached contains the log of the two tries with the 
> native mysql script and the ocf script (with parameters).
> 
> Thank you so much.

For the LSB RA: It returns 254 on monitor and stop. That's pretty bad. 
After that, the cluster will not touch that resource again as it cannot 
identify the status of the resource. crm_mon should show "unmanaged" for 
that resource.

When you added the mysql_ocf resource, you did not specify any 
constraints with it. I cannot see in the logs on which node the resource 
was tried to start, but I guess drbd0 was not mounted to /store at that 
time causing the start to timeout. After a failed start, the resource 
will not be touched again, either. At least not on that node.

Colocate and order the mysql resource to your drbd device. See the 
http://wiki.linux-ha.org/DRBD/HowTov2 for details. That should make it work.

Regards
Dominik