[Pacemaker] Slave didn't become master after master is down

[Digimer]

On 27/02/14 04:58 AM, 邓尧 wrote:
> I followed this manual:
> http://clusterlabs.org/doc/en-US/Pacemaker/1.1-pcs/html/Clusters_from_Scratch/index.html,
> except the pcsd installation, as it's unavailable on my CentOS 6.5 box.
> I also installed cman, because without it, command "pcs cluster start"
> would fail. Two resources were added to the cluster: a DRBD device and a
> filesystem resource, the resource setup steps are the same as written in
> the manual.
>
> Property "stonith-enabled" was set to false, "no-quorum-policy" was set
> to ignore.
>
> Everything seems to be fine until I tried to test failover:
> If I "standby" the master, the slave became master for a very short
> while (about one or two seconds), then it become slave, however after I
> unstandby the original master, the slave then became a master, and the
> original master became slave. The behaviour is illustrated as the graph:
>
> node0(master)  node1(slave)  <== the original state.
> ====================
> node0(slave)  node1(master)   <== after command "pcs cluster standby node0"
> =====================
> node0(stopped) node1(slave)  <== after a short while
> ====================
> node0(slave) node1(master)  <== after command "pcs cluster unstandby node0"
>
> I also tried rebooting the master node, the slave only became master
> after the original master is successfully rebooted.
> What's most likely cause of my problem ?
>
> Thanks
> Deng Yao

I've not setup cman + pacemaker on EL6, but I can say that fencing in 
cman is required normally. Please configure fencing/stonith and see if 
it works (fence_pcmk in cman, real stonith in pacemaker). You will need 
this anyway, and it's best to configure it from the start.

-- 
Digimer
Papers and Projects: https://alteeve.ca/w/
What if the cure for cancer is trapped in the mind of a person without 
access to education?