1

For environment and config pls check
HA pacemaker monitor_5000 on zabserver-a 'not running' (7): call=53, status=complete, last-rc-change='Mon Jul 13 07:51:32 2015', queued=0ms, exec=0ms

I am now adding a service which is controlled by supervisord So I am adding supervisord as a resource in pcs config

The problem is that in the passive node I can still start supervisord and it keeps running whereas I understand normal behavior is that pcs should shut it down.

Secondly, I need all services to failover to passive node even if one of them dies and I am getting scattered services on noda-a and node-b.

What is the best solution to solve this, and am I doing anything wrong in my set up?

Community
  • 1
John Test
  • 89
  • 1
  • 3
  • 14
  • "The problem is that in the passive node I can still start supervisord and it keeps running whereas I understand normal behavior is that pcs should shut it down." Could you clarify? What do you mean by "start supervisord"? Start it from the shell / init.d / something similar? Or via `pcs`? – gxx Nov 10 '15 at 20:17
  • @gf_ via init.d/service start. Actually I found out that is not normal behavior and that it is doable. So now I need to find a way to prevent it. – John Test Nov 10 '15 at 21:49
  • 1
    AFAIK, this _is_ "normal behavior", however, you wouldn't want to do this, if your resource is managed via `pcs`. Wouldn't it be sufficient, to disable the start of `supervisord` at boot time? Could you point me to some docs, stating that this isn't "normal behavior"? Regarding your second point: Did you had a look at [Colocating Sets of Resources](http://clusterlabs.org/doc/en-US/Pacemaker/1.1/html/Pacemaker_Explained/s-resource-sets-colocation.html) and/or [Groups - A Syntactic Shortcut](http://clusterlabs.org/doc/en-US/Pacemaker/1.1/html/Pacemaker_Explained/ch10.html#group-resources)? – gxx Nov 10 '15 at 22:18
  • @gf_ it was a typo. I meant to say it is normal behavior. Thanks for the links. Do these settings go in cluster.conf ? – John Test Nov 10 '15 at 23:15
  • 1
    As these settings work with your resources, one way to set these would be via a cluster shell, `pcs` in your case. Have a look into `pcs config show`: In there, you'll find a section about "Colocation Constraints" (so, colocations go in there). Actually, `pcs config show`of your cluster in your linked thread shows as well a group called `zabbix-cluster`, so this concept isn't completely unknown to you, right? – gxx Nov 10 '15 at 23:39
  • @gf_ no it isn't completely unknown :) but I learn as I go and some things I have done and some I haven't. thank you for your help! – John Test Nov 11 '15 at 18:46
  • Great, sounds good. So, you're fine for the moment? – gxx Nov 11 '15 at 21:26

0 Answers0