[ClusterLabs] Howto restart resource

[Ken Gaillot]

On 08/29/2016 01:38 AM, Stefano Ruberti wrote:
> Dear all,
> 
> I have following situation and I need an advice from you:
> 
> in my Active/Passive Cluster  (Ubuntu_16.04  corosync + pacemaker , no pcs)
> 
> Node_A        Node_B
> Resource1    Resource1
> Resource2    Resource2
> Resource3    Resource3
> rsyslogd	     rsyslogd	
> 
> 1. is possible to configure rsyslogd to restarting when node (Node_A or Node_B) go on-line and take resource? 
> 
> 
> Best Regards
> Stefano

Is rsyslogd a cluster resource? That might not be ideal, since pacemaker
itself needs to use syslog (unless you've configured it not to), and you
probably want syslog running even if the cluster services are stopped
(for an upgrade, for example).

If your goal is simply to reload rsyslogd when a resource state changes,
I would think one of these approaches would be suitable:

Use the new alerts interface (pacemaker 1.1.15+) with an alert script
that reloads rsyslogd at the desired events;

Or, write a custom resource agent that reloads rsyslogd, and set up
colocation/ordering constraints between it and the desired resources.
Pacemaker will then "start" it after starting the resources, and "stop"
it before stopping the resources. Note that the agent must know whether
it is "running", and there is a ha_pseudo_resource() function in the OCF
libraries to handle that.