you can simply do a ‘clrs disable application-rs’. With this command the application will be stopped and stay offline.
Now you can do your maintenance. Which means you can start your application manually without cluster control. When you have finished your maintenance stop your application manually.
Then start up the application again with cluster control by using ‘clrs enable application-rs’.
Hi Juergen Thanks for the reply
Can unmonitoring the resource be used as well. Just wanted to get clarity on what is the difference between disabling a resource /unmonitoring a resource/offlining a resource
As per my understanding unmonitoring a resource will keep the resource running but will not be monitored by the cluster and then can be manually stopped and started till the monitor is turned back on again
yes, ‘clrs unmonitor’ and ‘clrs monitor’ is another option. In this case only the monitor for the resource will be stopped. This can also be enough to do the maintenance. But if somebody switchover/failover the resource group which include the unmonitored resource then the resource will also switch to the other node. So, it depends on your maintenance and what you expect.
when doing planned maintenance (like applying a patch or upgrading the application software) for an application under cluster control, the safest way is to:
- disable the cluster resource (clrs disable <resourcename>)
- perform the maintenance
- verify everything is working as expected
- enable the cluster resource again (clrs enable <resourcename>)
There is also the possibility to suspend the whole resource group, where the application resource is configured in, by doing a "clrg suspend <resourcegroupname".
Then the cluster framework will not try to restart this resource group or trigger failover for it.
The resource group can be resumed by "clrg resume <resourcegroupname>".
But here it really depends on the scope of change you apply to an application, if the data service monitor for the application resource can really continue to perform monitoring as before.
In some cases the data service monitor is not only probe based, but does also perform PMF based monitoring, ie. it monitors the specific process ids (pids) that got started by its start method.
If you then manually stop the application, PMF might no longer see any of those pids, and thus determine a failed resource.
That is why in general for performing maintenance on an application, you should start with disabling the resource. And enable it back once the maintenance is finished.
Thanks for teh detailed reply
I have disabled a resource with clrs disable resource
Now the status for the resource is showing offline on both nodes
If I enable the resource now where will it start the application running under this resource and show online as the application running under the resource has to come up on the same node where the other resources for this application are running
If you just disabled the specific application resource, and left other resources enabled and the corresponding resource group online, then if you enable the application resource, it will be started on the node where the resource group is currently online.
It is always the resource group that will determine where the resources configured for this resource group switch online.