[ClusterLabs] Antw: CRM location specification and errors

[Kristoffer Grönlund]

Andrei Borzenkov <arvidjaar at gmail.com> writes:

> В Fri, 26 Jun 2015 10:12:00 +0200
> Kristoffer Grönlund <kgronlund at suse.com> пишет:
>
>> Andrei Borzenkov <arvidjaar at gmail.com> writes:
>> 
>> > В Fri, 26 Jun 2015 08:28:14 +0200
>> > "Ulrich Windl" <Ulrich.Windl at rz.uni-regensburg.de> пишет:
>> >
>> >> Hi!
>> >> 
>> >> I guess the cluster is running monitor operations on the forbidden node to make sure no resources run there: Meaning: If you had started those resources on the forbidden node, the cluster would stop them. To find out it runs the monitors.
>> >
>> > Unfortunately it does not. And that was quite a nasty surprise for me
>> > when I started to use pacemaker. It does run monitor only once (probe)
>> > on initial startup. So no, in pacemaker there is no protection against
>> > accident manual starting of resources that I am aware of.
>> >
>> >
>> 
>> Just as with normal monitor operations, you have to configure it if you
>> want it. Pacemaker doesn't monitor resources on any nodes unless
>> configured to do so.
>> 
>> op monitor role=Stopped
>> 
>
> I do not see "role" mentioned anywhere on this page:
>
> http://clusterlabs.org/doc/en-US/Pacemaker/1.1-pcs/html/Pacemaker_Explained/_resource_operations.html
>
> How does it look like in CIB?
>

I'll avoid suggesting that the documentation needs to be updated since I
risk having to do it myself ;)

You can see an example of what it looks like in the CIB in this test
case:

https://github.com/ClusterLabs/pacemaker/blob/master/pengine/test10/stopped-monitor-31.xml

-- 
// Kristoffer Grönlund
// kgronlund at suse.com