[ClusterLabs] disable failover

[Ken Gaillot]

On 10/01/2015 09:54 AM, Vijay Partha wrote:
> For example. Lets have a cluster of 2 nodes node A and node B. Say on node
> A i have resource A running. If node A goes down i dont want the resource A
> to start on node B.

I assume the goal is to do this temporarily, for example, to perform
some maintenance on resource A? (If not, why put it in HA in the first
place?)

You have a few options for temporary maintenance:

* You can make a particular resource or resources "unmanaged", which
means Pacemaker will no longer try to start or stop them. To do this,
set the resource's "is-managed" meta-attribute to false. You might also
want to disable any recurring monitor operations on them, by setting the
monitor operation's "enabled" option to false.

* You can put the entire cluster into maintenance mode, in which case
all resources are made unmanaged. To do this, set the "maintenance-mode"
cluster option to true.

You can start and stop services as desired at that point, however you
shouldn't move a service when it is unmanaged (i.e. start a service on a
different node than the cluster last thought it was on).

You can also put a node into standby mode to do maintenance on the node
itself (e.g. reboot for a kernel update), but that will move all
resources to the other node.

Of course, remember to undo those changes when done with maintenance,
and realize that Pacemaker may then decide to move resources around if
circumstances call for it.

> On Thu, Oct 1, 2015 at 8:18 PM, Andrei Borzenkov <arvidjaar at gmail.com>
> wrote:
> 
>> On Thu, Oct 1, 2015 at 5:30 PM, Vijay Partha <vijaysarathy94 at gmail.com>
>> wrote:
>>> Hi,
>>>
>>> I want to know how to disable failover. If a node undergoes a failover
>> the
>>> resources running on the node should not be started on the other node in
>> the
>>> cluster. How can this be achieved.
>>>
>>
>> What exactly "node undergoes failover" means? Nodes do not failover -
>> resources may fail over between nodes.