[DRBD-user] 3-Node DRBD Without Stacking
pavlos.parissis at gmail.com
Wed Nov 17 14:24:34 CET 2010
On 16 November 2010 18:45, Felix Frank <ff at mpexnet.de> wrote:
> On 11/16/2010 05:55 PM, Robinson, Eric wrote:
>>>> Is it possible to deploy a 3-node CRM-based cluster where:
>>>> -- nodes A and C share resource R1 on /dev/drbd0
>>>> -- nodes B and C share resource R2 on /dev/drbd1
>>>> -- resource constraints prevent R1 from running on node B and
>>>> resource R2 from running on node A?
>>> I don't think I understand what you're trying to do.
>>> Please elaborate.
>>> Why would you need to prevent R1 from "running" on
>>> Host B if the resource is shared between A and C?
>>> B is not at all configured to connect to R1, is it?
>> It's a three-node CRM cluster where node C acts as a failover for nodes
>> A and B, but nodes A and B do not fail over to each other. However, in a
>> three-node CRM cluster, all resources are allowed to run on all nodes by
>> default. For example, even though resource R1 only physically exists on
>> nodes A and C, if node A fails Pacemaker might try to start it on node B
>> if we don't put in some kind of constraints. You follow?
> I think I do now.
> The ocf:linbit:drbd RA should be smart enough to realize that R1 is not
> configured (in drbd.conf) to ever run on R1. But don't quote me on that.
Nope, there is not such intelligence in the RC because PE takes this
kind of decisions.
> If you are still afraid it might take a nonsensical action and bring a
> DRBD device up on a node that is not configured to do that, feel free to
> add a constraint like
> location p_drbd_r1_not_on_hostB -inf: p_drbd_r1 hostB
> or something along those lines.
> drbd-user mailing list
> drbd-user at lists.linbit.com
More information about the drbd-user