[DRBD-user] One-line doubt when clusterizing DRBD resources...

Giuseppe Ragusa giuseppe.ragusa at hotmail.com
Fri Jul 4 17:46:07 CEST 2014

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

> Date: Fri, 4 Jul 2014 14:18:48 +0200
> From: lars.ellenberg at linbit.com
> To: drbd-user at lists.linbit.com
> Subject: Re: [DRBD-user] One-line doubt when clusterizing DRBD resources...
> Note that in most parts of the world,
> "doubt" is NOT the same as "question"
> 	;-)

Apart from an absolute lack of English fluency on my part, it was a blatant attempt to look knowledgeable and attract someone (who really is) on a a supposedly subtle subject ;->

A very successful attempt, it seems :-))

> > Do they need to be halted on both nodes with "drbdadm down res_name"
> > before stopping drbd service and clusterizing them all?
> Not necessarily.
> Depends on what you want, what you expect, and what you do.
> If you have existing DRBD resource,
> which are in active use,
> but do not use a cluster manager yet,
> you now want to add a cluster manager (pacemaker),
> and you expect it to take over control, without interfering,
> then you should experiment with this in a test environment first.

Actually I was starting from scratch and testing I did (while reading tons of your posts etc. ;> ), but a different problem afflicted my tests (wrong fence-peer handler) and at one time I was afraid that this "dubious practice" could instead be part of the problem.

> What you can do is start to configure pacemaker in "maintenance-mode",
> and once you are positive that it is set up the way you want it,
> take it out of maintenance-mode.
> At which point it will "reprobe" the state of the world (ok, this
> cluster), and if it finds all resources already active and in line with
> the configured policies, it will not take any action.

Really brilliant!
Many thanks for this suggestion: I will surely consider this strategy from now on, even when starting from scratch.

What I actually did was instead (I only have DRBD-backed KVM resources on CentOS 6.5):

*) begin with cluster (CMAN+Pacemaker) stopped/unconfigured on both nodes
*) manually start DRBD service on both nodes
*) create-md and up the resources on both nodes
*) make a resource primary on a selected node
*) virt-install/test there then shut down VM
*) make the resource secondary on the above-selected node
*) down the resource on both nodes
*) repeat for all resources
*) manually stop DRBD service on both nodes
*) start/configure cluster on both nodes
*) batch-define resources (pcs -f resource_cfg...) and test them "live" one at a time

On the "doubt" itself I know understand (from your answer and from tests, other problems corrected) that it is not necessary to totally quiesce DRBD (if it finds a consistent status when evaluating the clustered resources).

Many thanks again.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20140704/c06f32c6/attachment.htm>

More information about the drbd-user mailing list