[Pacemaker] Build dlm_controld for pacemaker stack (dlm_controld.pcmk)

[Bernardo Cabezas Serra]

El 2012-10-30 02:14, Andrew Beekhof escribió:

> Have you read Section 8.1 and 8.2 of Clusters from Scratch?
>
http://www.clusterlabs.org/doc/en-US/Pacemaker/1.1-plugin/html/Clusters_from_Scratch/ch08.html#id985586tl:dr
> - its very important that all parts of the stack are making decisions
> based on the same membership and quorum data.
>
> There were/are three options to achieve this:
> 1. everyone talks to pacemaker
> 2. everyone talks to cman
> 3. everyone talks to corosync
>
> [...]
>
> Option 3 is where things are headed, however the only distro that 
> ships
> it today is Fedora-17 (and shortly 18).
> In this scenario, all components obtain membership and quorum 
> directly
> from corosync.
> So far OCFS2 is the only component that hasn't been updated to 
> support
> this.
> This option requires corosync 2.x + pacemaker 1.1.7 or later
>
> Does that help?

Yes, thanks a lot, now I have things muuuch more clear.

I have a working ocfs2 active/active cluster with option (2), on top of 
ubuntu 12.04:
https://wiki.ubuntu.com/ClusterStack/Precise
Works pretty well, but I had some issues with fenced daemon, so I 
wanted to try the "head" option, but I misunderstood differences between 
option (1) and (3).

Just for information if someone tries it, ubuntu 12.04 LTS uses 
cluster-glue (1.0.8), openais (1.1.4), corosync (1.4.2), pacemaker 
(1.1.6) on top of cman (3.1.7), and seems also a good distro for option 
(2).
I think my option will be to solve current problems and stay on (2) on 
top of ubuntu.


However, as it is a new cluster that has to have support for years, I 
will spend some more time in trying to make option (3), work  building 
from sources on ubuntu (maybe trying FC17, and maybe switching to GFS2. 
Will share if success.

Seems that clusterlabs build guide 
(http://www.clusterlabs.org/wiki/Install#From_Source), does not 
contemplate the dlm part, and that confused me. Will try now with things 
more clear, and comments from Vladislav post about dlm4.


>> (2) What is the future roadmap about this?> Will future corosync2.0 
>> cluster have dlm issues addressed?
>The dlm works fine with corosync 2.0 - otherwise GFS2 wouldn't work (I 
> can assure you it does).

OK
Thanks again.

Bernardo.