[Pacemaker] Raid RA Changes to Enable ms configuration -- need some assistance plz.

[Andrew Beekhof]

On 10 Oct 2014, at 12:08 am, Errol Neal <eneal at businessgrade.com> wrote:

> Hi all. I was hoping to get some help with my configuration. I'm kinda 
> stuck at the moment. 
> 
> I've made some changes to the Raid1 RA to implement a ms style 
> configuration (you can see a diff here http://pastebin.com/Q2nbF6Rg 
> against the RA in the github repo)
> 
> I've modeled my ms implementation heavily on the SCST RA from the ESOS 
> project (thanks Marc!). 
> 
> Here is my full pacemaker config:
> 
> http://pastebin.com/jw6WTpZz
> 
> I'm running on CentOS 6.5 using Pacemaker, and CMAN available from the 
> distro.  
> 
> I'll explain a little bit about what I'm doing then tell you folks where 
> I think I need a bit of assistance or a push in the right direction
> 
> The Raid1 RA changes I made are because I want to assemble the same 
> RAID1 members on two different hosts. The slave should assemble it in 
> readonly mode. Using the SCST RA, I can offer two paths to the same LUN 
> from two distinct systems, one standby, one active. In order to keep 
> things good and consistent, the master of the MD resource also needs to 
> be master of the SCST resource (i'm using the example here 
> http://clusterlabs.org/doc/en-
> US/Pacemaker/1.1/html/Pacemaker_Explained/s-resource-sets-
> colocation.html).
> So if you can't already tell, this is YAA to create a highly available 
> open storage system :)
> 
> My understanding is that in order for N to start, N+1 must already be 
> running. So my configuration (to me) reads that the ms_md0 master 
> resource must be started and running before the ms_scst1 resource will 
> be started (as master) and these services will be force on the same 
> node. Please correct me if my understanding is incorrect.

I see only one ordering constraint, and thats between dlm_clone and clvm_clone.
Colocation != ordering.

> When both 
> nodes are up and running, the master roles are not split so I *think* my 
> configuration is being honored, which leads me to my next issue. 
> 
> In my modified RA, I'm not sure I understand how to promote/demote 
> properly. For example, when I put a node on standby, the remaining node 
> doesn't get promoted. I'm not sure why, so I'm asking the experts. 
> 
> I'd really appreciate any feedback, advice, etc you folks can give. 
> 
> Thanks, 
> 
> Errol Neal
> 
> 
> 
> 
> 
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 841 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20141013/1acc009f/attachment-0007.sig>