[one-dev] announcing addon plan: shared-lvm-single-lock

[Nicolas AGIUS]

Hi,

Maybe, you should have a look at the CXM project [1] and the CXM driver [2].
It's a little bit old now, but it has been designed to work with cLVM on XEN hosts, and provide load-balancing and automatic failover.

It can also work without cLVM. In this setup, Opennebula is not part of the cLVM cluster, all Xen nodes are read-writing LVM metadatas. That's working good, but it's dangerous and slower.

Cheers,
Nicolas AGIUS

[1] https://github.com/nagius/cxm
[2] http://opennebula.org/software:ecosystem:cxm_drivers


--------------------------------------------
En date de : Mar 3.12.13, Jaime Melis <jmelis at opennebula.org> a écrit :

 Objet: Re: [one-dev] announcing addon plan: shared-lvm-single-lock
 À: "Mihály Héder" <mihaly.heder at sztaki.mta.hu>
 Cc: dev at lists.opennebula.org, cc at hbit.sztaki.hu
 Date: Mardi 3 décembre 2013, 12h24
 
 Hi Mihály,
 I'm sincerely looking forward to seeing this
 addon and I think it would help to many OpenNebula
 users.
 Do you have anything already done? Maybe I could
 take a quick look at it and give you some feedback?
 
 
 As a general recommendation I'd like to point
 out that it would be great if you could reuse code from
 other DS and TM drivers.
 cheers,
 Jaime
 
 
 
 
 On Wed, Nov 27, 2013
 at 11:17 PM, Mihály Héder <mihaly.heder at sztaki.mta.hu>
 wrote:
 
 
 Hi, 
 
 I want to announce my plans to create
 addon-shared-lvm-single-lock which would be the current
 version of the patches detailed here: http://wiki.opennebula.org/shared_lvm
 
 
 
 
 In a nutshell: the patch lets OpenNebula to use
 commercial off the shelf SANs that are providing a few
 pre-configured LUN-s as block devices over iSCSI, AoE or FC.
 
 
 The block devices in question are mounted on all the
 nodes and on the frontend, too. The device is split up to
 logical volumes that are only active on the node that runs
 the VM in question but are present everywhere. This allows
 live migration that is facilitated by migration hooks which
 activate and deactivate volumes as the instance travels. 
 
 
 
 This way, in certain cases the use of cLVM can be spared as
 all nodes except the frontend can run with read-only lvm
 metadata setting, meaning that local locking is sufficient
 on the frontend. This is only as long as other storage
 drivers don't try to modify any lvm vgs on the nodes -
 they won't work because of the read-only metadata.
 
 
 
 
 Please comment on the matter! Thanks!
 
 Cheers
 Mihály Héder
 MTA SZTAKI HBIT
 
 
 _______________________________________________
 
 Dev mailing list
 
 Dev at lists.opennebula.org
 
 http://lists.opennebula.org/listinfo.cgi/dev-opennebula.org
 
 
 
 
 
 -- 
 Jaime Melis
 Project Engineer
 OpenNebula - Flexible Enterprise Cloud Made Simple
 www.OpenNebula.org | jmelis at opennebula.org
 
 
 
 
 -----La pièce jointe associée suit-----
 
 _______________________________________________
 Dev mailing list
 Dev at lists.opennebula.org
 http://lists.opennebula.org/listinfo.cgi/dev-opennebula.org