[one-users] fibre channel datastore

[Jaime Melis]

Hello Denis,

1. Do I have to connect the FC datastore to opennebula server ?
>
No. You don't need to (you could, but there's really no point).

> 2. Which driver should I use on compute nodes ?
>
The LVM drivers.

> 3. Will my virtual machines be persistent during infrastructure reboots ?

Yes.

The similar setup you have described is exactly what you want to replicate
with OpenNebula. You need the LVM drivers for this. I have just added a
diagram to the LVM guide [1] to show you exactly the setup you need. The
idea is to export the same LUN to all the OpenNebula nodes and create a
cLVM* on top of it. OpenNebula will speak to just one of the nodes of that
cluster ($HOST parameter in the Datastore template). If the FC
configuration is persistent, and the cLVM configuration is also persistent,
then your VMs will persist during reboots.

Note that the underlying storage doesn't affect OpenNebula at all, you
could do this with FC, iSCSI, or even a NAS, as long as you export the same
block device to all the hosts (are configure it to persist after reboots).

By the way, I strongly recommend reading this article, written by the
people at MTA SZTAKI, explaining many things about LVM deployment. You will
find a lot of hints, best practices and troubleshooting tips in there.

* Maybe you don't need cLVM, after reading the article [2] you will be able
to understand the pros and cons of using it.

[1] http://opennebula.org/documentation:rel4.0:lvm_ds
[2] http://wiki.opennebula.org/shared_lvm

Cheers,
Jaime

On Tue, May 21, 2013 at 9:35 AM, Denis J. Cirulis
<denis.cirulis at gmail.com>wrote:

> Hello,
>
> I have to setup proof of concept cloud using opennebula and zfs san.
> I can not understand the correct scenario of running virtual machines from
> FC:
>
> 1. Do I have to connect the FC datastore to opennebula server ?
> 2. Which driver should I use on compute nodes ?
> 3. Will my virtual machines be persistent during infrastructure reboots ?
>
> I already had similar setup but with plain libvirt and kvm, the concept
> was to use one lun per datastore from fc storage using fc hba and switch,
> which was advertised to compute nodes as lvm vg, then each virtual machine
> had 1+n logical volumes from this vg as its hard drive. Backup was
> performed via lvm snapshot and dd/rsync. There was a possibility to migrate
> storage from one node to another without vm downtime using virsh block-copy
> --live.
>
> What are the correct steps to achieve the same functionality on opennebula
> 4.0 ?
> I'm running CentOS 6.4 both on opennebula server and compute nodes.
>
> Thanks in advance!
>
> _______________________________________________
> Users mailing list
> Users at lists.opennebula.org
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>
>
-- 
Join us at OpenNebulaConf2013 <http://opennebulaconf.com/> in Berlin, 24-26
September, 2013
--
Jaime Melis
Project Engineer
OpenNebula - The Open Source Toolkit for Cloud Computing
www.OpenNebula.org | jmelis at opennebula.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20130521/7f8a797b/attachment-0001.htm>