[Gluster-users] VM Images Over GlusterFS

Alexander Atticus alexander.atticus at gmail.com
Mon Jan 19 16:18:23 UTC 2009


I'm mostly concerned about the auto-healing feature.  I'm assuming you
have some kind of redundancy setup in your GlusterFS configuration.
What happens when you have a VM running on an image served from
Gluster, and you simulate a crash on one of your bricks? Does the VM
stay up?

My impression is that it may but if you reboot the VM the VM after you
bring the crashed brick back up it will block until the VM disk image
has a chance to heal (sync).

What I wanted to setup here is Striping over all bricks with each
brick setup to AFR to another for redundancy.  I don't know whether
this will deliver the availability I'm looking for or whether I need
HA or something else as well.  Essentially I don't want VMs to notice
at all when a brick goes down and I'm not sure this is possible?

Alexander

On Mon, Jan 19, 2009 at 6:03 AM, Enrico Valsecchi <admin at hostyle.it> wrote:
> Matthias Teege ha scritto:
>>> I am wondering if there is anyone on list that is successfully serving
>>> virtual machine images (Xen, KVM, VMWare, etc.) over GlusterFS.  I am
>>>
>>
>> I using it with Xen and vserver. With Xen and large disk images it
>> works. Live migration and failover works. Healing is a major problem. It
>> blocks the image for sometime so you have downtime anyway. I have to
>> write a heartbeat ressource to manage the healing before restart the
>> VM. Linux Vserver works to but there are performance issues im working
>> on. It is no problem to "dd" a file to gluster with reasonable speed but
>> in "realtime" vserver is much faster on drbd then on gluster. Maybe I hit
>> a "lot of small files" Problem here or there is a problem with vservers
>> COW mechanism. I use 1.3.12 only.
>>
> Hello!
>
> I have tested with success Xen with GlusterFs.
> I does not have a cleaned idea if my scenario is same at your.
> I have a xen server attached with Infiniband and RDMA access on 2
> GlusterFS servers, used to store xen images files, and configured in AFR
> by Client.
> GlusterFS version are 1.3.12, xen tested are 3.2.0, 3.2.1, 3.3.0.
> Images file are mounted in xen with tap:aio disk option, because with
> this system have more performance that "file" option and,
> in order to more post found in the net, it's more safe.
> Only way to use tap:aio on xen are to mount GlusterFS with "-d" option.
>
> Now, where problems you have found?
>
>
> Enrico
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users
>




More information about the Gluster-users mailing list