On 06/29/2012 04:57 PM, Itamar Heim wrote:
On 06/29/2012 11:45 PM, Robert Middleswarth wrote:
> On 06/29/2012 03:15 PM, Itamar Heim wrote:
>> On 06/29/2012 06:15 PM, Nathan Stratton wrote:
>>> I must admit the first part of managing GlusterFS with 3.3 works very
>>> well out of the box on oVirt 3.1.
>>>
>>> Turn on GlusterFS in the cluster
>>> Add your nodes
>>> Manually add the peers in Gluster
>>> Edit /etc/nfsmount.conf and set Defaultproto=tcp and Defaultvers=3
>>> Create your gluster volume
>>> Start the volume
>>> Mount it in on one of your nodes in say /mnt
>>> chown 36.36 /mnt
>>>
>>> The only problem now is that you have to now use the volume as a NFS
>>> server.
>>
>> why not as posixfs?
> Because I helped him get it working and I was never able to get posixfs
> working but was able to get nfs working.
i have these steps courtesy of haim ateya worth trying:
* add new volume constructed from one brick
- volume name = myVol
- brick dir = /opt/myBrick
* start volume
- ssh to node
* chown -R vdsm:kvm /opt/myBrick
- using ovirt do the following):
* create new DC (Posix compliant FS)
* create new cluster
* add new host (virt)
* new domain dialog
- path = node:/myVol
- VFS Type = glusterfs
- Mount Options = vers=3
* if start vm failedt to selinux - disable SELINUX on host
you would need the newer kernel with direct_io iirc
The mount options is useful to
know. I have moved over to centos 6.2
since F17 was freezing up on me every 10 to 12 hours.
I am guessing I will need to wait for CentOS 6.3 to get direct IO
support and hopefully live snapshots as well. I might test glusterfs as
well. PS I use localhost:/volumename for NFS.
Thanks
Robert