On Mon, Feb 3, 2014 at 11:33 PM, Itamar Heim <iheim(a)redhat.com> wrote:
On 02/03/2014 01:29 PM, Andrew Lau wrote:
> On Mon, Feb 3, 2014 at 11:27 PM, Itamar Heim <iheim(a)redhat.com
> <mailto:iheim@redhat.com>>wrote:
>
> On 02/03/2014 01:25 PM, Andrew Lau wrote:
>
> On Mon, Feb 3, 2014 at 11:23 PM, Itamar Heim <iheim(a)redhat.com
> <mailto:iheim@redhat.com>
> <mailto:iheim@redhat.com <mailto:iheim@redhat.com>>>__wrote:
>
>
>
> On 02/03/2014 01:19 PM, Andrew Lau wrote:
>
> The issue was a split-brain issue on the dom_md/ids
> file causing an
> input/output error, thanks!
>
>
> is this with gluster?
>
>
> Yup a 2 brick gluster replicated instance serving the NFS
> server, sorry
> was meant to say I resolved it too.
>
>
> you have to use a gluster with quorum, or this will happen often
>
>
>
> Yeah, I disabled quorum temporarily because I'm only using a two host
> scenario and I need to have the case scenario where one is to be
> shutdown the VMs won't end up in a paused state.
>
iiuc, without quorum you'll get both hosted engine and the SPM into split
brains
I've been using the non-quorum method for a while though and it's seemed
to
work all right, and this only case I had the split-brain was because I was
actually messing with gluster and deleting full brick contents to test
cgroups.