[ovirt-users] [Ovirt 4.1] Hosted engine with glusterfs (2 node)
Simone Tiraboschi
stirabos at redhat.com
Tue Jul 11 07:27:21 UTC 2017
On Tue, Jul 11, 2017 at 7:17 AM, TranceWorldLogic . <
tranceworldlogic at gmail.com> wrote:
> Hi,
>
> I was trying to setup hosted engine on 2 host machine but it not allowing
> to setup hosted engine on 2 node gluster file system.
>
> I have also modified vdsm.com to allow 2 replica count.
>
> But latter I found in code it is hard coded to 3 as shown below:
> ------------------- -------------------------------
> ------------------------------ ---------------
> src/plugins/gr-he-setup/storage/nfs.py
>
> if* replicaCount != '3'*:
> raise RuntimeError(
> _(
> 'GlusterFS Volume is not using replica 3'
> )
> )
> ------------------- -------------------------------
> ------------------------------ ---------------
>
> Can I get to know the reason to hard coded to 3 ?
> from online I got to know that split brain issue not get resolve
> automatically using 2 node gluster setup. But please let me know more (why
> 3 ? why not 2 or 4 ?)
>
Simply because two is even and three is odd.
In a split condition over only two nodes, if node A says A and nodes B says
B you cannot decide if the truth is A or B.
With three nodes, if node says A but nodes B AND nodes C says B, the truth
is B by design.
That's why you need three nodes with at least 2 of them up and working.
If you want to save something on storage side you could evaluate two
regular nodes plus an arbiter one.
>
> If I am missing something please let me know.
> Also want to understand, can quorum can solve 2 node gluster issue
> (split-brain) ?
>
> Thanks,
> ~Rohit
>
>
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20170711/dea7d45f/attachment-0001.html>
More information about the Users
mailing list