On Tue, Jan 31, 2017 at 2:20 PM, Fernando Frediani <fernando.frediani@upx.com.br> wrote:
Hello

Note that Gluster hasn't all the VSAN features yet so you will be able to replicate data.

Also I personally think replica=3 is overkill and waste of space for mos scenarios. 2 should be enough and give a raid 1 like.

I'm not sure it's like RAID 1. In RAID 1, when one disk fails, I'm pretty sure I know where the correct data is.
In Gluster, if you have a split brain, where's the correct data?

I suggest perhaps using an arbiter.
Y.
 



On 31/01/2017 06:29, Nicolas Ecarnot wrote:
Le 31/01/2017 à 09:15, Anantha Raghava a écrit :
Hi,

We are trying to create a setup that uses the internal disks of the
hosts / nodes, yet provide the high availability, replication and
failover using oVirt. The setup we are typing to build is close to
VMWare VSAN which allows for all the above just using the internal disks
of the ESXi servers.

Can we achieve something similar with oVirt with Gluster?

Absolutely. One of our oVirt setup is done this way.
Three hosts are set up as glusterFS servers (replica-3), as well as oVirt nodes.
We choose to add a fourth host as an standalone engine, but you can choose to use a VM for that (hyperconverge setup).

I have no experience on similar setup with a random number of nodes, neither if this can be achievable (some kind of network RAID-10)... (?)


_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users