----- Original Message -----
From: "Itamar Heim" <iheim(a)redhat.com>
To: "Sven Kieske" <S.Kieske(a)mittwald.de>, users(a)ovirt.org, "Eli
Mesika" <emesika(a)redhat.com>, "Barak Azulay"
<bazulay(a)redhat.com>
Sent: Tuesday, February 25, 2014 12:14:48 PM
Subject: Re: [Users] oVirt 3.5 planning
On 02/25/2014 11:55 AM, Sven Kieske wrote:
> RFE] Allow to perform fence operations from engine (or from a host in
> another DC)
> the functionality of the current fencing feature is so crippled that
> it is plain useless in local storage data centers, because you can't
> use it at all (no other fencing host in cluster)!
>
https://bugzilla.redhat.com/show_bug.cgi?id=1054778
eli - thoughts on above?
The fencing mechanism certainly does not work for local storage domain,
It looks like this was neglected since the fencing mechanism was designed to prevent data
corruption .. and this use case (single host in DC) we are not worried about it.
In first look it looks like an easy implementation - just to add a system wide proxy
selection policy,
However when thinking on adding this option I can think of a few issues that needs to be
discussed first:
- Can we assume network connectivity between hosts in different DCs ?
- If this is not the case than can/should we enable specific DCs that we know has this
connectivity ?
Or do we just select a system wide proxy and try to fence ... this might have timeout
issues in case there is no connectivity ... and fencing might take a long time to perform.
Thanks
Barak Azulay