[ovirt-users] Host In Peril?

Duckworth, Douglas C duckd at tulane.edu
Mon Nov 2 20:54:18 UTC 2015


Ido

We are using RHEV 3.5.4 with 6.7 Hosts.

I deployed another Host for the Hosted Engine VM, moved the Hosted
Engine VM to that new Host, then fixed the networks on that bad Host.

Setting the Bad host to "Local" Maintenance Mode did not trigger VM
startup on the other Host.  So we shut it down manually as described
under section "Maintenance Flows:"

http://www.ovirt.org/Features/Self_Hosted_Engine

All is now good in our DR site Cluster.

Thank you for your informative explanation!

Best
Doug

-- 
Thanks

Douglas Charles Duckworth
Unix Administrator
Tulane University
Technology Services
1555 Poydras Ave
NOLA -- 70112

E: duckd at tulane.edu
O: 504-988-9341
F: 504-988-8505

On 11/01/2015 01:25 AM, Ido Barkan wrote:
> 
> On Fri, Oct 30, 2015 at 12:04 AM, Duckworth, Douglas C <duckd at tulane.edu
> <mailto:duckd at tulane.edu>> wrote:
> 
>     Doug
> 
> 
> 
> Hi Doug,
> You haven't sated your vdsm/oVirt version so I my answers might be
> somewhat generic.
> 1) once a network configuration (net-conf) is given by the engine, and
> is configured by vdsm,
>     vdsm stores it under /var/run/vdsm/netconf/{nets,bonds}. This is
> used as an intermediate
>     place fro this configuration, and allows the user to verify that his
> requested net-conf
>     indeed serves his goals.
>     If the worst has happen, and all communication is now broken, and
> since this path does
>     not survive a reboot, it leaves the user with the emergency ultimate
> option of rebooting
>     the host and let it restooe the last knownworking net-conf (see2).
> 2) once engine calls setSafeNetworkConfig API verb, this info is copied
> into /var/lib/vdsm/persistence/netconf/
>     in an atomic operation. Since this path survives  a host reboot, the
> host will restore
>     all networks from that place.
>     If I am not mistaking, the engine defaults to call this API
> immediately after each netwrok
>     setup (it can be toggled in a checkbox in the UI).
> 
> In general, the ultimate knoledge about the required networks sits in
> the administrator mind,
> and hopefully, in the engine database. So the best options you guys have
> is adding the required netwroks
> one by one from the engine, and call setSafeNetworkConfig after each
> verified successful state.
> This will ensure a proper state+backup on your host.
> 
> -- 
> Thanks,
> Ido Barkan



More information about the Users mailing list