[ovirt-users] Ovirt 3.5.x Power failure recovery

[Itamar Heim]

On 02/09/2015 04:50 AM, Raymond wrote:
> Hi,
> Long time read-only user here :)
> Unfortunately I need some info which I can't seem to find in the archives and on Google.
>
> In the past 5 months we’ve had two very rare occasions of power failures in my hometown.
> First one was a ms during spike or drop, just enough to cause a reboot, the other was a 30min failure.
>
> Problem
> After the power failure my ovirt nodes boot into CentOS 6.6 and are running fine.
> In both cases my ovirt-engine-VM wouldn't "start" due to service Postgresql not starting.
> I tried some things before reverting back to an older DB snapshot that was on the VM disk, but still not working.
> Postgresql is not starting due to "Duplicate UUID"
>
> My solution
> Calm down girlfriend and don't sleep
> Reinstall 1 node, create new ovirt-engine, recreate VM's and copy VM disk data from old store to new store.
> Add second node
>
> Other solutions I can think of
> 1. Start cluster without reinstall ;)
> 2. Buy UPS (€150 one time, extra €60 yearly power usage = €450 in 5 years)
>
> At this moment I’m 1 click away from buying the UPS, but I prefer a more elegant solution.
>
> After the last power failure I did NOT reinstall my second ovirt node and the old engine is also on storage available.
> Is there someone that wants some data to troubleshoot/analyse?
>
> Or is this just one of those things? Buy the UPS and get on with your life?
>
> Short HW overview
> 2x ovirt 3.5.x (i3/32GB ram,1Gb eth VM network, dual x520 NFS eth)
> 1x NFS (i3/4GB ram, 5TB SAS and 6TB SATA on Dell PERC,1Gb eth mgmt, dual x520 10Gb)
>
> 1Gb is used for VM networking, the 10Gb is connected via DAC cables and runs NFS-storage.
> This works great! Whole cluster runs below 120w and VM disk performance is around 700MB/s :)
> I bought all the HW with power usage in mind, PicoPSU's and 35w CPU's in all nodes.
> So adding 30w "extra" just for the UPS feels a bit like killing an elephant with tissues...
>
> kind regards
> Raymond
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>

if your DC is 3.5, you don't need to copy the vm disks, just install new 
engine and import the data domain
http://www.ovirt.org/Features/ImportStorageDomain

(i assume there is a way to fix the postgres issue itself as well)