[ovirt-users] 3.6 : Moving the hosted-engine to another storage

Simone Tiraboschi stirabos at redhat.com
Thu Jul 28 09:13:43 UTC 2016


On Thu, Jul 28, 2016 at 10:59 AM, Alexis HAUSER
<alexis.hauser at telecom-bretagne.eu> wrote:
>>Unfortunately we know that migrating from HE to HE is not as simple as
>>from physical to HE:
>>https://bugzilla.redhat.com/show_bug.cgi?id=1240466#c21
>>In general the issue is that the DB backup form the old hosted-engine
>>VM contains a lot of references to the previous hosted-engine env and
>>you cannot simply remove/edit them from the engine since they are lock
>>so you have to manually remove them from the DB which is quite
>>risky/error prone.
>
> This is a bit scaring. In case of issue with engine and trying to recover, it could also happen.

The backup and restore procedure on the same env is well testes, the
issue is simply if you need to restore on a different environment
(it's basically a migration) cause you have to remove all the
references to the old env.

> What other way would you suggest for backing the engine VM and being sure to be able to restore it as it was without errors ? Have you ever tried to backup/restore from rsync ?
> If there are data in the DB written when you're performing it, do you think it can cause issues ? If ovirt-engine service is stopped, is that problem avoided ?

Postgres is transactional so if you do it in the proper way I don't
see issue but stopping the ovirt-engine service will of course help.
But the issue is not that you risk data corruption, the issue is that
when you import a backup of the engine DB to a different env, that
backup says the hosted-engine storage domain is still the old one
since it was that in the previous env and so on and you can simply
edit the hosted-engine strage domain location from the engine itself
since we are preventing it.

>>In the mean time I'd suggest, if feasible, to redeploy a new
>>hosted-engine env and reattach there your storage domains and your
>>hosts.
>>This will imply a downtime.
>
> Ok, I think I'll do that. A downtime isn't a problem right now, as I'm still at a pre-production step. (preparing it for production soon)



More information about the Users mailing list