migration of services from linode server to phx2 lab

[Michael Scherer]

Le mardi 16 juin 2015 à 03:38 -0400, Eyal Edri a écrit :
> 
> ----- Original Message -----
> > From: "Karsten Wade" <kwade at redhat.com>
> > To: infra at ovirt.org
> > Sent: Tuesday, June 16, 2015 6:29:04 AM
> > Subject: Re: migration of services from linode server to phx2 lab
> > 
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> > 
> > Hi,
> > 
> > Between myself (original linode01.ovirt.org admin) and Michael (misc,
> > aka knower-of-stuff), what can we do to get us off this Linode
> > instance? From what I can tell, at least Mailman is running from there.
> > 
> > If we still need service failover, can we switch to another Red
> > Hat-provided service such as an OpenShift or OpenStack instance?
> 
> Hi Karsten,
> I know this has been taking for too long, but unfortunately there are many reasons
> and obstacles that prevented us from moving, which i'll explain below,
> while there is a risk right now of moving services to PHX, but i think we can try.
> 
> Blocking items:
>  - were missing more hypervisors to the production DC ( we finally got them installed last week, and are now in final stages of bringing them up)
>  - storage issue - currently NFS storage is quite slow, we are testing moving to a mixed mode of local + nfs for the jenkins slaves,
>                    though might be that the production service won't get affected too much - worth a try.
>  - lack of monitoring for servers, which might add risk if we hit an issue.
> 
> there are some other issues, but not blocking imo.
> 
> Here is what needs to be done (per service to migrate off linode), and we'd appreciate any help given.
> 1. create new VM on production DC in PHX (assign dns/ip/etc.. )
> 2. create puppet manifests to manage that vm, so it will be easy to reproduce it and maintain it
> 3. install the relevant software on it (for e.g mailman/ircbot/etc...)
> 4. test to see it works
> 5. do the actual migration (downtime of existing service, and bringing up the new one)

So, last time I did look, the puppet setup was a bit "unfinished" rather
overkill ( using r10k to deploy count as overkill when you do not have
stage/prod/qa, at least for me ). Did stuff changed, or should first
some effort be made to fix/automate/improve the puppet setup ?


-- 
Michael Scherer
Sysadmin, Community Infrastructure and Platform, OSAS

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://lists.ovirt.org/pipermail/infra/attachments/20150617/ac0eb278/attachment.sig>