[Ci-users] [CentOS-devel] Changes on CentOS CI and next steps

Tue Jul 26 13:19:46 UTC 2022
Niels de Vos <ndevos at redhat.com>

On Tue, Jul 26, 2022 at 12:17:33PM +0200, Fabian Arrotin wrote:
> On 26/07/2022 11:12, Niels de Vos wrote:
> > On Tue, Jul 26, 2022 at 10:18:54AM +0200, Fabian Arrotin wrote:
> > > On 11/07/2022 18:52, Niels de Vos wrote:
> > > > On Wed, Jun 15, 2022 at 08:49:34AM -0300, Camila Granella wrote:
> > > > > *TL;DR: CentOS CI is going hardwareless and if you wish your project
> > > > > remains using it, we need your opt-in by August 2022.
> > > > 
> > > > The Ceph-CSI project is a very happy user of the current CI
> > > > infrastructure, and we definitely wish to remain using it. We run a set
> > > > of jobs on a mix of bare-metal systems provided by Duffy and some
> > > > OpenShift native (containerized).
> > > > 
> > > > With recent minikube versions we might be able to run inside a VM,
> > > > however we do require setting up a (minimal) Ceph cluster per job. The
> > > > VMs we run on bare-metal get extra disks for the backing storage. If
> > > > that is an option for the VMs we obtain through Duffy, we should be able
> > > > to adapt our jobs.
> > > > 
> > > > Many thanks!
> > > > Niels
> > > > 
> > > 
> > > The EC2 VMs we'll provision will not get any extra disk[s] (not planned so
> > > far but we might reconsider and adapt as it should be easy).
> > > The other option is to ask for a bare-metal node to Duffy and you just
> > > continue to provision yourself VMs on top as you are doing right now.
> > > What do you think should match your needs ?
> > 
> > A bare-metal node is the easiest for us. We might be able to run the
> > tests on a VM, but that usually is way less stable and requires more
> > tuning and correction of the CI jobs as updates cause (temporary)
> > breakage. A VM needs to have sufficient CPU, RAM and storage to that we
> > can create block-devices and run Ceph on top of that (in containers?).
> > 
> > Thanks,
> > Niels
> > 
> 
> So you can still ask for bare-metal (select the correct pool in new duffy
> api) and so nothing should change for you.

Yes, and we're happy with that. I expect we need to prepare for a
VM-only solution in the future, right?

Cheers,
Niels