On 14/03/2023 10:30, Vladimir Benes wrote:

On Tue, 2023-03-14 at 08:13 +0100, Fabian Arrotin wrote:

...

On 13/03/2023 20:59, Vladimir Benes wrote:

...

Hi all, as we are testing new features in C9S it would be nice to have the latest kernel (or new enough) in EC2 environment. But sadly, it's not the case.

Today's NetworkManager MR testing failed as it contains quite new feature that landed in RHEL9.2 (in NM and kernel, too) and thus C9S some weeks ago.

so I checked the latest kernel available and it's 5.14.0- 286.el9.x86_64

so I checked kernel from c9s on EC2 and found Linux 5.14.0- 142.el9.x86_64 which is from August 2022.

https://kojihub.stream.centos.org/kojifiles/packages/kernel/5.14.0/

Is there any plan to move c9s deployments up2date?

Thank you so much! Vladimir

https://jenkins-networkmanager.apps.ocp.cloud.ci.centos.org/job/NetworkManag...

I see updated package that you can install ? kernel-core                   x86_64  5.14.0-285.el9

That's the latest kernel that the Stream team pushed out, and that is available for you to install

PS : in the past , for CentOS CI, we were updating through cloud-init but tenants said they'd update themselves the duffy nodes, and *asked* for the updates to not be applied through cloud-init as it was then competing with their own jobs, and were failing dnf transactions ...

So the base image is older, and we are supposed to update and reboot, got it. Not entirely sure where exactly do the reboot to not kill running stuff on the node. Could you point me to some howto, please?

Just an idea, would it be possible to have an option to have fully updated image when asking for EC2 instance via duffy to save time and resources? Not sure if others would see this useful, but we definitely would. RFE?

Thank you for explanation, Vladimir

Hi Vladimir,

Well, we'll not have different pools with different settings depending on tenants' needs, so we provide same image for everybody and based on tenants' feedback/input we can change things for all. That was the case when we had "dnf update -y" as part of cloud-init when initializing duffy nodes, but due to some discussions/threads on this list, it was then disabled again :

- https://lists.centos.org/pipermail/ci-users/2022-September/004608.html - https://lists.centos.org/pipermail/ci-users/2022-November/004624.html

OTOH, the fact that the "dnf update -y" step was taking so long is because upstream CentOS Stream AMI images weren't rolled on a frequently basis and so provisioned ec2 instances had to updates quite a lot of packages, even for the basic cloud image.

It seems that it's changing (see current thread on centos-devel list about that too) so hopefully we'd be able to update the AMI list for Duffy on a regular basis (once official images will be rolled out more frequently that is) and so enable back the dnf update part too (shorter or even then maybe not needed at all)