NO-JIRA: overrides-c10s: pin kernel to 6.12.0-71 #7
Conversation
This is a workaround for [1]. We are hitting a qemu bug since the introduction of the kernel patch [2]. With the newer kernel that does not contain this patch, the tests are working. [1] openshift/os#1818 [2] torvalds/linux@6aa989a
|
@jcapiitao: This pull request explicitly references no jira issue. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the openshift-eng/jira-lifecycle-plugin repository. |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: jcapiitao The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
|
||
| #packages: | ||
| packages-ppc64le: |
There was a problem hiding this comment.
I think we usually do a separate file for arch-specifix overrides
There was a problem hiding this comment.
Hmm I used the same pattern as in openshift/os@5e54e65
In coreos-assembler codebase, I see only manifest-lock.overrides.{arch}.yaml which is handled. In f-c-c and this repo, I don't see manifests or overrides file with arches in filename, but maybe it was the case in the past.
There was a problem hiding this comment.
I think we usually do a separate file for arch-specifix overrides
We do for FCOS where we actually use manifest-lock.overrides.{arch}.yaml as input to the rpm-ostree compose.
For EL we don't have lockfiles so we have to resort to putting the NVRs in the packages: definition.
There was a problem hiding this comment.
but when we do add overrides.. this would be the place to do it for c10s.
If we're having the same problem in RHEL then the rhel-10.1 version would be the place for it.
Now that we can do conditional-includes we could consider consolidating those 4 files into one.
| # This needs to be removed once https://gitlab.com/qemu-project/qemu/-/issues/2966 is fixed. | ||
| [c10s-kernel-6.12.0-71] | ||
| name=CentOS Stream 10 - kernel-6.12.0-71.el10 | ||
| baseurl=https://jcapitao.fedorapeople.org/c10s-kernel-6.12.0-71/ |
There was a problem hiding this comment.
I don't think we'd ever want to pull RPMS from non official places so I don't think we should merge this as is.
I guess the problem is that the CentOS repos don't have older versions of packages in them?
There was a problem hiding this comment.
ahh yes:
rhel-coreos-config/c10s-mirror.repo
Lines 1 to 4 in e5e42d5
We can just switch to the mirror versions of the repo for now while we wait. Another option is to add a "fast-track" view into that repo just for the kernel-* package so we can continue to get newer versions of everything else.
There was a problem hiding this comment.
but I see that over in openshift/os#1818 (comment) it doesn't have every version of the package :(
There was a problem hiding this comment.
yes, I tried compose and mirror repo URLs before hosting the kernel onto my own fedorapeople.org space.
|
That particular commit is only in the (yet to be released) kernel 6.15. I think you are saying that our kernels in c10s and RHEL backported that patch and that's why we're seeing the problem here. If that's the case, what I don't understand is why we're not seeing a similar issue when testing ppc64le in rawhide (that does have 6.15 RC kernels). |
yes, it's already backport in c10s kernel as per [1] (i.e see
I was wondering the same question, so I went a little bit further on the troubleshooting, and I found out [2]. [1] https://gitlab.com/redhat/centos-stream/rpms/kernel/-/commit/dc7a191ebb6074eb0107fbf255c99b303aff106f |
Sounds good to me. I guess we should open a RHEL bug for that to make the request? |
I filed a bug in RHEL https://issues.redhat.com/browse/RHEL-92470 asking to backport the patch. |
|
PR needs rebase. Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. |
This is a workaround for [1]. We are hitting a QEMU bug since the introduction of the kernel patch [2].
With the newer kernel that does not contain this patch, the tests are working.
Note that this is a workaround targeting our tooling (i.e how we run our kola tests), but on the other hand, it covers the usecase of the enduser booting the image with QEMU.
I'm using my fedorapeople.org namespace to serve the packages, let me know if there is a more proper way.
[1] openshift/os#1818
[2] torvalds/linux@6aa989a