Mention RPM upgrade path for prior patch releases #21898
Conversation
k8s-ci-robot
added
cncf-cla: yes
k8s-ci-robot
added
the
sig/docs
saschagrunert
force-pushed
the
install-upgrade
branch
2 times, most recently
from
3f4531f to
a2d2ff8
Compare
|
Deploy preview for kubernetes-io-master-staging ready! Built with commit f74a5ae https://deploy-preview-21898--kubernetes-io-master-staging.netlify.app |
|
Deploy preview for kubernetes-io-master-staging ready! Built with commit b4d8718 https://deploy-preview-21898--kubernetes-io-master-staging.netlify.app |
saschagrunert
force-pushed
the
install-upgrade
branch
3 times, most recently
from
6bac111 to
2286971
Compare
content/en/docs/setup/production-environment/tools/kubeadm/install-kubeadm.md
Outdated
Show resolved
Hide resolved
saschagrunert
force-pushed
the
install-upgrade
branch
from
2286971 to
c508883
Compare
|
@saschagrunert Should we also consider adding a note to the upgrade docs: https://kubernetes.io/docs/tasks/administer-cluster/kubeadm/kubeadm-upgrade/? Or should that be done by the SCL/kubeadm team instead? |
Hm, do you think we could handle the removal of the package directly in kubeadm? This way the user would not have a need for a manual intervention. I'll check if I can find something. |
I'm not really sure, but AFAIK, kubeadm intentionally doesn't manage any packages and instead just focuses on the Kubernetes (components) itself. From the docs:
We should probably ping someone from SCL and check what they would think about this idea. |
Good point, I'll focus now on updating the mentioned docs too. |
saschagrunert
force-pushed
the
install-upgrade
branch
2 times, most recently
from
0f86c43 to
37ee0de
Compare
| - If you're running CentOS, RHEL, Fedora or directly use the | ||
| [yum.k8s.io](https://yum.k8s.io) repositories: The `kubernetes-cni` package | ||
| was obsoleted with the release of Kubernetes v1.18.4, v1.17.7 and v1.16.11 on | ||
| June 17th, 2020. This means in case of doing an upgrade from prior versions, | ||
| please ensure that the package gets **manually** removed before doing the | ||
| upgrade by executing `yum remove kubernetes-cni-0.7.5-0`. |
There was a problem hiding this comment.
@saschagrunert I've thought about this a bit more. I can't test this because I don't have any custom RPM repo, so this is based on the assumption and has to be verified.
Looking at the output from your PR to fix packages (kubernetes/release#1367 (comment)), once the kubernetes-cni package is gone, kubelet will be gone too, and mostly like kubeadm as well (because kubeadm has a dependency on kubernetes-cni).
kubeadm is not a big deal because the first step is to install the latest one, but kubelet is supposed to be installed after kubeadm upgrade (as per docs). If there's no kubelet on the instance, can kubeadm upgrade work or is it going to fail? I'd say it might fail because there's no kubelet to schedule all the new static pods, but I'm not sure about that.
If that's true, we might have to update docs to mention this as well.
|
/assign |
| doing an upgrade from prior versions, please ensure that the package gets | ||
| **manually** removed before doing the upgrade by executing `yum remove | ||
| kubernetes-cni-0.7.5-0`. | ||
| {{</ warning >}} |
There was a problem hiding this comment.
/cc @sftim
PTAL on the wording, AFAIK we do not include dates and PATCH version information in notes/warnings.
There was a problem hiding this comment.
nit: Again, will this or does this information appear in the release notes?
There was a problem hiding this comment.
I think we can just state that it is obsolete from now own, but we still have to mention from which upgrade path (changed the wording a bit).
We do not mention the obsoletion in the release notes, but I could follow-up with a dedicated PR to add it to them.
|
/sig cluster-lifecycle |
k8s-ci-robot
added
the
sig/cluster-lifecycle
|
how are, for example 1.18 users / readers of the 1.18 docs, supposed to know about this complication? |
|
@neolit123 , I do think this meets the rare criteria to be back ported to all impacted release docs. Has there been consensus from sig-release that this is the path forward? |
content/en/docs/setup/production-environment/tools/kubeadm/install-kubeadm.md
Show resolved
Hide resolved
content/en/docs/tasks/administer-cluster/kubeadm/kubeadm-upgrade.md
Outdated
Show resolved
Hide resolved
There is a manual intervention needed for upgrading from previous patch releases because the kubernetes-cni package got obsoleted. The docs now mention that and provide a possible workaround. Signed-off-by: Sascha Grunert <sgrunert@suse.com>
saschagrunert
force-pushed
the
install-upgrade
branch
from
37ee0de to
b4d8718
Compare
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
/hold Rethinking the overall solution. |
k8s-ci-robot
added
the
do-not-merge/hold
|
Closing, we decided to reintroduce |
|
@saschagrunert: Closed this PR. 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 kubernetes/test-infra repository. |
|
note another part of the docs was already amended with the package change for multiple languages: |
Thank you for the hint, I can follow up on that after we re-created the packages. |
There is a manual intervention needed for upgrading from previous patch
releases because the kubernetes-cni package got obsoleted. The docs now
mention that and provide a possible workaround.
Depends on kubernetes/release#1367