kops: Cluster cannot start with older versions of k8s
After merging https://github.com/kubernetes/kops/pull/10048, the new way of installing Docker is conflicting with the pre-installed Docker in the kope.io images.
https://prow.k8s.io/view/gcs/kubernetes-jenkins/logs/e2e-kops-aws-k8s-1-15/1320685447990480896
A similar problem existed in 1.17 and was fixed in https://github.com/kubernetes-sigs/image-builder/pull/273, by just copying the Docker deb files to /var/cache/nodeup/packages and letting kOps install them if the hash matches.
All kope.io images from 1.11 to 1.16, maybe also 1.17.
About this issue
- Original URL
- State: closed
- Created 4 years ago
- Comments: 18 (7 by maintainers)
Commits related to this issue
- Allow nodeup (and others) to replace in-use files By creating a tempfile and then moving the file into place, we both write more atomically and we can overwrite in-use files. Issue #10122 — committed to justinsb/kops by justinsb 3 years ago
- Install dbus if needed for protokube with containerd Issue #10122 — committed to justinsb/kops by justinsb 3 years ago
- Allow nodeup (and others) to replace in-use files By creating a tempfile and then moving the file into place, we both write more atomically and we can overwrite in-use files. Issue #10122 — committed to hakman/kops by justinsb 3 years ago
- alpha channel: Update older images To pick up the sudo fixes, also addresses some problems with docker preinstallation. Issue #10122 — committed to justinsb/kops by justinsb 3 years ago
Thanks for explaining. You may also disable the Docker and containerd services in the AMI. kOps will start them later.