Fix delete node permission on Fedora CoreOS node shutdown

* On cloud platforms, `delete-node.service` tries to delete the
local node (not always possible depending on preemption time)
* Since v1.18.3, kubelet TLS bootstrap generates a kubeconfig
in `/var/lib/kubelet` which should be used with kubectl in
the delete-node oneshot
This commit is contained in:
Dalton Hubble
2020-10-18 23:29:19 -07:00
parent eda78db08e
commit df17253e72
5 changed files with 15 additions and 7 deletions

View File

@ -4,13 +4,17 @@ Notable changes between versions.
## Latest
* Remove `asset_dir` variable (default off in [v1.17.0](https://github.com/poseidon/typhoon/pull/595), deprecated in [v1.18.0](https://github.com/poseidon/typhoon/pull/678))
* Remove `asset_dir` variable (defaulted off in [v1.17.0](https://github.com/poseidon/typhoon/pull/595), deprecated in [v1.18.0](https://github.com/poseidon/typhoon/pull/678))
### Fedora CoreOS
* Fix local node delete oneshot on node shutdown ([#856](https://github.com/poseidon/typhoon/pull/855))
### Flatcar Linux
* Change `kubelet.service` container runner from rkt to docker ([#855](https://github.com/poseidon/typhoon/pull/855))
* Change `delete-node.service` to be inlined and use docker ([#855](https://github.com/poseidon/typhoon/pull/855))
* Fix mount to restore permission to delete the local node on shutdown (cloud-only)
* Change `delete-node.service` to use docker and an inline ExecStart ([#855](https://github.com/poseidon/typhoon/pull/855))
* Fix local node delete oneshot on node shutdown ([#855](https://github.com/poseidon/typhoon/pull/855))
## v1.19.3