Merge pull request #52929 from krousey/gciupgrades

Automatic merge from submit-queue (batch tested with PRs 52880, 52855, 52761, 52885, 52929). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>..

Add cos as an alias for gci in the upgrade script

This was causing some issues when upgrading from a GCI image. This is the same conversion happening in config-defaults.sh.

https://k8s-gubernator.appspot.com/build/kubernetes-jenkins/logs/ci-kubernetes-e2e-gce-stable2-stable1-upgrade-cluster-new/205

The node image was being left at COS, and when we  went to build the kube-env, we only check against "gci". This caused us to not fully construct the environment for nodes and then they couldn't fully come up after an upgrade.

I've already fixed the CI test suites to explicitly specify "gci", but this auto-detection logic should be fixed too.

Fixes: #52930
pull/6/head
Kubernetes Submit Queue 2017-09-24 04:30:15 -07:00 committed by GitHub
commit 0dc096a9d1
1 changed files with 5 additions and 0 deletions

View File

@ -253,6 +253,11 @@ function setup-base-image() {
if [[ "${env_os_distro}" == "false" ]]; then
echo "== Ensuring that new Node base OS image matched the existing Node base OS image"
NODE_OS_DISTRIBUTION=$(get-node-os "${NODE_NAMES[0]}")
if [[ "${NODE_OS_DISTRIBUTION}" == "cos" ]]; then
NODE_OS_DISTRIBUTION="gci"
fi
source "${KUBE_ROOT}/cluster/gce/${NODE_OS_DISTRIBUTION}/node-helper.sh"
# Reset the node image based on current os distro
set-node-image