run TNC as a static pod before cluster having state#3077
run TNC as a static pod before cluster having state#3077squat merged 4 commits intocoreos:masterfrom
Conversation
|
Can one of the admins verify this patch? |
|
@enxebre tnc dev has been made public. |
|
ok to test |
| namespace: kube-system | ||
| labels: | ||
| k8s-app: tectonic-node-controller | ||
| managed-by-channel-operator: "true" |
There was a problem hiding this comment.
This static pod is not managed by the channel operator
| - name: cluster-config | ||
| hostPath: | ||
| path: /opt/tectonic/tnc-config | ||
| nodeSelector: |
There was a problem hiding this comment.
A static pod does not need a node selector; it is manually scheduled
| nodeSelector: | ||
| node-role.kubernetes.io/master: "" | ||
| tolerations: | ||
| - key: "node-role.kubernetes.io/master" |
There was a problem hiding this comment.
Are you sure the toleration is required? I suspect this is not needed for static pods
| - key: "node-role.kubernetes.io/master" | ||
| operator: "Exists" | ||
| effect: "NoSchedule" | ||
|
|
steps/assets/resources/rm-assets.sh
Outdated
| /usr/bin/aws --region="$REGION" s3 cp /tmp/assets.zip s3://"$LOCATION/ign/v1/role/master" | ||
| ' | ||
| # Remove TNC static pod | ||
| rm /etc/kubernetes/manifests/tectonic-node-controller-pod.yaml |
There was a problem hiding this comment.
I think this is a little abusive of the rm-assets script, which is only concerned with removing assets from object storage. Until we create the cleanup service, I think this would be a better fit for the end of the tectonic service, which is responsible for launching all tectonic components. This way the removal of the static pod is more directly coupled with the creation of the operators
|
I saw that you briefly posted the golang code enabling the etcd step. Will that come in a separate PR? |
4 participants
This needs the TNC image to be public as there's no state for holding secrets at this time. Fix INST-935