bhouse-nexthop opened a new issue, #117:
URL: https://github.com/apache/cloudstack-kubernetes-provider/issues/117

   `deployment.yaml` (also what CloudStack's 
`create-kubernetes-binaries-iso.sh` downloads as `provider.yaml` for CKS) can't 
bring up the CCM on a fresh cluster.
   
   | Missing | Effect on a new cluster |
   |---|---|
   | `hostNetwork: true` | Nodes stay tainted `uninitialized` until the CCM 
runs, and the CNI and CoreDNS may not be up yet. A pod-network CCM can't get an 
IP or reach the API, so nothing gets initialised. |
   | `dnsPolicy` for host networking | With host networking the CCM must use 
the node's resolver: CoreDNS isn't running yet, and the CloudStack API endpoint 
is an external name. |
   | Toleration for `node-role.kubernetes.io/control-plane:NoSchedule` (and the 
pre-1.25 `master` taint) | With workers not yet joined or not initialised, 
control-plane nodes are often the only place the CCM can go. |
   | Tolerations for `node.kubernetes.io/not-ready:NoSchedule` and 
`node.kubernetes.io/network-unavailable:NoSchedule` | Nodes are `NotReady` (and 
may be `NetworkUnavailable`) until the CNI is up, so the CCM pod is never 
scheduled. |
   
   The `node.cloudprovider.kubernetes.io/uninitialized` toleration is already 
there.
   
   The AWS, OpenStack and vSphere CCM manifests all run the CCM on the host 
network and on control-plane nodes.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to