Upgrading CTE for Kubernetes

Introduction

Upgrading CTE for Kubernetes to the latest release is accomplished by invoking the deploy.sh script with the desired release tag. The CTE for Kubernetes node server cannot be automatically upgraded. Doing this will result in application failures for existing pods using CTE volumes on the nodes. You must use a manual procedure to stop the pods running on the nodes and delete the CTE for Kubernetes node server pods on that node.

Redeploy CTE for Kubernetes

To redeploy CTE for Kubernetes, type:

    ./deploy.sh -t <version#>.<build#>

For example:

    ./deploy.sh -t 1.0.0.1223

Drain and update a node

To upgrade the nodes:

1. Get a list of nodes currently running CTE for Kubernetes, type:

   kubectl get pods -n kube-system -o wide -l app=cte-csi-node --no-headers -o custom-columns=":spec.nodeName"
   

   Response Example:

   ub20-master
   ub20-work1
   

2. Drain a node from the list, type:

   kubectl drain --delete-emptydir-data --ignore-daemonsets <nodeName>
   

   Note

   The drain command pauses while it waits for the node to be drained.

   Response example:

   node/ub20-work1 cordoned
   node/ub20-work1 drained
   

3. Use a compound kubectl command to delete the CTE for Kubernetes node server on this node:

   kubectl delete -n kube-system pod $(kubectl get pods -n kube-system --field-selector spec.nodeName=<nodeName> -l app=cte-csi-node --no-headers -o custom-columns=":metadata.name")
   

   • The kubectl get pods needs to be bounded to only output the CTE for Kubernetes-node label for this node

   • Use --field-selector spec.nodeName=<nodeName> to specify the CTE for Kubernetes node to drain.

   • Use -l app=cte-csi-node to select only the CTE for Kubernetes node server from that node.

   • Use --no-headers -o custom-columns=":metadata.name" to tell kubectl to only output the pod names.

   • Pass the output of that command to the kubectl delete pod which triggers a new CTE for Kubernetes node server pod to be created with the desired image.

4. Verify that the pod has started successfully, type:

   kubectl get pods -n kube-system --field-selector spec.nodeName=<nodeName> -l app=cte-csi-node -o wide
   

   Response example:

   NAME                 READY   STATUS    RESTARTS   AGE   IP               NODE         NOMINATED NODE   READINESS GATES
   cte-csi-node-kgwkj   4/4     Running   0          14m   192.168.77.163   ub20-work1   <none>           <none>
   

5. Return node to working status, type:

   kubectl uncordon <nodeName>
   

   Response example:

   node/ub20-work1 uncordoned