Related Issues
Test Longhorn Upgrade with Pre-existing RWX Volume Workloads and Storage Network Configured
Verify the behavior of Longhorn RWX volume workloads during and after the cluster upgrade process with existing storage network configured. Ensure no disruption to existing workload pods during the upgrade.
Given A cluster with Longhorn v1.6.2 installed.
> kubectl -n longhorn-system get daemonsets.apps longhorn-manager -o yaml | grep image:
image: longhornio/longhorn-manager:v1.6.2
And The storage-network setting is set to a valid NAD.
> kubectl -n longhorn-system get setting storage-network
NAME VALUE AGE
storage-network kube-system/demo-192-168-0-0 66s
And The auto-delete-pod-when-volume-detached-unexpectedly setting is set to true.
> kubectl -n longhorn-system get setting auto-delete-pod-when-volume-detached-unexpectedly
NAME VALUE AGE
auto-delete-pod-when-volume-detached-unexpectedly true 85s
And RWX volume workload 1 is created.
> kubectl get pod
NAME READY STATUS RESTARTS AGE
rwx-1-6c6dd9764c-jrqvh 2/2 Running 0 98s
rwx-1-6c6dd9764c-k66f8 2/2 Running 0 98s
rwx-1-6c6dd9764c-rqnkf 2/2 Running 0 98s
When update Longhorn to v1.7.0.
Then Longhorn CSI plugin pods are not annotated with the storage network.
> kubectl -n longhorn-system get pod longhorn-csi-plugin-682cr -o yaml | grep k8s.v1.cni.cncf.io/networks
And workload 1 is not restarted.
> kubectl -n longhorn-system get pod -l app=longhorn-csi-plugin
NAME READY STATUS RESTARTS AGE
longhorn-csi-plugin-682cr 3/3 Running 0 22s
longhorn-csi-plugin-bkbbr 3/3 Running 0 22s
longhorn-csi-plugin-dkzjz 3/3 Running 0 22s
> kubectl get pod
NAME READY STATUS RESTARTS AGE
rwx-1-6c6dd9764c-jrqvh 2/2 Running 0 3m32s
rwx-1-6c6dd9764c-k66f8 2/2 Running 0 3m32s
rwx-1-6c6dd9764c-rqnkf 2/2 Running 0 3m32s
And workload 1 data is accessible.
> kubectl exec -it rwx-1-6c6dd9764c-jrqvh -- ls /data
Defaulted container "run" out of: run, nginx
index.html lost+found