longhorn-driver-deployer error: Node is not support mount propagation
If longhorn-instance-manager image name is more than 63 characters long, it will fail to deploy, and longhorn-driver-deployer pod will be in CrashLoopBackOff.
Checking Longhorn driver deployer logs will report the following:
time="2020-03-13T22:49:22Z" level=warning msg="Got an error when checking MountPropagation with node status, Node XXX is not support mount propagation"
time="2020-03-13T22:49:22Z" level=fatal msg="Error deploying driver: CSI cannot be deployed because MountPropagation is not set: Node <NODE_NAME> is not support mount propagation"
Issue can be conformed by checking Longhorn manager log, you should be able to see the following logs:
“Dropping Longhorn node longhorn-system/NODE_NAME out of the queue: fail to sync node for longhorn-system/NODE_NAME:
InstanceManager.longhorn.io "instance-manager-e-605e9473" is invalid: metadata.labels: Invalid value:
"PRIVATE_REGISTRY_URL-PREFIX-longhorn-instance-manager-v1_20200301": must be no more than 63 characters”
Longhorn instance manager: metadata.labels must be no more than 63 characters
Using a long registry URL may cause Longhorn installation error
Longhorn manager would report errors in the log when this happened:
"instance-manager-e-xxxxxxxx" is invalid: metadata.labels: Invalid value: "<PRIVATE_REGISTRY_URL>-longhornio-longhorn-instance-manager-v1_20200301": must be no more than 63 characters
Longhorn instance manager pods have labels with key longhorn.io/instance-manager-image and value REGISTRY_URL-USER-IMAGE_NAME-TAG