In TKGI environment where out-of-tree (EBS) CSI driver is installed following the doc below errors are observed on ebs-csi-node
pod logs.
$ kubectl logs ebs-csi-node-xxx -n kube-system
Defaulted container "ebs-plugin" out of: ebs-plugin, node-driver-registrar, liveness-probe
I1203 09:02:27.936582 1 main.go:153] "Initializing metadata"
I1203 09:02:31.453210 1 metadata.go:48] "Retrieved metadata from IMDS"
I1203 09:02:31.454222 1 driver.go:69] "Driver Information" Driver="ebs.csi.aws.com" Version="v1.37.0"
E1203 09:02:32.865556 1 node.go:857] "Unexpected failure when attempting to remove node taint(s)" err="isAllocatableSet: driver not found on node
E1203 09:02:33.455947 1 node.go:857] "Unexpected failure when attempting to remove node taint(s)" err="isAllocatableSet: driver not found on node
E1203 09:02:34.509197 1 node.go:857] "Unexpected failure when attempting to remove node taint(s)" err="isAllocatableSet: driver not found on node
E1203 09:02:36.553760 1 node.go:857] "Unexpected failure when attempting to remove node taint(s)" err="isAllocatableSet: driver not found on node
On reviewing the csi node configuration spec.drivers
is null
$ kubectl get csinodes -A -o yaml
apiVersion: v1
items:
apiVersion: storage.k8s.io/v1
kind: CSINode
...
spec:
drivers: null
TKGI on AWS
kubelet path in TKGi is /var/vcap/data/kubelet
, but the manifest in these csi driver uses /var/lib/kubelet,
so csi-driver fails to register to kubelet.
Please add option "--set linux.kubelet=/var/vcap/data/kubelet"
for helm install.
Please replace "/var/lib/kubelet"
with "/var/vcap/data/kubelet"
in the manifest for manual install.
Same resolution applies for Azure environment when installing csi following below documents.
azuredisk-csi-driver
azurefile-csi-driver