如何为Kubernetes配置'efs-provider'？

Question

I have followed the steps from this guide this guide to deploy the efs-provider for Kubernetes and bind an EFS filesystem. I have not succed.

I am implementing Kubernetes with Amazon EKS and I use EC2 instances as worker nodes, all are deployed using eksctl .

After I applied this adjusted manifest file , the result is:

$ kubectl get pods
NAME                               READY   STATUS    RESTARTS
efs-provisioner-#########-#####   1/1     Running   0       

$ kubectl get pvc
NAME       STATUS    VOLUME
test-pvc   Pending   efs-storage

No matter how much time I wait, the status of my PVC stucks in Pending .

After the creation of a Kubernetes cluster and worker nodes and the configuration of the EFS filesystem, I apply the efs-provider manifest with all the variables pointing to the EFS filesystem. In the StorageClass configuration file is specified the spec.AccessModes field as ReadWriteMany .

At this point my efs-provider pod is running without errors and the status of the PVC is Pending . What can it be? How can I configure the efs-provider to use the EFS filesystem? How much should I wait to get the PVC status in Bound ?

---

Update

About the configuration of the Amazon Web Services, these is what I have done:

• After the creation of the EFS filesystem, I have created a mount point for each subnet where my nodes are.
• To each mount point is attached a security group with a inbound rule to grant the access to the NFS port (2049) from the security group of each nodegroup.

The description of my EFS security group is:

{
    "Description": "Communication between the control plane and worker nodes in cluster",
    "GroupName": "##################",
    "IpPermissions": [
        {
        "FromPort": 2049,
        "IpProtocol": "tcp",
        "IpRanges": [],
        "Ipv6Ranges": [],
        "PrefixListIds": [],
        "ToPort": 2049,
        "UserIdGroupPairs": [
            {
            "GroupId": "sg-##################",
            "UserId": "##################"
            }
        ]
        }
    ],
    "OwnerId": "##################",
    "GroupId": "sg-##################",
    "IpPermissionsEgress": [
        {
        "IpProtocol": "-1",
        "IpRanges": [
            {
            "CidrIp": "0.0.0.0/0"
            }
        ],
        "Ipv6Ranges": [],
        "PrefixListIds": [],
        "UserIdGroupPairs": []
        }
    ],
    "VpcId": "vpc-##################"
}

Deployment

The output of the kubectl describe deploy ${DEPLOY_NAME} command is:

$ DEPLOY_NAME=efs-provisioner; \
> kubectl describe deploy ${DEPLOY_NAME}
Name:               efs-provisioner
Namespace:          default
CreationTimestamp:  ####################
Labels:             app=efs-provisioner
Annotations:        deployment.kubernetes.io/revision: 1
kubectl.kubernetes.io/last-applied-configuration:
{"apiVersion":"extensions/v1beta1","kind":"Deployment","metadata":{"annotations":{},"name":"efs-provisioner","namespace":"default"},"spec"...
Selector:           app=efs-provisioner
Replicas:           1 desired | 1 updated | 1 total | 1 available | 0 unavailable
StrategyType:       Recreate
MinReadySeconds:    0
Pod Template:
Labels:           app=efs-provisioner
Service Account:  efs-provisioner
Containers:
efs-provisioner:
Image:      quay.io/external_storage/efs-provisioner:latest
Port:       <none>
Host Port:  <none>
Environment:
FILE_SYSTEM_ID:    <set to the key 'file.system.id' of config map 'efs-provisioner'>    Optional: false
AWS_REGION:        <set to the key 'aws.region' of config map 'efs-provisioner'>        Optional: false
DNS_NAME:          <set to the key 'dns.name' of config map 'efs-provisioner'>          Optional: true
PROVISIONER_NAME:  <set to the key 'provisioner.name' of config map 'efs-provisioner'>  Optional: false
Mounts:
/persistentvolumes from pv-volume (rw)
Volumes:
pv-volume:
Type:      NFS (an NFS mount that lasts the lifetime of a pod)
Server:    fs-#########.efs.##########.amazonaws.com
Path:      /
ReadOnly:  false
Conditions:
Type           Status  Reason
----           ------  ------
Available      True    MinimumReplicasAvailable
OldReplicaSets:  <none>
NewReplicaSet:   efs-provisioner-576c67cf7b (1/1 replicas created)
Events:
Type    Reason             Age   From                   Message
----    ------             ----  ----                   -------
Normal  ScalingReplicaSet  106s  deployment-controller  Scaled up replica set efs-provisioner-576c67cf7b to 1

Pod Logs

The output of the kubectl logs ${POD_NAME} command is:

$ POD_NAME=efs-provisioner-576c67cf7b-5jm95; \
> kubectl logs ${POD_NAME}
E0708 16:03:46.841229       1 efs-provisioner.go:69] fs-#########.efs.##########.amazonaws.com
I0708 16:03:47.049194       1 leaderelection.go:187] attempting to acquire leader lease  default/kubernetes.io-aws-efs...
I0708 16:03:47.061830       1 leaderelection.go:196] successfully acquired lease default/kubernetes.io-aws-efs
I0708 16:03:47.062791       1 controller.go:571] Starting provisioner controller kubernetes.io/aws-efs_efs-provisioner-576c67cf7b-5jm95_f7c5689f-a199-11e9-a152-def1285e1be5!
I0708 16:03:47.062877       1 event.go:221] Event(v1.ObjectReference{Kind:"Endpoints", Namespace:"default", Name:"kubernetes.io-aws-efs", UID:"f7c682cd-a199-11e9-80bd-1640944916e4", APIVersion:"v1", ResourceVersion:"3914", FieldPath:""}): type: 'Normal' reason: 'LeaderElection' efs-provisioner-576c67cf7b-5jm95_f7c5689f-a199-11e9-a152-def1285e1be5 became leader
I0708 16:03:47.162998       1 controller.go:620] Started provisioner controller kubernetes.io/aws-efs_efs-provisioner-576c67cf7b-5jm95_f7c5689f-a199-11e9-a152-def1285e1be5!

StorageClass

The output of the kubectl describe sc ${STORAGE_CLASS_NAME} command is:

$ STORAGE_CLASS_NAME=aws-efs; \
> kubectl describe sc ${STORAGE_CLASS_NAME}
Name:            aws-efs
IsDefaultClass:  No
Annotations:     kubectl.kubernetes.io/last-applied-configuration={"apiVersion":"storage.k8s.io/v1","kind":"StorageClass","metadata":{"annotations":{},"name":"aws-efs"},"provisioner":"aws-efs"}
Provisioner:           aws-efs
Parameters:            <none>
AllowVolumeExpansion:  <unset>
MountOptions:          <none>
ReclaimPolicy:         Delete
VolumeBindingMode:     Immediate
Events:                <none>

PersistentVolumeClaim

The output of the kubectl describe pvc ${PVC_NAME} command is:

$ PVC_NAME=efs; \
> kubectl describe pvc ${PVC_NAME}
Name:          efs
Namespace:     default
StorageClass:  aws-efs
Status:        Pending
Volume:
Labels:        <none>
Annotations:   kubectl.kubernetes.io/last-applied-configuration:
{"apiVersion":"v1","kind":"PersistentVolumeClaim","metadata":{"annotations":{"volume.beta.kubernetes.io/storage-class":"aws-efs"},"name":"...
volume.beta.kubernetes.io/storage-class: aws-efs
Finalizers:    [kubernetes.io/pvc-protection]
Capacity:
Access Modes:
Events:
Type       Reason              Age                 From                         Message
----       ------              ----                ----                         -------
Warning    ProvisioningFailed  43s (x12 over 11m)  persistentvolume-controller  no volume plugin matched
Mounted By:  <none>

About the questions

1. Do you have the EFS filesystem id properly configured for your efs-provisioner ?

   • Yes, both (from the fs and the configured) match.

2. Do you have the proper IAM credentials to access this EFS?

   • Yes, my user has and also the eksctl tool configures it.

3. Does that EFS path specified for your provisioner exist?

   • Yes, is only the root (/) path.

4. Did you add an EFS endpoint to the subnet that your worker node(s) are running on, or ensure that your EFS subnets have an Internet Gateway attached?

   • Yes, I have added the EFS endpoints to the subnet that my worker node(s) are running on.

5. Did you set your security group to allow the Inbound for NFS port(s)?

   • Yes.

Answer 1

I have solved my issue by replacing the provisioner name of my StorageClass from kubernetes.io/aws-efs to only aws-efs .

As we can read on this issue comment on Github posted by wongma7 :

The issue is that provisioner is kubernetes.io/aws-efs . It can't begin with kubernetes.io as that is reserved by kubernetes.

That solves the ProvisioningFailed on the Events produced on the PersistentVolumeClaim by the persistentvolume-controller .