csi: document how to use existing volumes

* Add Readme with examples on how to use existing volumes
* Add an option to prevent formatting an existing volume (needs more
testing)

KNOWN ISSUES:

* Deleting a `PVC` will delete the volume if the reclaim policy is set
to `Delete`, however the `PV` will be not deleted. This seems to be a
bug in K8S
* Using an existing, attached volume might cause confusion if the POD is
scheduled to a different node. It's advised that user only use existing
**detached** volumes

CHANGELOG.md

@@ -1,5 +1,10 @@
 ## unreleased
 
+* Add a [tutorial](examples/kubernetes/pod-single-existing-volume/README.md) on how to re-use an existing volume. Also a new option is introduced to prevent formatting an existing volume.
+  [[GH-87]](https://github.com/digitalocean/csi-digitalocean/pull/87)
+* Handle case if a volume is already attached to a droplet
+  [[GH-87]](https://github.com/digitalocean/csi-digitalocean/pull/87)
+
 ## v0.2.0 - 2018.09.05
 
 * Add support to CSI Spec `v0.3.0`. This includes many new changes, make sure 

README.md

@@ -187,7 +187,7 @@ $ make test
 If you want to test your changes, create a new image with the version set to `dev`:
 
 ```
-$ VERSION=dev make publish-dev
+$ VERSION=dev make publish
 ```
 
 This will create a binary with version `dev` and docker image pushed to

driver/controller.go

@@ -180,6 +180,10 @@ func (d *Driver) DeleteVolume(ctx context.Context, req *csi.DeleteVolumeRequest)
 	if err != nil {
 		if resp != nil && resp.StatusCode == http.StatusNotFound {
 			// we assume it's deleted already for idempotency
+			ll.WithFields(logrus.Fields{
+				"error": err,
+				"resp":  resp,
+			}).Warn("assuming volume is deleted already")
 			return &csi.DeleteVolumeResponse{}, nil
 		}
 		return nil, err
@@ -227,7 +231,7 @@ func (d *Driver) ControllerPublishVolume(ctx context.Context, req *csi.Controlle
 	ll.Info("controller publish volume called")
 
 	// check if volume exist before trying to attach it
-	_, resp, err := d.doClient.Storage.GetVolume(ctx, req.VolumeId)
+	vol, resp, err := d.doClient.Storage.GetVolume(ctx, req.VolumeId)
 	if err != nil {
 		if resp != nil && resp.StatusCode == http.StatusNotFound {
 			return nil, status.Errorf(codes.NotFound, "volume %q not found", req.VolumeId)
@@ -244,6 +248,22 @@ func (d *Driver) ControllerPublishVolume(ctx context.Context, req *csi.Controlle
 		return nil, err
 	}
 
+	attachedID := 0
+	for _, id := range vol.DropletIDs {
+		attachedID = id
+		if id == dropletID {
+			ll.Info("volume is already attached")
+			return &csi.ControllerPublishVolumeResponse{}, nil
+		}
+	}
+
+	// droplet is attached to a different node, return an error
+	if attachedID != 0 {
+		return nil, status.Errorf(codes.FailedPrecondition,
+			"volume is attached to a wrong droplet(%q), dettach the volume to fix it", attachedID)
+	}
+
+	// attach the volume to the correct node
 	action, resp, err := d.doClient.StorageActions.Attach(ctx, req.VolumeId, dropletID)
 	if err != nil {
 		// don't do anything if attached

examples/kubernetes/pod-single-existing-volume/README.md

@@ -0,0 +1,147 @@
+# Use an existing volume
+
+Below you will find the instruction on how to use an existing DigitalOcean Block
+Storage with your Kubernetes cluster.
+
+## Known issues
+
+* Deleting a `PVC` will delete the volume if the reclaim policy is set
+to `Delete`, however the bound `PV` will be not deleted. This seems to be a
+bug in K8S.
+* Using an existing, attached volume might cause confusion if the `POD` is
+scheduled to a different node. It's advised to use only existing **detached**
+volumes. If you have an attached volume, please make sure to detach it, so
+Kubernetes attaches it to the correct droplet. The CSI plugin will return an
+error if the volume is attached to a wrong droplet.
+
+## Example
+
+To use an existing volume, we have to create manually a `PersistentVolume` (PV)
+resource. Here is an example `PersistenVolume` resource for an existing volume:
+
+```
+kind: PersistentVolume
+apiVersion: v1
+metadata:
+  name: volume-nyc1-01
+  annotations:
+    # fake it by indicating this is provisioned dynamically, so the system
+    # works properly
+    pv.kubernetes.io/provisioned-by: com.digitalocean.csi.dobs
+spec:
+  storageClassName: do-block-storage
+  # by default, the volume will be not deleted if you delete the PVC, change to
+  # "Delete" if you wish the volume to be deleted automatically with the PVC
+  persistentVolumeReclaimPolicy: Delete
+  capacity:
+    storage: 5Gi
+  accessModes:
+    - ReadWriteOnce
+  csi:
+    driver: com.digitalocean.csi.dobs
+    fsType: ext4
+    volumeHandle: 1952d58a-c714-11e8-bc0c-0a58ac14421e
+    volumeAttributes:
+      com.digitalocean.csi/format: "true"
+```
+
+Couple of things to note,
+
+* `volumeHandle` is the volume ID you want to reuse. Make sure it matches exactly the volume you're targeting. You can list the ID's of your volumes via doctl: `doctl compute volume list`
+* `volumeAttributes` has a special, csi-digitalocean specific annotation called `com.digitalocean.csi/format`. If you add this key, the CSI plugin make sure to **not format** the volume. If you don't add this, it'll be formatted.
+* `storage` make sure it's set to the same storage size as your existing DigitalOcean Block Storage volume.
+
+Create a file with this content, naming it `pv.yaml` and deploy it:
+
+```
+kubectl create -f pv.yaml
+```
+
+View information about the `PersistentVolume`:
+
+```
+$ kubectl get pv volume-nyc1-01
+
+NAME             CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS      CLAIM     STORAGECLASS       REASON    AGE
+volume-nyc1-01   5Gi        RWO            Delete           Available             do-block-storage             15s
+```
+
+The status is `Available`. This means it has not yet been bound to a
+PersistentVolumeClaim. Now we can proceed to create our PVC:
+
+
+```
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: csi-pod-pvc
+spec:
+  accessModes:
+  - ReadWriteOnce
+  resources:
+    requests:
+      storage: 5Gi
+  storageClassName: do-block-storage
+```
+
+This is the same (just like our other examples). When you create `PVC`,
+Kubernetes will try to match it to an existing `PV`. Because the
+storageClassName and the storage size matches our `PV` descriptions, kubernetes
+will bind this `PVC` to our manually create `PV`. CSI **will not** create a new
+volume because of the existing `PV`.
+
+Create the PersistentVolumeClaim:
+
+```
+kubectl create -f pvc.yaml
+```
+
+Now look at the PersistentVolumeClaim (PVC):
+
+```
+kubectl get pvc task-pv-claim
+NAME          STATUS    VOLUME           CAPACITY   ACCESS MODES   STORAGECLASS       AGE
+csi-pod-pvc   Bound     volume-nyc1-01   5Gi        RWO            do-block-storage   5s
+```
+
+As you see, the output shows that the PVC is bound to our PersistentVolume, `volume-nyc1-01`.
+
+Finally, define your pod that refers to this PVC:
+
+```
+kind: Pod
+apiVersion: v1
+metadata:
+  name: my-csi-app
+spec:
+  containers:
+    - name: my-frontend
+      image: busybox
+      volumeMounts:
+      - mountPath: "/data"
+        name: my-do-volume
+      command: [ "sleep", "1000000" ]
+  volumes:
+    - name: my-do-volume
+      persistentVolumeClaim:
+        claimName: csi-pod-pvc 
+```
+
+
+Check if the pod is running successfully:
+
+
+```
+$ kubectl describe pods/my-csi-app
+```
+
+Write inside the app container:
+
+```
+$ kubectl exec -ti my-csi-app /bin/sh
+/ # touch /data/hello-world
+/ # exit
+$ kubectl exec -ti my-csi-app /bin/sh
+/ # ls /data
+hello-world
+```

examples/kubernetes/pod-single-existing-volume/pod.yaml

@@ -0,0 +1,16 @@
+kind: Pod
+apiVersion: v1
+metadata:
+  name: my-csi-app
+spec:
+  containers:
+    - name: my-frontend
+      image: busybox
+      volumeMounts:
+      - mountPath: "/data"
+        name: my-do-volume
+      command: [ "sleep", "1000000" ]
+  volumes:
+    - name: my-do-volume
+      persistentVolumeClaim:
+        claimName: csi-pod-pvc 

examples/kubernetes/pod-single-existing-volume/pv.yaml

@@ -0,0 +1,23 @@
+kind: PersistentVolume
+apiVersion: v1
+metadata:
+  name: volume-nyc1-01
+  annotations:
+    # fake it by indicating this is provisioned dynamically, so the system
+    # works properly
+    pv.kubernetes.io/provisioned-by: com.digitalocean.csi.dobs
+spec:
+  storageClassName: do-block-storage
+  # by default, the volume will be not deleted if you delete the PVC, change to
+  # "Delete" if you wish the volume to be deleted automatically with the PVC
+  persistentVolumeReclaimPolicy: Delete
+  capacity:
+    storage: 5Gi
+  accessModes:
+    - ReadWriteOnce
+  csi:
+    driver: com.digitalocean.csi.dobs
+    fsType: ext4
+    volumeHandle: 1952d58a-c714-11e8-bc0c-0a58ac14421e
+    volumeAttributes:
+      com.digitalocean.csi/noformat: "true"

examples/kubernetes/pod-single-existing-volume/pvc.yaml

@@ -0,0 +1,11 @@
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: csi-pod-pvc
+spec:
+  accessModes:
+  - ReadWriteOnce
+  resources:
+    requests:
+      storage: 5Gi
+  storageClassName: do-block-storage