Running a stateful container on Storwize systems

This information provides an example of running a stateful container on Storwize systems using the IBM block storage CSI driver.

About this task

This example illustrates a basic configuration required for running a stateful container with volumes provisioned on a Storwize and Spectrum Virtualize storage services.

Note: This example can be used for any Storwize, Spectrum Virtualize, or FlashSystem as Storwize storage service.

Creating a k8s secret storwize-array1 for the storage system.
Creating a storage class gold.
Creating a PersistentVolumeClaim (PVC) demo-pvc that uses the storage class gold and show some details on the created PVC and persistent volume (PV).
Creating a StatefulSet application demo-statefulset and observing the mountpoint / multipath device that was created by the driver.
Writing data inside demo-statefulset, and then deleting and recreating demo-statefulset, verifying that the data still exists.

Procedure

Open a command-line terminal.

Create an array secret.

$> cat demo-secret-storwize-array1.yaml
kind: Secret
apiVersion: v1
metadata:
  name: storwize-array1
  namespace: kube-system
type: Opaque
stringData:
   management_address: <VALUE-2,VALUE-3> # replace with valid storage system managment address
   username: <VALUE-4>                   # replace with valid username
data:
   password: <VALUE-5 base64>            # replace with valid password
  
$> kubectl create -f demo-secret-storwize-array1.yaml
secret/storwize-array1 created

Create a storage class.

$> cat demo-storageclass-gold-storwize.yaml
kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: gold
  
provisioner: block.csi.ibm.com
parameters:
  SpaceEfficiency: thick # SpaceEfficiency values are: thick, thin, deduplicated, and compressed.
  pool: gold

  csi.storage.k8s.io/provisioner-secret-name: storwize-array1
  csi.storage.k8s.io/provisioner-secret-namespace: kube-system
  csi.storage.k8s.io/controller-publish-secret-name: storwize-array1
  csi.storage.k8s.io/controller-publish-secret-namespace: kube-system

  csi.storage.k8s.io/fstype: xfs   # Optional. Values are ext4/xfs. The default is ext4.
  volume_name_prefix: demo1        # Optional.

$> kubectl create -f demo-storageclass-gold-storwize.yaml
storageclass.storage.k8s.io/gold created

Create a PVC demo-pvc-gold with the size of 1 Gb.

$> cat demo-pvc-gold.yaml
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: pvc-demo
spec:
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi
  storageClassName: gold


$> kubectl apply -f demo-pvc-gold.yaml
persistentvolumeclaim/demo-pvc created

Display the existing PVC and the created persistent volume (PV).

$> kubectl get pv,pvc
NAME                                                        CAPACITY   ACCESS MODES
persistentvolume/pvc-a04bd32f-bd0f-11e9-a1f5-005056a45d5f   1Gi        RWO

RECLAIM POLICY   STATUS   CLAIM              STORAGECLASS   REASON   AGE
Delete           Bound    default/demo-pvc   gold                    78s

NAME                             STATUS   VOLUME                                     CAPACITY   
persistentvolumeclaim/demo-pvc   Bound    pvc-a04bd32f-bd0f-11e9-a1f5-005056a45d5f   1Gi

ACCESS MODES   STORAGECLASS   AGE
RWO            gold           78s

$> kubectl describe persistentvolume/pvc-a04bd32f-bd0f-11e9-a1f5-005056a45d5f
Name:            pvc-a04bd32f-bd0f-11e9-a1f5-005056a45d5f
Labels:          <none>
Annotations:     pv.kubernetes.io/provisioned-by: block.csi.ibm.com
Finalizers:      [kubernetes.io/pv-protection]
StorageClass:    gold
Status:          Bound
Claim:           default/demo-pvc
Reclaim Policy:  Delete
Access Modes:    RWO
VolumeMode:      Filesystem
Capacity:        1Gi
Node Affinity:   <none>
Message:         
Source:
    Type:              CSI (a Container Storage Interface (CSI) volume source)
    Driver:            block.csi.ibm.com
    VolumeHandle:      SVC:6001738CFC9035EB0000000000D1F68F
    ReadOnly:          false
    VolumeAttributes:      array_name=<IP>
                           pool_name=gold
                           storage.kubernetes.io/csiProvisionerIdentity=1565550204603-8081-
                           block.csi.ibm.com
                           storage_type=SVC
                           volume_name=demo1_pvc-a04bd32f-bd0f-11e9-a1f5-005056a45d5f
Events:                <none>

Create a StatefulSet application demo-statefulset, using the demo-pvc.

$> cat demo-statefulset-with-demo-pvc.yml
kind: StatefulSet
apiVersion: apps/v1
metadata:
  name: demo-statefulset
spec:
  selector:
    matchLabels:
      app: demo-statefulset
  serviceName: demo-statefulset
  replicas: 1
  template:
    metadata:
      labels:
        app: demo-statefulset
    spec:
      containers:
      - name: container1
        image: registry.access.redhat.com/ubi8/ubi:latest
        command: [ "/bin/sh", "-c", "--" ]
        args: [ "while true; do sleep 30; done;" ]
        volumeMounts:
          - name: demo-pvc
            mountPath: "/data"
      volumes:
      - name: demo-pvc
        persistentVolumeClaim:
          claimName: demo-pvc

      #nodeSelector:
      #  kubernetes.io/hostname: NODESELECTOR


$> kubectl create -f demo-statefulset-with-demo-pvc.yml
statefulset/demo-statefulset created

Check the newly created pod.

Display the newly created pod (make sure the pod status is Running).

$> kubectl get pod demo-statefulset-0
NAME                 READY   STATUS    RESTARTS   AGE
demo-statefulset-0   1/1     Running   0          43s

Check the mountpoint inside the pod.

$> kubectl exec demo-statefulset-0 -- bash -c "df -h /data"
Filesystem          Size  Used Avail Use% Mounted on
/dev/mapper/mpathz 1014M   33M  982M   4% /data

$> kubectl exec demo-statefulset-0 -- bash -c "mount | grep /data"
/dev/mapper/mpathz on /data type xfs (rw,relatime,seclabel,attr2,inode64,noquota)

Write data to the persistent volume of the pod.

The PV should be mounted inside the pod at /data.

$> kubectl exec demo-statefulset-0 touch /data/FILE
$> kubectl exec demo-statefulset-0 ls /data/FILE
File

Log into the worker node that has the running pod and display the newly attached volume on the node.

Verify which worker node is running the pod demo-statefulset-0.

$> kubectl describe pod demo-statefulset-0| grep "^Node:"
Node: k8s-node1/hostname

Establish an SSH connection and log into the worker node.
```
$> ssh k8s-node1
```

List the multipath devices on the worker node. Note the same mpathz, as mentioned in step 7.b.

$>[k8s-node1]  multipath -ll
mpathz (36001738cfc9035eb0000000000d1f68f) dm-3 IBM     ,2145 (for SVC)         
size=1.0G features='1 queue_if_no_path' hwhandler='0' wp=rw
`-+- policy='service-time 0' prio=1 status=active
  |- 37:0:0:12 sdc 8:32 active ready running
  `- 36:0:0:12 sdb 8:16 active ready running

$>[k8s-node1] ls -l /dev/mapper/mpathz
lrwxrwxrwx. 1 root root 7 Aug 12 19:29 /dev/mapper/mpathz -> ../dm-3

List the physical devices of the multipath mpathz and its mountpoint on the host. (This is the /data inside the stateful pod).

$>[k8s-node1]  lsblk /dev/sdb /dev/sdc
NAME     MAJ:MIN RM SIZE RO TYPE  MOUNTPOINT
sdb        8:16   0   1G  0 disk  
└─mpathz 253:3    0   1G  0 mpath /var/lib/kubelet/pods/d67d22b8-bd10-11e9-a1f5-005056a45d5f/volumes/kubernetes.io~csi/pvc-a04bd32f-bd0f-11e9-a1f5
sdc        8:32   0   1G  0 disk  
└─mpathz 253:3    0   1G  0 mpath /var/lib/kubelet/pods/d67d22b8-bd10-11e9-a1f5-005056a45d5f/volumes/kubernetes.io~csi/pvc-a04bd32f-bd0f-11e9-a1f5

View the PV mounted on this host.

Note: All PV mountpoints look like: /var/lib/kubelet/pods/*/volumes/kubernetes.io~csi/pvc-*/mount

$>[k8s-node1]  df | egrep pvc
/dev/mapper/mpathz      1038336    32944   1005392   4% /var/lib/kubelet/pods/d67d22b8-bd10-11e9-a1f5-005056a45d5f/volumes/kubernetes.io~csi/pvc-a04bd32f-bd0f-11e9-a1f5-005056a45d5f/mount

Details about the driver internal metadata file .stageInfo.json is stored in the k8s PV node stage path /var/lib/kubelet/plugins/kubernetes.io/csi/pv/<PVC-ID>/globalmount/.stageInfo.json. The CSI driver creates the metadata file during the NodeStage API and is used at later stages by the NodePublishVolume, NodeUnPublishVolume and NodeUnStage CSI APIs later on.
```
$> cat /var/lib/kubelet/plugins/kubernetes.io/csi/pv/pvc-711b6fef-bcf9-11e9-a1f5-005056a45d5f/globalmount/.stageInfo.json
{"connectivity":"iscsi","mpathDevice":"dm-3","sysDevices":",sdb,sdc"}
```

Delete StatefulSet and then restart, in order to validate data (/data/FILE) remains in the persistent volume.

$> kubectl delete statefulset/demo-statefulset
statefulset/demo-statefulset deleted

### Wait until the pod is deleted. Once deleted the '"demo-statefulset" not found' is returned.
$> kubectl get statefulset/demo-statefulset
NAME                 READY   STATUS        RESTARTS   AGE
demo-statefulset-0   0/1     Terminating   0          91m


###### Establish an SSH connection and log into the worker node in order to see that the multipath was deleted and that the PV mountpoint no longer exists.

$> ssh k8s-node1

$>[k8s-node1] df | egrep pvc
$>[k8s-node1] multipath -ll
$>[k8s-node1] lsblk /dev/sdb /dev/sdc
lsblk: /dev/sdb: not a block device
lsblk: /dev/sdc: not a block device


###### Recreate the statefulset and verify that /data/FILE exists.
$> kubectl create -f demo-statefulset-with-demo-pvc.yml
statefulset/demo-statefulset created

$> kubectl exec demo-statefulset-0 ls /data/FILE
File

Delete StatefulSet and the PVC.

$> kubectl delete statefulset/demo-statefulset
statefulset/demo-statefulset deleted

$> kubectl get statefulset/demo-statefulset
No resources found.

$> kubectl delete pvc/demo-pvc
persistentvolumeclaim/demo-pvc deleted

$> kubectl get pv,pvc
No resources found.