cluster-operator

StorageOS cluster-operator

The StorageOS Cluster Operator deploys and configures a StorageOS cluster on Kubernetes.

For quick installation of cluster operator, use the cluster operator helm chart.

Pre-requisites

• Kubernetes 1.9+
• Kubernetes must be configured to allow (configured by default in 1.10+):
  • Privileged mode containers (enabled by default)
  • Feature gate: MountPropagation=true. This can be done by appending --feature-gates MountPropagation=true to the kube-apiserver and kubelet services.

Refer to the StorageOS prerequisites docs for more information.

Setup/Development:

1. Install operator-sdk.
2. Run operator-sdk generate k8s if there's a change in api type.
3. Build operator container with operator-sdk build storageos/cluster-operator:<tag>
4. Apply the manifests in deploy/ to install the operator
   • Apply service_account.yaml and role_binding.yaml to create a service account and to grant all the permissions.
   • Apply crds/*_storageoscluster_crd.yaml to define the custom resources.
   • Apply operator.yaml to install the operator. Change the container image in this file when installing a new operator.
   • Apply crds/*_storageoscluster_cr.yaml to create a StorageOSCluster custom resource.

NOTE: Installing StorageOS on Minikube is not currently supported due to missing kernel prerequisites.

For development, run the operator outside of the k8s cluster by running:

$ make local-run

Build operator container image:

$ make image/cluster-operator OPERATOR_IMAGE=storageos/cluster-operator:test

This builds all the components and copies the binaries into the same container.

After creating a resource, query the resource:

$ kubectl get storageoscluster
NAME                READY     STATUS    AGE
example-storageos   3/3       Running   4m

Inspect a StorageOSCluster Resource

Get all the details about the cluster:

$ kubectl describe storageoscluster/example-storageos
Name:         example-storageos
Namespace:    default
Labels:       <none>
Annotations:  kubectl.kubernetes.io/last-applied-configuration={"apiVersion":"storageos.com/v1alpha1","kind":"StorageOSCluster","metadata":{"annotations":{},"name":"example-storageos","namespace":"default"},"spec":{"...
API Version:  storageos.com/v1alpha1
Kind:         StorageOSCluster
Metadata:
  Creation Timestamp:  2018-07-21T12:57:11Z
  Generation:          1
  Resource Version:    10939030
  Self Link:           /apis/storageos.com/v1alpha1/namespaces/default/storageosclusters/example-storageos
  UID:                 955b24a4-8ce5-11e8-956a-1866da35eee2
Spec:
  Join:  test07
Status:
  Node Health Status:
  ...
  ...
  Nodes:
    test09
    test08
    test07
  Phase:  Running
  Ready:  3/3
Events:   <none>

StorageOSCluster Resource Configuration

Once the StorageOS operator is running, a StorageOS cluster can be deployed by creating a Cluster Configuration. The parameters specified in the configuration will define how StorageOS is deployed, the rest of the installation details are handled by the operator.

The following tables lists the configurable spec parameters of the StorageOSCluster custom resource and their default values.

Upgrading a StorageOS Cluster

An existing StorageOS cluster can be upgraded to a new version of StorageOS by creating an Upgrade Configuration. The cluster-operator takes care of downloading the new container image and updating all the nodes with new version of StorageOS. An example of StorageOSUpgrade resource is storageos_v1alpha1_storageosupgrade_cr.yaml.

Only offline upgrade is supported for now by cluster-operator. During the upgrade, StorageOS maintenance mode is enabled, the applications that use StorageOS volumes are scaled down and the whole StorageOS cluster is restarted with a new version. Once the StorageOS cluster becomes usable, the applications are scaled up to their previous configuration. Once the update is complete, make sure to delete the upgrade resource to put the StorageOS cluster in normal mode. This will disable the maintenance mode.

Once an upgrade resource is created, events related to the upgrade can be viewed in the upgrade object description. All the status and errors, if any, encountered during the upgrade are posted as events.

$ kubectl describe storageosupgrades example-storageosupgrade
Name:         example-storageosupgrade
Namespace:    default
Labels:       <none>
Annotations:  kubectl.kubernetes.io/last-applied-configuration={"apiVersion":"storageos.com/v1alpha1","kind":"StorageOSUpgrade","metadata":{"annotations":{},"name":"example-storageosupgrade","namespace":"default"},...
API Version:  storageos.com/v1alpha1
Kind:         StorageOSUpgrade
...
Spec:
  New Image:  storageos/node:1.0.0
Events:
  Type    Reason           Age   From                Message
  ----    ------           ----  ----                -------
  Normal  PullImage         4m    storageos-upgrader  Pulling the new container image
  Normal  PauseClusterCtrl  2m    storageos-upgrader  Pausing the cluster controller and enabling cluster maintenance mode
  Normal  UpgradeInit       2m    storageos-upgrader  StorageOS upgrade of cluster example-storageos started
  Normal  UpgradeComplete   0s    storageos-upgrader  StorageOS upgraded to storageos/node:1.0.0. Delete upgrade object to disable cluster maintenance mode

StorageOSUpgrade Resource Configuration

The following table lists the configurable spec parameters of the StorageOSUpgrade custom resource and their default values.

Cleanup Old Configurations

StorageOS creates and saves its files at /var/lib/storageos on the hosts. This also contains some configurations of the cluster. To do a fresh install of StorageOS, these files need to be deleted.

WARNING: This will delete any existing data and won't be recoverable.

NOTE: When using an external etcd, the data related to storageos should also be removed.

ETCDCTL_API=3 /usr/local/bin/etcdctl --endpoints http://storageos-etcd-server:2379 del --prefix storageos

The cluster-operator provides a Jobresources, that can execute certain tasks on all the nodes or on some selected nodes. This can be used to easily perform cleanup task. An example would be to create a Job resource:

apiVersion: storageos.com/v1alpha1
kind: Job
metadata:
  name: cleanup-job
spec:
  image: darkowlzz/cleanup:v0.0.2
  args: ["/var/lib/storageos"]
  mountPath: "/var/lib"
  hostPath: "/var/lib"
  completionWord: "done"
  nodeSelectorTerms:
    - matchExpressions:
      - key: node-role.kubernetes.io/worker
        operator: In
        values:
        - "true"

When applied, this job will run darkowlzz/cleanup container on the nodes that have label node-role.kubernetes.io/worker with value "true", mounting /var/lib and passing the argument /var/lib/storageos. This will run rm -rf /var/lib/storageos in the selected nodes and cleanup all the storageos files. To run it on all the nodes, remove the nodeSelectorTerms attribute. On completion, the resource description shows that the task is completed and can be deleted.

$ kubectl describe jobs.storageos.com cleanup-job
Name:         cleanup-job
Namespace:    default
...
...
Spec:
  Completion Word:  
  Args:
    /var/lib/storageos
  Host Path:            /var/lib
  Image:                darkowlzz/cleanup:v0.0.2
  ...
Status:
  Completed:  true
Events:
  Type    Reason        Age   From                       Message
  ----    ------        ----  ----                       -------
  Normal  JobCompleted  39s   storageoscluster-operator  Job Completed. Safe to delete.

Deleting the resource, will terminate all the pods that were created to run the task.

Internally, this Job is backed by a controller that creates pods using DaemonSet. And the job containers have to be built in a specific way, in order to achieve this behavior.

In the above example, the cleanup container runs a shell script(script.sh):

#!/bin/ash

set -euo pipefail

# Gracefully handle the TERM signal sent when deleting the daemonset
trap 'exit' TERM

# This is the main command that's run by this script on
# all the nodes.
rm -rf $1

# Let the monitoring script know we're done.
echo "done"

# this is a workaround to prevent the container from exiting
# and k8s restarting the daemonset pod
while true; do sleep 1; done

And the container image is made with Dockerfile:

FROM alpine:3.6
COPY script.sh .
RUN chmod u+x script.sh
ENTRYPOINT ["./script.sh"]

The script, after running the main command, enters into a sleep state, instead of exiting. This is needed because we don't want the container to exit and start again and again. Once completed, it echos "done". This is read by the Job controller to figure out when the task is completed. Once all the pods have completed the task, the Job status is completed and it can be deleted.

This can be extended to do other similar cluster management operations. This is also used internally in the cluster upgrade process.

Job(jobs.storageos.com) Resource Configuration

The following table lists the configurable spec parameters of the Job custom resource and their default values.

TLS Support

To enable TLS, ensure that an ingress controller is installed in the cluster. Set ingress.enable and ingress.tls to true. Store the TLS cert and key as part of the storageos secret as:

apiVersion: v1
kind: Secret
metadata:
  name: "storageos-api"
...
...
data:
  # echo -n '<secret>' | base64
  ...
  ...
  # Add base64 encoded TLS cert and key.
  tls.crt:
  tls.key:

CSI

StorageOS also supports the Container Storage Interface (CSI) to communicate with Kubernetes. Only versions 1.10+ are supported. CSI ensures forward compatibility with future releases of Kubernetes, as vendor-specific drivers will soon be deprecated from Kubernetes. However, some functionalities are not supported yet.

To enable CSI, set csi.enable to true in the StorageOSCluster resource config.

apiVersion: "storageos.com/v1alpha1"
kind: "StorageOSCluster"
metadata:
  name: "example-storageos"
  namespace: "default"
spec:
  secretRefName: "storageos-api"
  secretRefNamespace: "default"
  csi:
    enable: true

CSI Credentials

To enable CSI Credentials, ensure that CSI is enabled by setting csi.enable to true. Based on the type of credentials to enable, set the csi fields to true:

apiVersion: "storageos.com/v1alpha1"
kind: "StorageOSCluster"
metadata:
  name: "example-storageos"
  namespace: "default"
spec:
  ...
  ...
  csi:
    enable: true
    enableProvisionCreds: true
    enableControllerPublishCreds: true
    enableNodePublishCreds: true
  ...

Specify the CSI credentials as part of the storageos secret object as:

apiVersion: v1
kind: Secret
metadata:
  name: "storageos-api"
...
...
data:
  # echo -n '<secret>' | base64
  ...
  ...
  csiProvisionUsername:
  csiProvisionPassword:
  csiControllerPublishUsername:
  csiControllerPublishPassword:
  csiNodePublishUsername:
  csiNodePublishPassword: