Skip to content

06-operators.md

Latest commit

 

History

History
181 lines (135 loc) · 3.77 KB

06-operators.md

File metadata and controls

181 lines (135 loc) · 3.77 KB

Operators

Centrally we have the Operator Lifecycle Manager installed. It provides us various operators ready to be consumed within our project.

We will deploy a etcd-cluster and see what we can do.

Subscription

To be able to consume an operator we will create a subscription within a project. This will enable the CRDs, as well as install the operator that watches our CRDs as well as the deployed clusters.

Create project:

oc new-project userXY-operator

Subscribe to the etcd operator, by creating a file called etcd-subscription.yaml with the following content:

# etcd-subscription.yaml
apiVersion: operators.coreos.com/v1alpha1
kind: Subscription
metadata:
  generateName: etcd-
  namespace: userXY-operator
  name: etcd
spec:
  source: rh-operators
  name: etcd
  startingCSV: etcdoperator.v0.9.2
  channel: alpha

Now subscribe using the file:

oc create -f etcd-subscription.yaml
oc get pods

You can also subscribe to the operator through the Console, though not all integration might be ready, due to being technical preview by default.

You can see the running operator:

oc get pods

Once this is done, you are able to deploy a cluster, by using a crd.

You can do this by creating a file called etcd-cluster.yaml with the following content:

# etcd-cluster.yaml
apiVersion: "etcd.database.coreos.com/v1beta2"
kind: "EtcdCluster"
metadata:
  name: "example-etcd-cluster"
spec:
  size: 3
  version: "3.1.10"
oc create -f etcd-cluster.yaml

Your cluster will be bootstrapped and will become ready:

oc get pods -w
oc get service

Describing the configured CRD gives us more information about the deployment:

oc describe EtcdCluster example-etcd-cluster

We can also store something within our cluster:

oc rsh example-etcd-cluster-XYZ
export ETCDCTL_API=3
etcdctl get foo
etcdctl put foo bar
etcdctl get foo

Recovery

The operator is watching the deployed cluster and will recover it from failures. This is not a feature driven by a StatefulSet or a DeploymentConfig, rather the operator watches the deployed cluster and ensures it is kept in the desired overall state:

oc delete pod example-etcd-cluster-XYZ
oc get pods
oc rsh example-etcd-cluster-ABC
export ETCDCTL_API=3
etcdctl get foo

See how the changes got logged.

oc describe EtcdCluster example-etcd-cluster

Updating

But well it looks like we didn't deploy a recent enough version:

oc describe pod -l app=etcd | grep -E 'Image:.*etcd' | uniq
    Image:         quay.io/coreos/etcd:v3.1.10

Let's update the cluster. This is done by patching the CRD to a new version

oc get EtcdCluster example-etcd-cluster -o yaml
# check the version

Now we can either edit the current deployment or supply an update to the spec:

# etcd-update.yaml
apiVersion: "etcd.database.coreos.com/v1beta2"
kind: "EtcdCluster"
metadata:
  name: "example-etcd-cluster"
spec:
  size: 3
  version: "3.2.13"
oc apply -f etcd-update.yaml

Now watch how each member is being updated to 3.2.13 until all of them are updated:

oc describe EtcdCluster example-etcd-cluster
# watch the events of the resource or the overall events
oc get events
# in the end all images should be updated
oc describe pod -l app=etcd | grep -E 'Image:.*etcd' | uniq
    Image:         quay.io/coreos/etcd:v3.2.13

Scale Up

Works exactly the same way:

# etcd-scaleup.yaml
apiVersion: "etcd.database.coreos.com/v1beta2"
kind: "EtcdCluster"
metadata:
  name: "example-etcd-cluster"
spec:
  size: 5
  version: "3.2.13"
oc apply -f etcd-scaleup.yaml
oc get events -w
# until all are up
oc get pods -l app=etcd

Scale Down

Is also supported with this operator. Do you figure out how?