Merge pull request #9298 from ingvagabund/describe-how-to-deploy-cluster-autoscaler

How to deploy the cluster autoscaler

playbooks/openshift-cluster-autoscaler/README.md

@@ -0,0 +1,101 @@
+# Cluster autoscaler
+
+This directory contains Ansible playbooks for deployment of [cluster autoscaler](https://github.com/openshift/kubernetes-autoscaler).
+
+It's assumed a user is familiar with the cluster deployment.
+For cluster deployment in AWS cloud provider you can follow instructions provided in [AWS playbooks](https://github.com/openshift/openshift-ansible/tree/master/playbooks/aws).
+
+## Running
+
+The autoscaler can be deployed directly by running the corresponding playbook:
+
+```sh
+ansible-playbook -i inventory playbooks/openshift-cluster-autoscaler/config.yml
+```
+
+or during your cluster deployment by setting `openshift_cluster_autoscaler_install=True`
+in the corresponding inventory file.
+
+Currently, the autoscaler is deployed only on nodes that have the `infra` [node role](https://github.com/openshift/openshift-ansible#node-group-definition-and-mapping) assigned (forced by node selector set to `node-role.kubernetes.io/infra: "true"`, see `openshift_cluster_autoscaler_node_selector` variable of the `openshift_cluster_autoscaler` role).
+In order to run the autoscaler inside the AWS cloud provider,
+one has to attach an [IAM role](https://aws.amazon.com/iam/details/manage-roles/) to the `infra` node that has appropriate permissions to scale down nodes in  [Auto Scaling Group](https://docs.aws.amazon.com/autoscaling/ec2/userguide/AutoScalingGroup.html) belonging to the OpenShift compute node group.
+
+In case there is only one `infra` node, the process of attaching the IAM role
+reduces to running:
+* assuming the AWS credentials are available
+* assuming the IAM role with appropriate permissions is named `aws-autoscaler-role`
+* region set to `us-east-1`
+* with the infra node name set to `CLUSTERID infra group 1`
+
+```sh
+# to install the aws executable
+sudo pip install awscli
+# Get infra node instance ID (assuming there is only one)
+INFRA_NODE_INSTANCE_ID=$(aws --region us-east-1 ec2 describe-instances --filters 'Name=tag:openshift-node-group-config,Values=node-config-infra' "Name=tag:Name,Values=CLUSTERID infra group 1" | jq '.Reservations[0].Instances[0].InstanceId' --raw-output)
+# Get AssociationId for the AWS instance
+ASSOCIATION_ID=$(aws --region us-east-1 ec2 describe-iam-instance-profile-associations --filters "Name=instance-id,Values=${INFRA_NODE_INSTANCE_ID}" | jq '.IamInstanceProfileAssociations[0].AssociationId' --raw-output)
+# Attach the IAM role to the infra instance
+aws --region us-east-1 ec2 replace-iam-instance-profile-association --association-id ${ASSOCIATION_ID} --iam-instance-profile Name=aws-autoscaler-role
+```
+
+## Testing
+
+To exercise the basic autoscaling capability, one can create a workload in the same namespace as the autoscaler lives
+and observe creation of new nodes. Followed by deleting the workload and observing nodes being removed from the cluster.
+
+**Workload example**:
+```yaml
+apiVersion: extensions/v1beta1
+kind: Deployment
+metadata:
+  name: scale-up
+  labels:
+    app: scale-up
+spec:
+  replicas: 50
+  selector:
+    matchLabels:
+      app: scale-up
+  template:
+    metadata:
+      labels:
+        app: scale-up
+    spec:
+      containers:
+      - name: busybox
+        image: docker.io/library/busybox
+        resources:
+          requests:
+            memory: 2Gi
+        command:
+        - /bin/sh
+        - "-c"
+        - "echo 'this should be in the logs' && sleep 86400"
+      terminationGracePeriodSeconds: 0
+```
+
+Assuming the cluster autoscaler is deployed in `openshift-autoscaler` namespace,
+the following command will create the workload:
+
+```sh
+oc create -n openshift-autoscaler -f scale-up.yaml
+```
+
+To observe the cluster autoscaler's behavior closely, one can list its logs by running:
+
+```sh
+oc logs cluster-autoscaler-HASH
+```
+
+The pod name can be listed by running:
+
+```sh
+oc get pods --namespace openshift-autoscaler
+```
+
+To observe registration of nodes and their current state, one can run the following command
+on the master node:
+
+```sh
+watch -n 1 oc get nodes
+```