Deployment with EC3

In order to deploy an elastic Kubernetes cluster with the OSCAR platform, it is preferable to use the IM Dashboard. Alternatively, you can also use EC3, a tool that deploys elastic virtual clusters. EC3 uses the Infrastructure Manager (IM) to deploy such clusters on multiple Cloud back-ends. The installation details can be found here, though this section includes the relevant information to get you started.

Prepare EC3

Clone the EC3 repository:

git clone https://github.com/grycap/ec3

Download the OSCAR template into the ec3/templates folder:

cd ec3
wget -P templates https://raw.githubusercontent.com/grycap/oscar/master/templates/oscar.radl

Create an auth.txt authorization file with valid credentials to access your Cloud provider. As an example, to deploy on an OpenNebula-based Cloud site the contents of the file would be:

type = OpenNebula; host = opennebula-host:2633;
username = your-user;
password = you-password

Modify the corresponding RADL template in order to determine the appropriate configuration for your deployment:

Virtual Machine Image identifiers
Hardware Configuration

As an example, to deploy in OpenNebula, one would modify the ubuntu-opennebula.radl (or create a new one).

Deploy the cluster

To deploy the cluster, execute:

./ec3 launch oscar-cluster oscar ubuntu-opennebula -a auth.txt

This will take several minutes until the Kubernetes cluster and all the required services have been deployed. You will obtain the IP of the front-end of the cluster and a confirmation message that the front-end is ready. Notice that it will still take few minutes before the services in the Kubernetes cluster are up & running.

Check the cluster state

The cluster will be fully configured when all the Kubernetes pods are in the Running state.

 ./ec3 ssh oscar-cluster
 sudo kubectl get pods --all-namespaces

Notice that initially only the front-end node of the cluster is deployed. As soon as the OSCAR framework is deployed, together with its services, the CLUES elasticity manager powers on a new (working) node on which these services will be run.

You can see the status of the provisioned node(s) by issuing:

 clues status

which obtains:

| node            |state| enabled |time stable|(cpu,mem) used |(cpu,mem) total|
|-----------------|-----|---------|-----------|---------------|---------------|
| wn1.localdomain | used| enabled | 00h00'49" | 0.0,825229312 | 1,1992404992  |
| wn2.localdomain | off | enabled | 00h06'43" | 0,0           | 1,1073741824  |
| wn3.localdomain | off | enabled | 00h06'43" | 0,0           | 1,1073741824  |
| wn4.localdomain | off | enabled | 00h06'43" | 0,0           | 1,1073741824  |
| wn5.localdomain | off | enabled | 00h06'43" | 0,0           | 1,1073741824  |

The working nodes transition from off to powon and, finally, to the used status.

Default Service Endpoints

Once the OSCAR framework is running on the Kubernetes cluster, the endpoints described in the following table should be available. Most of the passwords/tokens are dynamically generated at deployment time and made available in the /var/tmp folder of the front-end node of the cluster.

Service	Endpoint	Default User	Password File
OSCAR	https://{FRONT_NODE}	oscar	oscar_password
MinIO	https://{FRONT_NODE}:30300	minio	minio_secret_key
OpenFaaS	http://{FRONT_NODE}:31112	admin	gw_password
Kubernetes API	https://{FRONT_NODE}:6443		tokenpass
Kube. Dashboard	https://{FRONT_NODE}:30443		dashboard_token

Note that {FRONT_NODE} refers to the public IP of the front-end of the Kubernetes cluster.

For example, to get the OSCAR password, you can execute:

./ec3 ssh oscar-cluster cat /var/tmp/oscar_password