Link Search Menu Expand Document

CaaS Infrastructures

Container as a Service (CaaS) running in the cloud can be used by Nuvla to deploy applications.

If you already have a Kubernetes or Docker Swarm cluster, you can simply register it to host user container applications.

If you do not have a CaaS yet, simply follow these instructions.

Add CaaS or Other Infrastructure Services

Adding to (or registering with) Nuvla existing CaaS (or other infrastructure services) is simple. First, on the Cloud page, click the add button.

Add CaaS

Then, select the type of service you want to add. Currently, we support (Docker) Swarm and Kubernetes CaaS.

Add CaaS

From there, fill the form, including the endpoint to the service reachable from Nuvla.

Add CaaS

You can also share this service with users, using the share button (highlighted in red in the above screenshot).

The same procedure can be used to add other infrastructure services, such as an S3 (Simple Storage Service) compatible service (e.g. MinIO).

Container as a Service Infrastructures

Container as a Service (CaaS) can be used to deploy apps. We support both Docker Swarm and Kubernetes.

Nuvla also supports an advanced data management feature. For this to work, the CaaS used when deploying containers must include an S3 service and a distributed file system. To learn how to install and configure these services, please get in touch.

Docker Swarm Cluster

Docker Swarm clusters provide the computational platforms that Nuvla uses to deploy container-based applications.

A number of methods can be used to deploy a Docker Swarm cluster. See the Docker Swarm documentation for an overview of Docker Swarm and how to deploy it.

You may want to consider Docker Machine for installation; it automates the deployment of a Docker Swarm cluster on a cloud infrastructure.

The nuvla/deployment repository contains a script (deploy-swarm-exoscale.sh) to deploy a Swarm cluster with Docker Machine. Clone this repository:

git clone https://github.com/nuvla/deployment.git

to a machine that will be convenient for managing your cluster.

This script uses Docker Machine to deploy a Swarm cluster on the Exoscale cloud. Customise this script to change the cloud driver or the sizes of machines deployed.

In your cloned repository, descend into the swarm subdirectory and copy env-example.sh to env.sh.

cd deployment/swarm
cp env-example.sh env.sh

Edit this file, changing the values of the variables to customize your installation. The SSH_KEY and EXOSCALE_* variables are the most important for the Swarm deployment.

Docker Machine uses SSH to communicate with the virtual machines of the cluster. By default the key ${HOME}/.ssh/id_rsa will be used (or created if it does not exist). If you want to use a different key, then set the environmental variable SSH_KEY in the env.sh file.

WARNING: Use an SSH key without a password. If you use one with a password, you will be prompted for it, repeatedly. To generate a new SSH key without a password just set SSH_KEY to a file that does not exist.

After your changes, run:

source env.sh

to set all of the environmental variables for the Swarm management script.

The command to use to create the cluster is:

./swarm-deploy-exoscale.sh deploy 3

This creates a cluster with one master and two workers (three nodes in total). If you do not provide the second argument, it defaults to one.

The size and number of worker nodes to deploy depends entirely on the foreseen workload. The script uses “Small” instances that have 2 vCPUs, 2 GB RAM, and 50 GB disk.

You will want to note the IP addresses of the Docker Swarm master and workers (if any). You can recover these IP addresses by running the command docker-machine ls if necessary.

You can access your machines with:

docker-machine ssh dockermachine-1556097484

using the name of the machine provided in the machine listing. This may be a non-root account; to become root use the command sudo su -.

WARNING: This procedure creates a docker-machine security group in Exoscale associated to the Docker virtual machines. However, this security group does not open the default ports used to map to deployed containers. These start from 30000. Therefore, in order to reach from the Internet services running in containers deployed in the Swarm cluster at Exoscale, you have to update the security group to allow incoming connectivity on these ports (say 30000-32000).

To tear down the Swarm cluster use the command:

./swarm-deploy-exoscale.sh terminate

This will terminate all the machines in your Swarm cluster. Use with care.