This documentation introduces how to run Docker container on FutureSystems. Currently we have deployed Docker swarm on Echo.
You will need an account on FutureSystems. To verify, try to see if you can log into india.futuresystems.org. You need to be a member of a valid FutureSystems project, and had submitted an ssh public key via the FutureSystems portal.
If your access to the india host has been verified, try to login to the docker swarm head node with the same username and key:
ssh FS_USER@149.165.150.76
NOTE: If you have access to india but not the docker swarm system, your project may not have been authorized to access the docker swarm cluster. Send a ticket to FutureSystems ticket system to request this.
Once logged in to the docker swarm head node, try to run:
docker run hello-world
to verify ‘docker run’ works.
While ‘docker run’ can start a container and you may even attach to its console, the recommended way to use a docker swarm cluster is to create a service and have it run on the swarm cluster. The service will be scheduled to one or many number of the nodes of the swarm cluster, based on the configuration. It’s also easy to scale up the service when more swarm nodes are available. Docker swarm really makes it easier for service/application developers to focus on the functionality development but not worrying about how and where to bind the service to some resources/server. The deployment, access, and scaling up/down when necessary, are all managed transparently. Thus achieving the new paradigm of ‘serverless computing’.
As an example, the following command creates a service and deploy it to the swarm cluster:
docker service create –name notebook_test -p 9001:8888 jupyter/datascience-notebook start-notebook.sh –NotebookApp.password=NOTEBOOK_PASS
It pulls a published image from docker cloud, starts a container and runs a script to start the service inside the container with necessary parameters. The option “-p 9001:8888” maps the service port inside the container (8888) to an external port of the cluster node (9001) so the service could be accessed from the Internet. In this example, you can then visit the URL:
to access the Jupyter notebook. Using the specified password when you create the service to login.
Please note the service will be dynamically deployed to a container instance, which would be allocated to a swarm node based on the allocation policy. Docker makes this process transparent to the user and even created mesh routing so you can access the service using the IP address of the management head node of the swarm cluster, no matter which actual physical node the service was deployed to.
This also implies that the external port number used has to be free at the time when the service was created.
Some useful related commands:
docker service ls
lists the currently running services.
docker service ps notebook_test
lists the detailed info of the container where the service is running.
docker node ps NODE
lists all the running containers of a node.
docker node ls
lists all the nodes in the swarm cluster.
To stop the service and the container:
docker service rm noteboot_test
You can create your own service and run it. To do so, start from a base image, e.g., a ubuntu image from the docker cloud. Then you could:
service, and create a new image from the changed instance using command ‘docker commit’.
service, and then build an image from it.
In reality, the first approach is probably useful when you are in the phase of develop and debug your application/service. Once you have the step by step instructions developped the latter approach is the recommended way.
Publish the image to the docker cloud by following this documentation:
Please make sure no sensitive information is included in the image to be published.
Alternatively you could publish the image internally to the swarm cluster.
TO DO
Once the image is published and available to the swarm cluster, you could start a new service from the image similar to the Jupyter Notebook example.