2024-09-25

I recently released a new CLI tool, Cloudy. Cloudy makes it easy to spin up one-off compute instances using various cloud providers (like AWS, Scaleway, etc).

In practice, it feels similar to tools like docker run or vagrant, but for running temporary compute instances in the cloud, instead of running locally.

Cloudy gives a nice CLI interface (with good Bash completion) for interactively managing your running compute instances.

This post gives a brief overview of how to use Cloudy, and lays out my future plans for the project. You can find out more information in the project's README.

How to Use Cloudy

Cloudy supports multiple cloud providers:

• AWS (support planned)
• Scaleway

The following instructions assume you're using Scaleway.

First, you'll need to setup a configuration file for Cloudy, as detailed in the README.

Once you've done that, you can create your first instance:

$ cloudy scaleway create \
    --zone nl-ams-1 \
    --instance-type PLAY2-NANO \
    --volume-size 75 \
    --image-id "1ec31ce3-bec3-4866-81fc-08d4b6966f9f"

This command creates a compute instance in Scaleway in zone nl-ams-1, using instance type PLAY2-NANO (https://www.vpsbenchmarks.com/hosters/scaleway/plans/play2-nano), with a 75 GB disk, and using an Ubuntu 24.04 disk image.

After creating the image, you can SSH to it with the command:

$ cloudy ssh

When you're finished, you can shut it down with the command:

$ cloudy destroy

Instance Setup Scripts

One of the main features of Cloudy is the ability to specify Instance Setup scripts. These are YAML files that contain cloud-init scripts to do things like install packages and run commands on first boot.

Here's an example of a Cloudy Instance Setup script:

~/.config/cloudy/instance-setups/hello-world.yaml:

short-description: An example instance-setup script

cloud-init-user-data: |
  #cloud-config

  # Install these packages using the default package manager of the instance.
  packages:
    - cowsay

  # Run these commands after bootup.
  runcmd:
    - |
      /usr/games/cowsay "hello world from Cloudy!" > /hello-from-cloudy

You can see that this script uses a cloud-init cloud-config file to install the package cowsay, and then use it to print a message to the file /hello-from-cloudy

You can specify an Instance Setup script when creating a new instance:

$ cloudy scaleway create \
    --zone nl-ams-1 \
    --instance-type PLAY2-NANO \
    --volume-size 75 \
    --image-id "1ec31ce3-bec3-4866-81fc-08d4b6966f9f" \
    --instance-setup hello-world

Instance Setup scripts give you quite a lot of flexibility in how your compute instances are setup.

Future Plans

I have a few plans for future improvements to Cloudy:

1. Support more cloud providers.

   It would be nice to support more cloud providers, like AWS, GCP, Azure, DigitalOcean, etc.

   It would also be nice to make Cloudy flexible enough so that support for additional cloud providers could be provided by plugins, similar to how Terraform has plugins for different cloud providers.

2. Provide more built-in Instance Setup scripts.

   Cloudy currently provides a few example Instance Setup scripts, but it would be nice to provide additional scripts that do common tasks, like setup popular web servers, databases, programming environments, etc.

3. Add more functionality for operations to be performed after booting an instance.

   It would be nice to add in support to Instance Setup scripts for doing things like copying files back to the host after creating an instance, running arbitrary commands on either the host or the cloud instance, etc.

   One use for this would be to copy files like public keys (after they are generated) from the cloud instance to the local host. This would make it easy to setup things like VPNs.

4. Add the ability to specify input variables to Instance Setup scripts.

   I'm imagining this would work similarly to Terraform variables.

Conclusion

Cloudy aims to be an easy-to-use imperative CLI tool for spinning up one-off compute instances in various cloud providers.

PRs and issues are welcome helping with any of the above features, or even just fixing small bugs or making improvements.

tags: haskell