Scheduled Disk Cleanup

[AnandChowdhary]

Is your feature request related to a problem? Please describe.
I have a Netdata monitor set up which sends me a notification when it thinks "Out of space time = 7h", which means I've probably done a bunch of deploys in a short period of time, and I should use the "Disk Cleanup" feature to remove older images.

Describe the solution you'd like
It would be great if I can schedule Disk Cleanup, say once a day, week, or month. I can set up how many recent images I want to keep, and it — like a cron job — deletes unused images automatically.

Describe alternatives you've considered
I can use an API or set up a cron job myself which hits the endpoint to clean up the disk automatically, but it's not very secure since Caprover uses a password-based authentication. It would be super useful to schedule it from within the Caprover UI.

[githubsaturn]

It's a good suggestion, meanwhile, you can use a 3rd party solution like this one:
blog: https://marcopeg.com/2019/docker-vacuum
github project: https://github.com/marcopeg/docker-vacuum

[grantcodes]

In case someone ends up wanting to set up docker vacuum with caprover like I did:

1. Create an app with persistent data
2. Don't expose as a webapp
3. Map /var/run/docker.sock to /var/run/docker.sock as a persistent directory
4. Deploy captain-definition file:

{
  "schemaVersion": 2,
  "imageName": "marcopeg/docker-vacuum"
}

[redrockzee]

In case someone ends up wanting to set up docker vacuum with caprover like I did:

1. Create an app with persistent data
2. Don't expose as a webapp
3. Map /var/run/docker.sock to /var/run/docker.sock as a persistent directory
4. Deploy captain-definition file:

{
  "schemaVersion": 2,
  "imageName": "marcopeg/docker-vacuum"
}

I had this working. Unfortunately, maybe with a docker update, the image started throwing errors.

Error: "Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?"

[negati-ve]

if an application is set to have 0 instances, would it be at risk to be cleaned up by Vaccum(with default settings)?

[cakeslice]

Yes please!

I think this is one of the last missing pieces to make CapRover a fully setup and forget system

[seiyria]

Not sure if it helps, but what I do is set this in my crontab:

0 3 * * * /usr/bin/docker system prune -af

[Turab]

@seiyria, that might be quite dangerous. If for a reason your apps had stopped, that command will remove all your app containers too. Maybe you can redeploy but there is also a chance for Caprover to break.

[maietta]

Not sure if it helps, but what I do is set this in my crontab:

0 3 * * * /usr/bin/docker system prune -af

This is exactly what I do. But I also use continuous deployment model so just in case something goes wrong, I'm only one click away from a quick re-build from last git commit. So far the docker system prune option has not let me down.

[maietta]

Not sure if it helps, but what I do is set this in my crontab:

0 3 * * * /usr/bin/docker system prune -af

This is exactly what I do. But I also use continuous deployment model so just in case something goes wrong, I'm only one click away from a quick re-build from last git commit. So far the docker system prune option has not let me down.

Update: This bit me. I had turned off a service to back it up then it was gone when I tried to scale it from 0 back to 1. So, I recommend nobody does this.

[Turab]

@seiyria, that might be quite dangerous. If for a reason your apps had stopped, that command will remove all your app containers too. Maybe you can redeploy but there is also a chance for Caprover to break.

As I said before, running that command is actually dangerous. As a temporary solution, I use Portainer. I log into it once in a while and remove unused images. That's a safe solution until Caprover develops a better way.

[s-kris]

In case someone ends up wanting to set up docker vacuum with caprover like I did:

1. Create an app with persistent data
2. Don't expose as a webapp
3. Map /var/run/docker.sock to /var/run/docker.sock as a persistent directory
4. Deploy captain-definition file:

{
  "schemaVersion": 2,
  "imageName": "marcopeg/docker-vacuum"
}

@grantcodes
getting this error: exec /bin/sh: exec format error

[githubsaturn]

Usually this error happens when the cpu architecture doesn't match. Are you running on an ARM server by any chance?

[s-kris]

Yes! I'm on Hetzner ARM64.

[githubsaturn]

You can try forking the project and use buildx, similar to how CapRover build works. We compile for all CPU architectures.

Here are the relevant lines:

caprover/dev-scripts/build_and_push_release.sh

Lines 65 to 71 in f9d880e

	docker run --rm --privileged multiarch/qemu-user-static --reset -p yes
	export DOCKER_CLI_EXPERIMENTAL=enabled
	docker buildx ls
	docker buildx create --name mybuilder
	docker buildx use mybuilder
	docker buildx build --platform linux/amd64,linux/arm64,linux/arm -t $IMAGE_NAME:$CAPROVER_VERSION -t $IMAGE_NAME:latest -f dockerfile-captain.release --push .

[s-kris]

Thank you!

However, I just forked the repo and added a captain-definition file to deploy using dockerfile. It seemed faster. :)

[githubsaturn]