No space left on device for docker-build workflow

[xmfcx]

Checklist

• I've read the contribution guidelines.
• I've searched other issues and no duplicate issues were found.
• I'm convinced that this is not my fault but a bug.

Description

• https://github.com/autowarefoundation/autoware/actions/runs/13546559639/job/37859542292?pr=5801
• https://github.com/autowarefoundation/autoware/actions/runs/13556055859/job/38212669829?pr=5819

The workflows are failing with Error: No space left on device.

Filesystem      Size  Used Avail Use% Mounted on
/dev/root        73G   21G   53G  28% /

The available space is plenty.

It fails at an early stage too.

Possible causes

• Did we add something in a near past?
• Have the docker build cache been overgrown?

cc. @youtalk @mitsudome-r @oguzkaganozt

[mitsudome-r]

The only thing that I can think of is CUDA upgrade, but I'm not sure why it started to fail at this timing.

[youtalk]

I hope #5830 resolves this issue.

[mitsudome-r]

I have investigated the size of build-cache with the following commands.

Before:
skopeo inspect --raw docker://ghcr.io/autowarefoundation/autoware-buildcache@sha256:60073907bce4f4fe9e7ce14a1d4d4cb3920a8aefe8089c4396897f1ad190c4fa | jq '[.manifests[].size] | add'
I get about 11.5 GB

Latest:
skopeo inspect --raw docker://ghcr.io/autowarefoundation/autoware-buildcache:amd64-main | jq '[.manifests[].size] | add'
I get about 4.2 GB

[xmfcx]

So this was caused by an overgrowing cache.

I'm not sure how to prevent it from overgrowing.
But a simple solution would be to remove the cache periodically, maybe weekly to prevent it from going over a size limit.
Or a daily workflow could check its size and remove it once it goes over a size limit.
I think daily checking is a solution until we have other ways to keep its size limited.
And checking its size can be done with the command @mitsudome-r shared above.