build: gha: always use GHA cache for Linuxkit Cache in hopes of diminishing DockerHub pull-rate limits trouble

[rpardini]

build: gha: always use GHA cache for Linuxkit Cache in hopes of diminishing DockerHub pull-rate limits trouble

• previously, it only used GHA cache for builds running on GH-hosted runners
• this now enables it for all builds, even on self-hosted runners
• next few builds might still fail, but further ones will have a cache (we've save-always: true) and should eventually work

Signed-off-by: Ricardo Pardini ricardo@pardini.net

[rpardini]

Alternatively to this, we could remove the vars CI_RUNNER_LK_ARM64 and CI_RUNNER_LK_AMD64, and thus let LK builds run on GitHub-hosted runners -- then those will have a cache.

The original logic was to disable cache for self-hosted runners as those are (usually/by default) not ephemeral, and always run in the same dir in the host, thus the cache dir should survive across GHA jobs -- this is somehow not happening here @jacobweinstock any idea why?

[rpardini]

the cache dir should survive across GHA jobs -- this is somehow not happening here

Oh damn. I always forget about this -- actions/checkout defaults to clean: true (and fetch-depth: 1!) and thus obliterates the local cache away. Adding with: { clean: false, fetch-depth: 0 } to all the action/checkout calls should also make the cache effective (and git pulls much faster over time) also on self-hosted runners.

Let me know which you prefer -- and if there's something on the runners to make them ephemeral somehow.

[rpardini]

Alternative to this: https://github.com/tinkerbell/hook/pull/231

[jacobweinstock]

Hey @rpardini , lets try this one! Thanks!