When running the mirror-images step from within the release-image workflow
we receive error messages such as
> msg="Failed to sync" target=ghcr.io/continuwuity/continuwuity:v0.5.0-rc.7
> source=forgejo.ellis.link/continuwuation/continuwuity:v0.5.0-rc.7
> error="failed to send blob post, ref
> ghcr.io/continuwuity/continuwuity@sha256:74976f7b85018b5abd867333bc783c7230d985a4b0af595bbf55964e25afe6ef:
> unauthorized"
So, we will need to define our credentials in the release-image workflow too
it seems, when we pull in the mirror-image workflow.
This is a test by adding all credentials that are defined in mirror-images.yml
Probably we don't need them all, but if this does not work, the whole approach
is flawed and we can remove everyting again.
If it works, we should remove unneccessary credentials until we found the
required ones.
Blind attempt at fixing the:
> 'runs-on' key not defined in Release Docker Image/merge-release
> 'runs-on' key not defined in Release Docker Image/merge-maxperf
Errors in the release-image CI
Run it once a night for regular cleanup etc and after each release-image
workflow instead of doing it blindly every 2 hours.
This will only run the mirror job if BOTH build types were
successful, but it is better than nothing, and we have the nightly run
to pick up everything else.
Thanks @jade for the quick feedback and hand-holding in creating this.
.well-known env var in example composes now point to `example.com:443`,
and are turned on by default for Caddy labels and the Traefik configs.
For the Caddyfile example, they're left uncommented. Added small
bulletpoint on how to use this env var. Consistencize all the
.well-known comments inside these composes.
I don't want to overly fix the example composes and get it wrong, so
this small change should be enough. Split-domain deployments are fully
outsourced to the Delegation page now, and some changes are also made
there to be more comprehensive.
Instead of /etc/resolv.conf on host. This works around the
systemd-resolved footgunning issue and provide an inline way to
configure resolvers, separate from that of the host system.
* Also add a note disclaiming that Docker is for demo purposes only
* Use port 8008 for Docker Quick Run instructions
* Remove the steps to create first admin user
I do believe "Docker - Quick Run" should be moved to after Docker
Compose section, but will want to discuss this before advancing
* Move all *.yml files to docs/public/deploying so they can be viewable
as raw files on the webpage
* Link to those files with "(view raw)"
* Move minimal environment variables to Docker Compose section
* Unheading the Docker Quick Run stuff to prepare deprioritizing them
* Remove redundant Registry column in Image table
* Rename reverse proxy subsections to have software name as first word
Renovate Bot