Periodic check of links about documentation

[enolfc]

yes and this is partially done (missing opening an issue if scheduled check fails)
See https://github.com/EGI-Federation/documentation/blob/main/.github/workflows/check-links.yml

[brucellino]

Hi! I might be able to look at this as a quick exercise to get myself familiar. Are you thinking about a cron trigger here to run one of the actions included in the existing workflow, or an outside agent which monitors the pages for missing links and opens issues against them?

[enolfc]

There is already some regular checks of the links, but the result is an email that gets ignored in my inbox. See last execution https://github.com/EGI-Federation/documentation/actions/runs/7948069227. It would be good to open an issue indeed instead of just getting the notification about broke execution

[brucellino]

Hm. 14 minutes for a run. Sound like something you want to do asynch. I don't suppose we deploy this to a staging environment perhaps?

[enolfc]

15 minutes is checking every link on the repo. For PRs, only links within the changed files are checked, so it goes rather fast. Fixing the broken links will also reduce greatly the run (you just avoid hitting the timeouts for the broken links).

[brucellino]

As I understood, there are links outside of this repo, which we do not control -- just plain dead links -- which we need to monitor for right?

So we can have the fast checks run only on the page that has changed -- we can parametrise this and it will run in seconds. But we don't have a trigger for changes which happen outside of the repo, so we need to schedule regular crawls for that (once a week, e.g.).

graph LR
  subgraph PR
    direction TB
    page[page changes] --> check_page[Check links in page] 
  end

  subgraph Cron
    direction TB
    cron[Scheduled check] --> check_all[Check all links in all pages]
  end

Loading

Failures on the cron path can open issues. Failures on the PR path break the build.

Sound right?

[brucellino]

nice. So, todo is:

1. Open issue if cron fails
2. only check links in changed file which has changed in PR (Looks like that goes in .github/workflows/build_pr_preview.yml ?).. that sounds like a job for a pre-commit hook -- see #635