ci: Add GitHub action to upload S3 website configuration

[StefanScherer]

Add a GitHub action to upload the S3 website configuration from a JSON file in this repo.
With that PR it's possible to maintain the redirections directly from this repo by making pull requests to

• _website-config-docs-stage.json for https://docs-stage.docker.com
• _website-config-docs.json for https://docs.docker.com

An example for the docs-stage.docker.com site with a redirect of all v17.03/ to /
Important keys are

• KeyPrefixEquals is set to v17.03/
• ReplaceKeyPrefixWith is set to empty string "" which means /
• HostName and Protocol must be set for each redirect rule, otherwise AWS redirects to the internal name. That's why we need two different JSON files.

  {
   "Condition": {
    "HttpErrorCodeReturnedEquals": null,
    "KeyPrefixEquals": "v17.03/"
   },
   "Redirect": {
    "HostName": "docs-stage.docker.com",
    "HttpRedirectCode": null,
    "Protocol": "https",
    "ReplaceKeyPrefixWith": "",
    "ReplaceKeyWith": null
   }
  },

[StefanScherer]

A temporary build that also ran in this PR branch was green: https://github.com/docker/docker.github.io/pull/10612/checks?check_run_id=573354449 and it successfully uploaded the website config for docs-stage.

[StefanScherer]

Ping @thaJeztah this is the replacement for the nginx redirect rules we had in the past. Now a similar config for S3 website is back in the repo to maintain it through PRs.

[thaJeztah]

Thanks! I just recalled we also need to take redirect_from entries in the /ee and /datacenter into account before we remove the content; here's a quick 'n dirty way to get which pages have redirects from outside of /(ee|datacenter|compliance);

echo "{" > extra-redirects.json
grep -E ': "/(compliance|datacenter|ee)/' <(curl -fsSL https://docs.docker.com/redirects.json | jq .) >> extra-redirects.json
echo '  "/dummy/": "/"' >> extra-redirects.json
echo "}" >> extra-redirects.json

extra-redirects.json.txt

[thaJeztah]

For those pages, I think we'd need dummy pages in /ee, /datacenter and /compliance

[thaJeztah]

LGTM

[StefanScherer]

This PR only adds automation to the current status. I've uploaded the website configs manually with a similar Go tool, there's no aws cli command for that.

For the three EE related folders I plan to do this in a new PR after the merge. We can simply remove the three folders and add three redirection rules to the new website*.json files. We will test this PR in docs-stage to see if everything gets a new URL prefix with the target domain name.

[thaJeztah]

We can simply remove the three folders and add three redirection rules to the new

Right, so that's the tricky bit; there's redirects that are currently in place that redirect to the /ee or /datacenter directory, but due to how Jekyll constructs redirects, those redirects are configured in the target page.

For example, https://docs.docker.com/reference/ucp/3.0/cli/support/ redirects to https://docs.docker.com/datacenter/ucp/3.0/reference/cli/support/

That redirect is defined in https://github.com/docker/docker.github.io/blame/663ea14a3ca321d8f4a257d3b5ca8648ba5aec2d/datacenter/ucp/3.0/reference/cli/support.md#L6

Which means that once that page is removed, the redirect is no longer in place. So for pages that define a redirect, we should either replace them with a "dummy" page that only has the redirect_from page set (so that Jekyll generates the redirect page), or commit the generated redirect HTML's to git

[thaJeztah]

Updated, with redirect "from" /ee/ and /datacenter/ removed

extra-redirects.json.txt

[StefanScherer]

I don't know how much effort we want to put into the forwarding to Mirantis. Do we have an "SLA" of how many of the old links should work/redirect to them? 😅