Skip to content

guardian/prout

Repository files navigation

Prout

"Has your pull request been deployed yet?" - Guardian blogpost

Tells you when your pull-requests are live. Tells you when they're not, and should be.

prout overduethenseen

Prout comes from the tenet that:

Developers are responsible for checking their changes on Production

This becomes more important, and easier once you move to a Continuous Deployment release process. Important, because now a developer can break the site simply by hitting 'Merge' on a pull request - but also easier because with such a small delay (say, less than 10 minutes) between merging the work and having it ready to view in a Production setting, the developer is in a much better place to review their work; it's still fresh in their mind.

While everyone on your team may agree with this philosophy, that 10 minute lag between merge and deploy can be enough time for a developer like me to get distracted ("Look, shiny thing!" or, more realistically, "What's the next bit of work?") and forget about promptly reviewing their changes on Production.

Prout simply notifies developers in their pull request that the code has been seen in Production (a slightly stronger statement than simply saying it's been deployed).

Configuration

Follow the 4-step program:

  1. Give prout-bot write-access to your repo (so it can set labels on your pull request)
  2. Add one or more .prout.json config files to your project
  3. Add callbacks to prout - ie a GitHub webhook and ideally also a post-deploy hook
  4. Expose the commit id of your build on your deployed site

Give prout-bot write-access

Giving prout write-access by going to Settings -> Collaborators and team -> Add teams and typing in 'prout'

Admin access to the repository is required to give prout-bot write-access. Please contact your team's Engineering Manager for this.

Add config file

Add a .prout.json file to any folder you want monitored in your repo:

{
  "checkpoints": {
    "DEV": { "url": "http://dev.mysite.com/", "overdue": "10M" },
    "PROD": { "url": "http://mysite.com/", "overdue": "1H" }
  }
}

When a pull-request changes a file anywhere under that folder, Prout will scan the checkpoints defined in your config file, and update the pull-request with labels and a comment as appropriate. The url you specify in the checkpoint will be fetched, and the contents of the response will be read- so long as you embed the commit id that response, Prout will be able to work out whether or not the PR has been deployed.

Add callbacks

Add Prout-hitting callbacks to GitHub and (optionally) post-deploy hooks to your deployment systems so that Prout can immediately check your site.

GitHub

Add a GitHub webhook with these settings:

  • Payload URL : https://prout-bot.herokuapp.com/api/hooks/github
  • Content type : application/json

The hook should be set to activate on Pull Request events.

Adding a GitHub web hook by navigating to Settings -> Webhooks -> Add webhook

Note that this can be done once at the Organization Webhook level, which removes the need for doing it on each individual repo. Prout will check that any repository has a .prout.json config file present before attempting to take any action on it.

Post-deploy hooks

Whatever deployment tool you use (RiffRaff, Heroku, etc) just set it to hit Prout as a post-deploy hook (for your repo on github.com/[owner]/[repo]):

https://prout-bot.herokuapp.com/api/update/[owner]/[repo]

Hitting that url (GET or POST) will always prompt Prout to scan the repository for outstanding pull-requests.

Expose the commit id

You must embed the commit id in your site - we do this on membership.theguardian.com for instance.

I use the sbt-buildinfo plugin to store the Git commit id in my stored artifact, and then expose that value on the production site. The ugly-looking SBT config is:

buildInfoKeys := Seq[BuildInfoKey](
      name,
        "gitCommitId" -> (Option(System.getenv("BUILD_VCS_NUMBER")) getOrElse (try {
        "git rev-parse HEAD".!!.trim
      } catch {
          case e: Exception => "unknown"
      }))
    )

Slack

Users can configure a Slack hook for Prout by creating a new Slack 'Incoming Webhook':

https://your-domain.slack.com/services/new/incoming-webhook

...this will get you a 'Webhook URL', which looks something like this:

https://hooks.slack.com/services/T05FTQF9H/B012N1Y2Y/p9VyRC1ZlTqNGuu

...stick that url into a GitHub webhook for your repo as the 'Payload URL':

https://github.com/my-org/my-repo/settings/hooks/new

...and then (optionally) disable the hook in GitHub! You don't actually want to send GitHub events to the hook - this is just a place to store the private url where Prout can find it. Note that Prout needs repo-admin access in order to read the hook data!

Status

Prout has a status page per repository available at https://prout-bot.herokuapp.com/view/{organisation}/{repository}. This lists useful information like if it was able to find .prout.json, which commit it thinks is currently deployed in production, etc.

Prout also logs lots of useful information which is available in the heroku backend.

Run your own instance of Prout

Deploy

...you want to run your own instance of Prout - the instance at https://prout-bot.herokuapp.com/ is really only for the Guardian.