This doc is for Peril who want to self-host. I'm using Heroku as the example, but it will work with most process-level server abstraction, e.g. now.sh (which Peril Staging works with) or anything which can run a server from a Docker container.
Sidenote: here's a note with terminology as it's a little tricky.
So, you will need to have:
- A heroku instance for Peril
- A unique GitHub App
- A repo where you can keep Peril settings
To create this, go to the URL (with your org):
https://github.com/organizations/[my_org]/settings/apps
or
https://github.com/settings/apps (for adding it to a user)
Then create a new app (use peril-[my-org]
, there is a global namespace for GitHub apps.)
You will need to give it a homepage (anything) and a webhook url. For the webhook url use something like:
https://[my_app_name].herokuapp.com/webhook
e.g. https://peril-artsy.herokuapp.com/webhook
You should also create a secret, this is just some text that Peril will use to confirm that GitHub is the entity sending you webhooks. It can be any string, so long as it's the same in the heroku setup also.
With respect to permissions, my recommendations are:
Checks: Read & Write
- Peril comments on PRs via the Checks APIRepo Metadata: Read
- no option on this oneRepository contents: Read
- Let Peril read your codeIssues: Read & Write
- let Peril see/amend new issuesPull Requests: Read & Write
- let Peril act on prsCommit Statuses: Read & Write
- let Peril provide green/red status indicators
And if you want to run Peril only for org members: Organization Members: Read
. Note: it's trivial to go and change
these later as you think of more rules that need different events.
With respect to webhooks, my recommendations are:
Issues
Pull Request
At the bottom is the option to only allow this integration to run on your org, you want this ticked probably.
Once this is created, generate and download the App's GitHub signing key and keep track of the numerical ID (in the About section), you'll need this later.
Now that you have created the GitHub app, you need to install it onto your GitHub account/org. On the app's page in github there is a sidebar which should have a link "Install App" - click that.
It'll show a list of potential places to install, which may just be your org. Click install for your org. Here you can choose either:
- install on all repos
- install on a few repos
Either is fine, for very large orgs, I'd start with a few repos, get it stable then move to all.
Completing this gives you the installation ID in the URL, grab this, you'll need it later:
https://github.com/organizations/[my_org]/settings/installations/74679
Ok, so, you need a heroku account. So # if you've not. This post will wait for you.
Click: . It's the "auto-heroku-ize" button that will walk you through setting up the environment variables for running Danger for just one org.
It should take you to a page like this:
You need to go through adding the environment variables. The app.json
is self-documenting about the variables that
need to be set.
One of them, DATABASE_JSON_FILE
will require you to create your config repo.
You need to have a repo which Peril has access to. This repo needs to have a settings JSON file. For now, here's an example JSON file which runs some Dangerfiles:
{
"settings": {
[your settings]
},
"rules": {
"pull_request": "orta/peril-bootstrap@dangerfiles/pr.js",
"issues": "orta/peril-bootstrap@dangerfiles/issue.js"
},
"repos" : {
"orta/ORStackView": {
"issues.opened": "lock_old_issues.ts"
}
}
}
You can look at the both the pull_request and issue, to verify there are no shenanigans.
This JSON file is split into 3 parts:
- Settings for Peril
- Rules for the organization
- Rules per repo
This setup will:
- Listen for the event
"pull_request"
, and will pull"dangerfiles/pr.js"
from the repo:orta/peril-bootstrap
. - Listen for the event
"issues"
, and will pull"dangerfiles/issue.js"
from the repo:orta/peril-bootstrap
. - Listen for the event
"issues"
event, and only if the action is"opened"
and will pull"lock_old_issues.ts"
from the same repo:orta/ORStackView
. So it would ignore issue updates or deletes.
You can actually use orta/example-peril
BTW, I have some dummy Dangerfiles on that repo exactly for this purpose. Save
the above JSON as peril-settings.json
. Add that to a repo, push it to master on your GitHub remote. Here's one
I did earlier.
Note: when you make updates to this file, you need to restart your Peril server, until #88 is resolved.
Read: Settings Repo Info.
Inside your repo create a new file for your first Dangerfile. Maybe have something like:
import { warn, danger } from "danger"
// Add a check to see if a changelog exists, then warn if it isn't in modified etc
- https://github.com/danger/peril-settings
- https://github.com/artsy/peril-settings
- https://github.com/CocoaPods/peril-settings
- https://github.com/Moya/moya-peril
- https://github.com/RxSwiftCommunity/peril
- https://github.com/ashfurrow/peril-settings (personal repo)
- https://github.com/orta/peril-settings (personal repo)
(Note: Peril can technically be installed on personal accounts, too, because an account is just a specific type of org from the GitHub API's perspective.)
Run heroku logs --app [my_heroku_peril_app]
.
Go back to your integration settings, you probably have two events so far. The 2nd one, which has a
X-GitHub-Event: integration_installation
will provide you with the env var for PERIL_ORG_INSTALLATION_ID
. You can
find the installation id inside "installation.id" in the JSON. To set the heroku env var, run
heroku config:add PERIL_ORG_INSTALLATION_ID="[my-id]" --app [my_heroku_peril_app]
Setting it will restart the server.
The GitHub API doesn't let the bot know what it's user ID is yet, so you'll need to pull this out of the API, sorry. Find an issue/PR where Peril has commented, then you need to see the API request for that issue/PR. For both issues and PRs, you use a URL like
https://api.github.com/repos/[org]/[repo]/issues/[pr_or_issue_id]/comments
Check Github API docs or here if you need more info on how to perform the request or here if you are having trouble authenticating.
Then you can scroll down on the API JSON response to find the "user"."id"
of the Peril user account. It will be a
comment where the "user"."type"
value is "Bot"
.
Set this on your server using: heroku config:add PERIL_BOT_USER_ID="[bot-id]" --app [my_heroku_peril_app]
.
Inside the Integration settings under Advanced, you can get a list of all events sent to Peril. You can open one of them, and then click on "Redeliver" to send it again.