This PR adds code to the CLI to enable fastly compute serve with "enable Pushpin", a new mode that locates and runs Pushpin before starting Viceroy to run guest code.

This is a companion PR to fastly/Viceroy#497, and will only function after that feature has landed.

How to enable it:

It can be invoked in one of two ways:

1. set a new flag --experimental-enable-pushpin
2. or, set a new flag in fastly.toml:

[local_server.pushpin]
enable = true

What it does:

• fastly compute serve starts Pushpin before starting Viceroy, and detects its startup. It also sets up to shut it down when the CLI process ends.

  • pushpin must be started at this time, because routes are set based on the backends defined in the fastly.toml manifest file.
  • "kill process" is done differently between Windows and Unix, so build flags are used to include the right implementation of killProcess.

• Pushpin routes are set up based on the fastly.toml manifest file.

  • A route is set up per backend.
  • The name of each backend is set as an id= condition. For example, a backend named origin will set id=origin. Viceroy handles Fanout handoff by forwarding a request for this backend to Pushpin and set the header Pushpin-Route: origin, allowing Pushpin to match the route.
  • If the backend definition's URL has a path segment, then it is set as a replace_beg configuration on the route. For example, if a backend is configured for http://localhost:3000/realtime, then this is set as replace_beg=realtime. Pushpin will forward a request for /api/user to this backend as http://localhost:3000/realtime/api/user.
  • over_http is set for every route target, as to always enable WebSocket-over-HTTP.
  • If the backend URL is HTTPS, then ssl is added.
  • If the backend definition sets override_host, then it will be set as the host value of the route target.

• When starting Viceroy, --local-pushpin-proxy-port=<port> is appended as a command-line argument, which starts Viceroy in "enable Pushpin" mode.

• Pushpin proxy runs by default on port 7677, but this can be overridden:

  • using proxy_port=<port> under local_server.pushpin section of the fastly.toml file, or
  • using --pushpin-proxy-port=<port>

• Pushpin's publishing endpoint runs by default on port 5561, but this can be overridden:

  • using publish_port=<port> under local_server.pushpin section of the fastly.toml file, or
  • using --pushpin-publish-port=<port>

• The pushpin binary is searched on the system path, but can be specified:

  • using pushpin_path=/path/to/pushpin under the local_server.pushpin section of the fastly.toml file, or
  • using --pushpin-path=/path/to/pushpin.

• The CLI writes the config and routes to temporary files in the OS temporary directory. This is cleaned up as part of shutdown.

• Pushpin runner places some runtime files into a temporary directory specified in pushpin.conf. We set this to an OS temporary directory, so that multiple instances can run if necessary. This is cleaned up as part of shutdown.

• Pushpin runner places logs into a directory specified in pushpin.conf. This is placed in the ./pushpin-logs directory under the current project

All Submissions:

• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same update/change?

New Feature Submissions:

• Does your submission pass tests?

Changes to Core Features:

• Have you added an explanation of what your changes do and why you'd like us to include them?
• Have you written new tests for your core changes, as applicable?
• Have you successfully run tests with your changes locally?

User Impact

• What is the user impact of this change?

Are there any considerations that need to be addressed for release?

The new feature is opt-in using the --experimental-enable-pushpin flag, and does nothing if the new flag is not set.