Heroku Exit

[awseward]

Context

---

Steps

• Move any state stores (DB/redis/what have you) to $new_host
• Move app to $new_host (Set up to deploy with fly.io #175)
• Set up a reverse proxy to forward clients from Heroku to $new_host
• Update client configs to point to $new_host (and/or handle HTTP 307 correctly)
  • ESP32 clients
  • Nim clients (Use nim >= 1.6.0 #172, call_status_checker 0.6.5 homebrew-tap#73)
  • hurl tests
• Goodbye, Heroku 😭

[awseward]

Re: Set up a reverse proxy […]

• It looks like the redirect isn't going to work out of the box for the ESP clients… If I'd need to reflash them to enable following redirects, that's no different from just flashing them with the new URL. 🙄 However, still probably worthwhile to turn on redirects for the next time something like this happens (we'll see if they follow 307/308 correctly or if they don't, like so many other HTTP client implementations I've come across… my guess is no 😔).

• hurl isn't handling HTTP 308 correctly. It looks like that's a libcurl thing, not a hurl thing; can maybe work around this with a custom libcurl. At a cursory glance, looks like the fix is going to be here.

• I haven't looked yet to see if if the HTTP client the launchd services are using will follow redirects. Would be sorta funny if none of the callers using the endpoint(s) will even be helped by setting up redirects in the first place. 🙃 It turns out this HTTP client also does not do the right thing with HTTP 308. 🙄 I guess that means that there's no real benefit to the proxy step outlined above.

  

  I'll probably just end up wrapping nim's std/httpclient.

  Fixed in https://github.com/awseward/call_status/releases/tag/0.6.5