Ops wants us to set up an endpoint that returns a 200 OK or 500 error based on some basic internal health checks, like if the database is up or not. If the endpoint doesn't return a 200, monitoring that ops will set up will notify us and them. We talked a bit with chartjes about possible checks we want, and came up with two: - Can the app connect to the database at all? - Are our public endpoints responding with 200s? I propose that the following are the public endpoints we really care about: - Self-repair endpoint: /<locale>/repair/ - Action Implementations: /api/v1/action/<action_name>/implementation/ - Bundles: /api/v1/fetch_bundle/ Open Questions: - Will API availability tests running from a webhead be a valid test as to whether the API is up or not? Can we be sure they're leaving the network and then coming in from the outside? Maybe this isn't a problem. - Is this endpoint okay to be publicly available? How will ops be hitting it?

This was fixed some time ago.