"Ready in a minute" remains if portal API doesn't start

[DonMartin76]

In cases where there's a misconfiguration in the static configuration which leads to the portal API not starting, the UI simply shows a "Ready in a minute, waiting for portal API" (or similar), but doesn't give a hint whether there's a problem (or which problem there is).

Proposal:

• At startup of the portal API, record error messages until the server has successfully started
• Catch errors at startup and instead of just quitting portal-api, go into an "error mode" which behaves as follows:
  • Report on the /health endpoint as 200 OK "healthy", but with a payload which makes it clear that it's not (so that services are routed e.g. in Kubernetes), but don't serve anything else than this end point
  • Return error information as JSON on the health end point and surface these to the UI
  • Kill the process after 60 seconds so that it would be restarted by the orchestrator

Advantages:

• The UI (wicked.portal) will display whether the portal API is not up and running, and also possibly why it is not running correctly

This has been a source of confusion for many; it's actually not the system itself which isn't working, it's usually a faulty configuration, a missing environment variable or something similar.

[DonMartin76]

There has been a change to this, even though the above behavior is still not implemented. Now the portal UI also restarts if it cannot reach the API in a certain amount of time (30 seconds).