A container that never turns "healthy" now explains why

v0.22.0-beta

A container that never turns "healthy" now explains why. Until now the failure amounted to "the new replica did not become healthy within 90s", with no indication of where to look. The agent now queries the container before giving up and distinguishes three causes: the application does answer on its port but exposes no /health route (the most frequent case, and the hardest to guess); the image does not contain wget, which the probe needs; or nothing listens on the port, which points back to the application's startup and its listening address.
The message states the port the image actually declares and the action to take. worker containers are not concerned: no probe is set for them, so they cannot trigger this diagnosis.