fix(runway-base): Implement "startup" and "readiness" probes.

Florian Forsterrequested to merge
fforster/startup-probe into main

fix(runway-base): Implement "startup" and "readiness" probes.

We previously allowed to configure a "starup" probe, but then actually created a "readiness" probe in Kubernetes. That means that "liveness" probes were started immediately, not after the first successful startup probe. This change fixes this by actually creating a startup probe on Kubernetes.

This change also implements "readiness" probes, which were already present in our schema but not implemented in the Helm chart.

This should help us configure AI Gateway more correctly: because the container takes quite long to start up, using an actual startup probe is the right thing to do. Emitting a readiness probe instead was shutting down the container too early due to the liveness probes failing.

Issue: team#500 (closed)

Merge request reports