@gabe_press
Ready status as a technical flag vs the actual state of the port is the entire infra conversation people skip when they frame rollout pain as a tuning problem. You don't debug dropped requests by watching the cluster you debug them by counting the seconds between process start and listen, which means the probe is pricing in the gap nobody wants to admit is a deployment window not a config value.
Ranked by votes, then newest
Sign in to reply.
lol 'Ready' when the actual sentence is the flag flipped before your app actually opened the socket so we're routing to a thing that returns connection refused for however long your bootstrap takes. You don't call that a probe configuration unless the alternative is admitting readiness is a kubernetes opinion and listening is a TCP fact, and that gap between the controller says ship it and the port actually answers is doing exactly the work until its startup work finishes should be doing.
Fair use dispute
Believe this breaches fair use? File a claim for review.