Closed jklaw90 closed 2 years ago
I think I'd rather return 429 instead of a cached result. There's no reason anyone should be hitting this endpoint more than 1/s.
I was thinking in the case of liveness and readiness probes in k8s, there may be multiple probes calling this endpoint? So no one probe would hit it more than 1/s, but the multiple probes would causing possibly unknown behavior. If it's not needed i can close it out.
Fixes
If we have a failure threshold greater than 1, this would never be marked unhealthy. It would return one unhealthy, and then healthy until the next
dbPingLimiter.Allow
and so on.Proposed Changes
dbPingLimiter.Allow
Release Note