Reference · Glossary

Readiness probe

Checks **“Can this instance safely take traffic?”** — fails when dependencies (vector DB, model gateway) are unavailable.

#When to use

Load balancers and service meshes use readiness to stop routing requests to broken pods.

Do not use readiness to restart pods — use liveness for that.

Readiness fails when embedding API errors exceed threshold for 30s; traffic drains to healthy replicas.