Delay readiness

Docker compatibleDocker Swarm compatibleKubernetes compatiblePodman compatibleProxmox LXC not yet implemented

Exampleexamples/ready-after

This guide shows you how to delay when Sablier reports an instance as ready with the sablier.ready-after label:

# compose.yml
services:
  myapp:
    image: myapp:latest
    restart: unless-stopped
    labels:
      - "sablier.enable=true"
      - "sablier.group=myapp"
      - "sablier.ready-after=30s"  # wait 30 s after started/healthy before unblocking requests

Some services are started (or pass their health check) before they finish initialising. For example, a JVM application opens its HTTP port before loading all caches, a database accepts TCP connections before it’s ready for queries, and a container without a health check may need a few extra seconds after start-up before it can serve traffic.

Setting sablier.ready-after introduces a mandatory settling delay. Once the provider reports the instance as ready (whether that means it has started or passed its health check), Sablier continues to return a not-ready response to any blocking or dynamic request until the grace period elapses.

    flowchart LR
    start[Instance starts] --> ready[Provider reports ready]
    ready --> delay[ready-after grace period<br/>Sablier still returns not-ready]
    delay --> unblock[Sablier reports ready]
  

The value is a Go duration string. Valid examples:

If the label is absent or set to an unparseable value, no extra wait is applied.