Traefik Proxy ↔ Docker Provider: Configuring Connection‑Draining Timeout During Graceful Reload

Goal

Determine whether the connection‑draining timeout that Traefik applies during a graceful reload can be configured or influenced, so that operators can guarantee zero‑downtime migration for a small application.

Uncertainty

The timeout is not exposed through Traefik’s HTTP API or provider interfaces; the default value (~30 seconds in v2.7) is internal and documentation does not state if it can be adjusted via static configuration, environment variables, or global settings, nor whether it applies uniformly to HTTP/2 streams.

Questions

• Is the connection‑draining timeout configurable via static file, environment variable, or any hidden setting in Traefik 2.x?
• If it is not configurable, what mechanisms exist to affect how long in‑flight requests are allowed to finish during a reload?
• Does this timeout govern the lifetime of individual HTTP/2 streams, or are streams handled differently during the drain period?