Intermittent connection pool exhaustion in Traefik serversTransport

Intermittent connection pool exhaustion in Traefik serversTransport

The goal is to identify a suitable value for the maxIdleConnsPerHost setting that prevents the transport pool from being exhausted during traffic spikes.

Constraints include the operating system’s file‑descriptor limit, the interaction with idleTimeout, and the absence of an automatic scaling algorithm in Traefik’s documentation.

• Which observable metrics indicate that maxIdleConnsPerHost is too low for the current load?
• How can an operator increase maxIdleConnsPerHost safely without exceeding the host’s file‑descriptor capacity?
• Is there a recommended heuristic or formula for setting maxIdleConnsPerHost based on measured peak concurrent requests per backend host?