# Deregistration Delayとは?

> deregistration delayは、削除中のtargetをドレインする間、load balancerが処理中のリクエストを引き続き処理する時間で、アクティブな接続をきれいに完了させます。

Source: https://latchkey.dev/ja/learn/ci-glossary/deregistration-delay-definition  
Updated: 2026-06-26

## なぜ重要か

drainingは、deployやscale-downの間のリクエストの取りこぼしや壊れた応答を回避します。短すぎると遅いリクエストを切断し、長すぎるとrolloutが停滞します。

## Where this shows up in a real pipeline

A definition is only useful if you can recognise the thing in your own logs. In practice this concept surfaces when a job behaves differently between a local run and a runner, which is where most CI debugging starts.

- Compare a local run against a CI run of the same command before assuming a tool is at fault.
- Runner environments differ in shell, PATH, TTY availability, available memory, and disk. Most CI-only behaviour traces to one of those five.
- Pin tool versions so an environment difference cannot be introduced by an unrelated update.

## FAQ

### What is What is deregistration Delay??

Deregistration delay, also called connection draining, is the interval a load balancer waits before fully removing a backend from rotation. During this window it stops sending new requests but lets existing ones complete. This prevents cutting off active connections when an instance is being replaced or scaled in.

### Why does What is deregistration Delay? matter in CI/CD?

Draining avoids dropped requests and broken responses during a deploy or scale-down. Setting it too short cuts off slow requests; too long stalls the rollout.

---

Latchkey runs CI/CD that repairs its own failures. Agent entry points: https://latchkey.dev/agent.txt, https://latchkey.dev/openapi.json, https://latchkey.dev/llms.txt