Docs

Volcengine Kubernetes Engine

Service and Ingress FAQs

How do I handle a dropped persistent connection?

How do I handle a dropped persistent connection?

Issue description

Services that require a persistent connection (for example, live streaming and online meetings) experience brief connection interruptions.

Cause analysis

The persistent connection is disrupted when Nginx restarts or reloads its configuration.

During an Nginx Ingress Controller scale-in, requests may exit before they are fully processed due to a lack of graceful shutdown support.

Solution

For connection disruptions caused by an Nginx restart or configuration reload, configure worker-shutdown-timeout to set the connection timeout for persistent connections after the reload. Modify the ConfigMap of the ingress-nginx-controller.

kubectl edit configmap -n ingress-nginx ingress-nginx-controller

apiVersion: v1
data:
disable-access-log: "true"
enable-multi-accept: "true"
enable-underscores-in-headers: "true"
ignore-invalid-headers: "false"
max-worker-connections: "500000"
max-worker-open-files: "1000000"
worker-cpu-affinity: auto
worker-shutdown-timeout: "240"

For dropped persistent connections caused by an Nginx Ingress Controller scale-in, if the client has a retry mechanism, it can recover automatically.

Last updated: 2026.07.23 14:31:46