Introduction

If you are using Traefik with Kubernetes and want longer backend timeouts, the answer depends on which Traefik generation and provider features you are actually using. In current Traefik Kubernetes setups, there usually is not a simple ingress annotation such as readtimeout or writetimeout that directly changes all backend timeout behavior. The modern pattern is to define a ServersTransport resource and reference it from the Ingress service annotation.

Why Old Annotation Advice Is Often Wrong

A lot of older answers mention annotations like readtimeout, writetimeout, or idletimeout directly on the Ingress. In current Traefik documentation for Kubernetes Ingress, those are not the main supported per-service timeout knobs.

Instead, Traefik exposes a service.serverstransport annotation that points to a ServersTransport definition. That resource carries the forwarding timeout settings for backend connections.

So if your existing annotation does nothing, the issue may be that you are following Traefik v1-era guidance against a v2 or v3 deployment model.

The Current Traefik Pattern

First, create a ServersTransport custom resource with the timeout values you want.

1apiVersion: traefik.io/v1alpha1
2kind: ServersTransport
3metadata:
4  name: slow-backend-timeouts
5  namespace: default
6spec:
7  forwardingTimeouts:
8    dialTimeout: 30s
9    responseHeaderTimeout: 300s
10    idleConnTimeout: 300s

Then reference it from your Ingress using the Traefik annotation:

1apiVersion: networking.k8s.io/v1
2kind: Ingress
3metadata:
4  name: my-app
5  namespace: default
6  annotations:
7    traefik.ingress.kubernetes.io/service.serverstransport: slow-backend-timeouts@kubernetescrd
8spec:
9  ingressClassName: traefik
10  rules:
11    - host: example.com
12      http:
13        paths:
14          - path: /
15            pathType: Prefix
16            backend:
17              service:
18                name: my-app-service
19                port:
20                  number: 80

That is the supported route when you need longer connection establishment or response-header wait times for the upstream service.

What the Timeout Fields Mean

The most relevant forwarding timeout settings are:

• 'dialTimeout: how long Traefik waits to establish the backend connection'
• 'responseHeaderTimeout: how long Traefik waits for backend response headers after sending the request'
• 'idleConnTimeout: how long an idle keep-alive backend connection can stay open'

These settings affect the connection between Traefik and your service, not every possible client-side timeout in the stack.

When You Need EntryPoint-Level Settings Instead

Some timeout behavior lives at the entrypoint or static configuration layer rather than the Kubernetes Ingress annotation layer. If your issue is incoming client connection behavior rather than upstream service behavior, you may need to change Traefik static configuration instead of the Ingress resource.

That is why timeout debugging in Traefik can be confusing: there is more than one place where timeouts can exist.

How to Verify the Configuration

After applying the ServersTransport, check these points:

• the CRD exists in the namespace you referenced
• the annotation uses the correct provider suffix such as @kubernetescrd
• your Traefik installation actually has the CRD provider enabled
• the service you are routing to is the one affected by the timeout

If the reference is wrong, Traefik may silently keep using its default transport behavior.

Common Pitfalls

A common mistake is copying direct timeout annotations from old Traefik examples and expecting them to work in current Kubernetes Ingress deployments.

Another mistake is changing only the Ingress resource when the real timeout bottleneck is at the application server, load balancer, or client layer.

A third issue is forgetting the @kubernetescrd suffix when referencing the ServersTransport. Without the provider suffix, Traefik may not resolve the object you intended.

Summary

• Current Traefik Kubernetes setups usually handle backend timeouts through ServersTransport, not legacy direct timeout annotations
• Reference the transport with traefik.ingress.kubernetes.io/service.serverstransport
• Use forwardingTimeouts for backend connection and response-header behavior
• EntryPoint or static settings may still matter for client-side timeout behavior
• Check version-specific Traefik docs before trusting older annotation examples