Services

A Kubernetes Service gives pods a stable internal address and port. Use services on RemoteGPU when other workloads or ingresses need to reach a deployment.

In RemoteGPU, services are namespace-scoped Kubernetes resources. They are your resources, whether you create them from the console or with kubectl.

Typical reasons to create a service

A service is useful for:

• Internal app-to-app traffic through a stable in-cluster target.
• A stable target while deployment pods restart or scale.
• A backend target before creating an ingress.

Choose a workflow

Use the console when you want guided service creation for an existing deployment. The console reads the deployment selector and creates a ClusterIP service.

Use kubectl when you want to define the service selector and ports directly from a manifest.

Create a service

From the console:

1. Open Kubernetes / Services.
2. Select a namespace.
3. Click Create service.
4. Choose the deployment you want to expose.
5. Provide the service details.

Console field	Description
Service name	Kubernetes service name in the namespace
Service port	The port clients use to reach the service
Target port	The container port or named target on the selected deployment
Backend response timeout	Ingress backend response timeout for this service

The console reads the deployment selector and creates a ClusterIP service for that deployment.

After that, it behaves like any other Kubernetes service in the namespace.

Supported service profile

Area	Supported	Not supported
Service type	ClusterIP	NodePort
Public exposure model	Ingress backed by a service	externalIPs
Ownership model	Namespace-scoped Kubernetes Service	Platform-managed service proxy

Backend response timeout

Service backend response timeout controls how long RemoteGPU ingress routing waits for a response from this backend service. Configure it in the console or set:

metadata:
  annotations:
    traefik.ingress.kubernetes.io/service.serverstransport: ingress-backend-response-timeout-300s@file

Supported values are ingress-backend-response-timeout-60s@file, ingress-backend-response-timeout-120s@file, ingress-backend-response-timeout-300s@file, and ingress-backend-response-timeout-600s@file. If omitted, RemoteGPU uses 60 seconds. Ingress resources should not carry this annotation.

Console and kubectl work on the same service list

The Services page shows the live services in the namespace.

That includes services created from the console and services created with kubectl.

Service updates and deletions made with kubectl appear in the console after refresh.

Read next

• Read Ingresses to publish the service with HTTP routing.
• Read Deployments to run or update the backend workload.