Ingresses

A Kubernetes Ingress routes external HTTP traffic to a service in your namespace. Use ingresses on RemoteGPU to publish service-backed web apps on a managed apps.remotegpu.ai hostname.

RemoteGPU supports Kubernetes Ingress for teams that manage Kubernetes resources directly. The console can help you create a simple ingress, and kubectl can manage the same resource directly.

Before you create an ingress

Before you create an ingress, make sure you have:

• a running workload, because ingress only routes traffic
• a ClusterIP service for the ingress backend
• a host that matches <prefix>--<namespace-token>.apps.remotegpu.ai
• the platform ingress class, traefik

HTTPS for managed hosts is served by the shared *.apps.remotegpu.ai wildcard certificate.

Choose a workflow

Use the console when you want guided ingress creation for an existing service. You choose the service, port, host prefix, and path in the UI.

Use kubectl when you manage ingress rules from manifests. Create a standard networking.k8s.io/v1 Ingress that follows the RemoteGPU ingress profile.

Create an ingress

From the console:

1. Open Kubernetes / Ingresses
2. Select a namespace
3. Click Create ingress
4. Choose the target service
5. Choose the backend port
6. Enter the ingress details

Console field	What it means
Ingress name	Kubernetes ingress name in the namespace
Host prefix	The left side of the enforced hostname pattern
Path	HTTP path for the backend rule
Max request body size	Ingress request limit. Default 8m, adjustable up to 16m
Request timeout	Upstream request timeout. Uses the ingress-controller default unless you select an override
Service	Existing service that should receive traffic
Port	Service port the ingress backend should target

The console shows the generated hostname before you create the ingress. For a namespace token such as 123456780000, a host prefix of demo-web becomes demo-web--123456780000.apps.remotegpu.ai.

TLS is provided automatically for generated apps.remotegpu.ai hosts.

Supported ingress profile

The standard ingress flow is:

Deployment
  -> Service
  -> Ingress
  -> public HTTP route

Area	Supported behavior
API version	networking.k8s.io/v1 Ingress
Backend type	Service-backed HTTP rules
Namespace boundary	The ingress must point to services in the same namespace
Hostnames	Hosts must match <prefix>--<namespace-token>.apps.remotegpu.ai for the target namespace
Ingress class	Use traefik
Path routing	Standard HTTP path routing with pathType
TLS	Managed apps.remotegpu.ai hosts use the platform wildcard certificate

Advanced ingress options

The console supports request body size and request timeout settings. The same settings can also be managed with kubectl by using the supported Traefik annotations and resources below.

For namespace-local Traefik Middleware references from an Ingress, prefix the middleware name with the current Kubernetes namespace and add the @kubernetescrd suffix. For example, in namespace rgpu-owner-default-12345678, a middleware named demo-body-limit becomes rgpu-owner-default-12345678-demo-body-limit@kubernetescrd.

For request body size limits, RemoteGPU accepts these buffering.maxRequestBodyBytes values:

• 1048576 (1m)
• 2097152 (2m)
• 4194304 (4m)
• 8388608 (8m)
• 16777216 (16m)

Without a buffering middleware, the managed ingress profile defaults to 8m.

For request timeouts, RemoteGPU accepts these platform shared Traefik references:

• kube-system-public-timeout-60s@kubernetescrd
• kube-system-public-timeout-120s@kubernetescrd
• kube-system-public-timeout-300s@kubernetescrd
• kube-system-public-timeout-600s@kubernetescrd

Do not add cert-manager.io/cluster-issuer or spec.tls for managed apps.remotegpu.ai hosts. HTTPS is terminated with the platform wildcard certificate.

The console workflow still offers body-size and timeout fields. RemoteGPU translates those selections into the allowlisted Traefik middleware and shared transport references above.

Limits

Pattern	Status
NodePort or externalIPs service exposure	Use ingress-backed HTTP routing instead
Hosts outside the namespace-scoped pattern	Rejected
defaultBackend	Not supported
Non-service ingress backends	Not supported
Per-ingress cert-manager annotations or custom TLS secrets	Not supported for managed apps.remotegpu.ai hosts
Legacy nginx.ingress.kubernetes.io/* annotations	Not supported
Unsupported Traefik annotations or CRDs	Not supported outside the documented allowlist
Custom-domain onboarding through the Kubernetes console flow	Not supported

How ingress shows up in the console

The console reads live ingress resources from your namespace.

Action	Result
Create ingress with kubectl	It appears in the console
Create ingress in the console	It appears in kubectl
Remove the target service or service port	The console shows an error state for the ingress

Troubleshooting

Symptom	What to check
Ingress creation is rejected because of the host	Use the namespace-scoped host pattern shown in the error response: <prefix>--<namespace-token>.apps.remotegpu.ai
The generated URL returns no application response	Confirm the ingress backend points to an existing ClusterIP service and that the service has ready endpoints
HTTPS configuration is rejected	Do not add spec.tls or cert-manager annotations for managed apps.remotegpu.ai hosts
Large requests fail	Use the console request body size setting, or apply one of the supported Traefik buffering values with kubectl
Long requests time out	Use the console request timeout setting, or apply one of the supported shared timeout references with kubectl

Read next

• Read Services to create the internal backend.
• Read Deployments to run the workload first.