Appearance
Kubernetes on RemoteGPU
Run containerized applications and batch tasks on RemoteGPU with Kubernetes. Start by creating a namespace, which groups your applications, networking, storage, and access permissions.
Use the console to fill in setup forms, or kubectl to apply YAML files and run commands from your terminal or CI system. Both manage the same resources, so changes made with one are visible in the other.
Before you begin
Sign in to the RemoteGPU console and choose a container image and a plan for your workload's CPU, memory, and GPU needs.
Quick start
Choose how you want to create and manage workloads.
- Open Kubernetes / Deployments.
- Select an existing namespace or create one.
- Follow Deployments to create a long-running app, Jobs for a one-off task, or CronJobs for scheduled work.
- Inspect the workload's status, pods, logs, and events in the console.
These console operations do not require an API key or kubeconfig.
To keep files when a pod is replaced, attach storage. Create a Service so other applications in the cluster can connect to yours. Add an Ingress if it needs a public web address. Store application credentials in Secrets.
Supported resources
Choose the Kubernetes resource for what you want to run, then check which tool you can use to manage it.
| Resource | Use | Console | kubectl |
|---|---|---|---|
Deployment | Long-running apps | Supported | Supported |
StatefulSet | Stable pod names and per-replica storage | Use kubectl | Supported |
Job | One-off tasks | Create and inspect | Supported |
CronJob | Scheduled tasks | Create and inspect | Supported |
Service | Internal network access | Supported | Supported |
Ingress | Public HTTP routing | Supported | Supported |
ConfigMap | Application configuration | Use kubectl | Supported |
Secret | Application credentials | Use kubectl | Supported |
PersistentVolumeClaim | Persistent data | Use kubectl | Supported |
Pod | Workload inspection | Inspect | Inspect only |
| Logs and events | Troubleshooting | Supported | Supported |
Important limits
Follow these requirements when creating resources.
| Area | Requirement | Details |
|---|---|---|
| Services | Use ClusterIP; NodePort and externalIPs are not supported | Services |
| Ingress hostnames | Use <prefix>--<namespace-token>.apps.remotegpu.ai | Ingresses |
| Ingress TLS | RemoteGPU sets up HTTPS for its generated hostnames; do not add your own TLS settings | Ingresses |
| Ingress routing | HTTP rules that route to a service, using the supported annotations | Ingresses |
| Secrets | Up to 128 per namespace; avoid reserved name prefixes | Secrets |
| Storage | Use supported storage classes and fixed sizes | Storage |
| Pod creation | Use Deployment, StatefulSet, Job, or CronJob; direct Pod creation is not supported | Deployments |
Read next
- Read Deployments to run or update an application.
- Read StatefulSets for stable pod names and per-replica storage.
- Read Services to expose a workload inside the cluster.
- Read Ingresses to publish a service with HTTP routing.
- Read Storage to create persistent volumes.
- Read Secrets to attach application credentials.
- Read Jobs for one-off batch work.
- Read CronJobs for scheduled batch work.