Jobs

Use a Kubernetes Job when you want to run work to completion instead of keeping a long-running service online.

Typical examples include data preparation, one-time model setup, migrations, rendering tasks, evaluation runs, and ad hoc maintenance work.

Choose a workflow

Workflow	Best for	Current RemoteGPU support
`Console`	Inspecting namespace state	Console support for creating and managing `Job` resources is not available yet
`kubectl`	Professional operators running native Kubernetes batch workloads	Supported today with namespace kubeconfig, API key auth, and admission policy checks

Use case	Why `Job` fits
Run work once and stop	A job tracks completion instead of keeping pods running forever
Retry failed work	Kubernetes can retry failed pods until completion criteria are met
GPU or CPU batch execution	The same SKU and admission rules apply to batch workloads
CI or scheduled automation triggers	You can create jobs from local tooling or a CI system with `kubectl`

If you need recurring execution, use a CronJob instead.

Area	Supported behavior
Resource type	`batch/v1` `Job`
Access path	`kubectl` with namespace kubeconfig and a Kubernetes exec API key
Scheduling controls	Standard Kubernetes job controls such as completions, backoff limit, and parallelism
SKU policy	The same RemoteGPU SKU label rules used for deployments also apply to job pod templates
Security policy	Job pod templates must pass the same admission restrictions as other workloads

RemoteGPU does not yet provide a dedicated console flow for creating or editing Job resources.

Use the console to create the namespace, download kubeconfig, inspect pods, events, and logs, then create the Job with kubectl.

Action	Result
Create a job with `kubectl`	The job runs in the target namespace under the same admission and quota rules as other workloads
Inspect the namespace in the console	You can still use the console for supporting resources such as pods, logs, and events
Re-run work	Create a new job instance or re-apply a new manifest, depending on your workflow

Topic	Guidance
Public traffic	Jobs are not a public HTTP surface; if the resulting application needs traffic, that is usually a deployment plus service plus ingress flow instead
Long-lived services	Use a deployment instead of a job
Repeated schedules	Use a CronJob for recurring work