Workload overview

Workloads in a serverless cluster

Built on a Kubernetes cluster, KSK is fully compatible with the Kubernetes-native resource creation and management methods. Workload creation and management in a serverless cluster are basically the same as those in a container cluster. However, because a serverless cluster does not contain physical nodes, some resource and parameter settings are not available for it:

• The DaemonSet workload is unavailable for a serverless cluster.
• Host directories and file systems cannot be used as volumes.
• When you create a workload, health check is unavailable in the container settings, and you cannot set the container to a privileged type.
• NodePort access to a workload is not supported.
• When you create a workload, actual resources will be allocated to each pod based on the resource limits in the container settings.

Workload types

You can create the following types of workloads in a serverless cluster based on your business requirements. For more information, see the corresponding topics.

Deployment

A Deployment manages stateless applications in the cluster. It provides declarative updates for pods and replica controllers in the cluster.

StatefulSet

A StatefulSet manages stateful applications in the cluster. It provides persistent identifiers for pods in the cluster. The identifiers will retain across rescheduling or destruction of the pods. When you create a persistent storage instance, you can map pods to volumes in a one-to-one manner by using identifiers.

Job

A job performs one-time batch processing. It ensures the successful termination of one or more pods that perform batch processing. Unlike a Deployment or a StatefulSet that perpetually runs, a job will automatically exit after completion. Therefore, it is more suitable for one-time processing such as data processing and migration.

CronJob

A CronJob runs a job at a given time point or periodically on a given schedule.