Kubernetes Workload Resources
External
Internal
Overview
Workload resources exist to create and manage Pods, which should not usually be created directly. Various workload resources implement different operational patterns. The pod-managing resources described here are some times referred to as higher-level pod controllers or "templated controllers".
From the simplest to the most complex ones, the workload resources are Jobs, CronJobs, ReplicaSets, Deployments, DaemonSets and StatefulSets.
Jobs and CronJobs are aimed mainly at running individual pods, as part of one-off or repeated executions.
A ReplicaSet is a workload resource aimed at maintaining a stable set of replica Pods running at any given time, guaranteeing the availability of a specified number of identical Pods.
The Deployment is a resource that delegates to a ReplicaSet. The deployment enhances a ReplicaSet semantics with things like deployment strategy. An older, deprecated alternative to Deployments are ReplicationControllers.
StatefulSet is the workload API object used to manage stateful applications. DaemonSet runs one instance of a pod on every node in the cluster.
Workload Resources
- Job
- CronJob
- ReplicaSet
- Deployment
- ReplicationController (deprecated)
- StatefulSet
- DaemonSet
Controllers
Each of the workload resources specified above is backed by a controller and state reconciliation loop. The controller for the resource handles replication and rollout and automatic healing in case of Pod failure. For example, if a Node fails, the controller notices that Pods on that Node have stopped working and creates a replacement Pod. The scheduler places the replacement Pod onto a healthy Node. All these controllers are supervised by the controller manager.