Workload types: StatefulSets, Jobs, CronJobs & DaemonSets

0 / 15 lessons — 0%

Lesson 12 / 15

A Deployment assumes every replica is interchangeable — kill pod 2 of 3, and a new, differently-named pod takes its place, nobody cares which. That assumption breaks for a handful of real workloads, which is why Kubernetes has three more specialized workload types.

StatefulSets — for apps that care about identity

A database cluster's node 0 often needs to stay node 0 — same name, same storage, even across restarts. A StatefulSet gives each replica a stable, predictable name (db-0, db-1, db-2) and its own PersistentVolumeClaim that follows it across rescheduling, plus ordered, one-at-a-time startup and shutdown.

Unlike a Deployment, each StatefulSet pod keeps its own name and its own storage.

Jobs & CronJobs — run to completion, not forever

A Deployment's pods are meant to run forever. A Job runs a pod until it finishes successfully, then stops — perfect for a one-off migration script or batch task. A CronJob is a Job that fires on a schedule, using standard cron syntax.


# job.yaml — run a database migration once, to completion
apiVersion: batch/v1
kind: Job
metadata:
  name: db-migrate
spec:
  template:
    spec:
      containers:
        - name: migrate
          image: myregistry/web:1.5.0
          command: ["npm", "run", "migrate"]
      restartPolicy: Never

# cronjob.yaml — same idea, but nightly at 2am
apiVersion: batch/v1
kind: CronJob
metadata:
  name: nightly-report
spec:
  schedule: "0 2 * * *"
  jobTemplate:
    spec:
      template:
        spec:
          containers:
            - name: report
              image: myregistry/report:1.0
          restartPolicy: OnFailure

DaemonSets — exactly one pod per node

Some things need to run on every node, no more, no less — a log shipper, a monitoring agent, a network plugin. A DaemonSet guarantees exactly one copy of its pod on each matching node, and automatically adds a copy when a new node joins the cluster.

Workload type	Use it when
Deployment	Stateless, interchangeable replicas — most web apps and APIs
StatefulSet	Each replica needs a stable identity and its own storage — databases, message queues
Job	A task that runs once and finishes — migrations, batch processing
CronJob	A Job on a recurring schedule — nightly reports, cleanup tasks
DaemonSet	Exactly one pod per node — log shippers, node-level monitoring

Try it yourselfRun kubectl get daemonsets -A on any real cluster you have access to — most clusters already run a DaemonSet or two for networking or monitoring, even if nobody on the team created it by hand.

← Lesson 11 Next: Lesson 13 →