Operating Batch Workloads

Running batch workloads reliably in production takes more than getting the Job/CronJob YAML right once - it needs an operating discipline around every run. For each execution, record enough to answer "what happened" later: an identifiable run/input reference, which attempt number this was, how long it took, and its final outcome. Design the workload to emit progress that survives losing a Pod partway through (a durable checkpoint somewhere outside the Pod itself), rather than keeping all progress only in memory. Critically, set up alerting for absence, not just failure: a CronJob that never starts at all - because of a missed schedule window, a paused CronJob nobody remembered to resume, or a controller outage - produces no error of any kind from the application's point of view, because the application never ran to report one. During an active incident, kubectl patch a CronJob's suspend: true field to safely stop new runs without deleting the object, and when investigating, check events on both the CronJob and its child Jobs - each layer can hide a different part of the story.

Analogy: Kubernetes is a thermostat, not a remote control. You declare the temperature; independent controllers measure reality and keep acting. Debugging means finding which sensor, rule or actuator prevents convergence.

Read the object as evidence

kubectl get RESOURCE NAME -n NAMESPACE -o yaml
kubectl describe RESOURCE NAME -n NAMESPACE
kubectl get events -n NAMESPACE --sort-by=.metadata.creationTimestamp
kubectl explain RESOURCE.spec

Do not memorize these as a ritual. The first command exposes desired and observed state, the second connects conditions and events, the third supplies a timeline, and the fourth checks the server's schema. Compare what the controller was asked to do with what it reports doing. Then test the narrowest hypothesis.

Scenario: A CronJob is configured with a startingDeadlineSeconds that turns out to be too small, and during an unrelated cluster outage the controller itself is briefly unavailable past that deadline. When it recovers, it correctly and intentionally skips the missed run rather than starting it dangerously late - this is the CronJob doing exactly what it was configured to do, not malfunctioning. But from the business's perspective, an expected report or job simply never happened, and nothing about the Kubernetes objects themselves reports an error. Monitoring has to actively detect missing expected executions ("this should have run by now and did not"), because Kubernetes has no concept of alerting on an absence.

Operate batch from records

Label Jobs with a run identifier, emit structured progress and durable checkpoints, and retain enough failed Pods for diagnosis. kubectl logs job/NAME --all-containers is useful for simple runs; parallel Jobs need aggregation by labels and timestamps. Alert on age, repeated failure, missed schedule and business-output freshness—not merely Job failed.

Scenario: The Job succeeds nightly but its source feed stopped updating a week ago. Infrastructure success masks stale business output. Validate input watermark and produced record counts.
Tip: A safe rerun procedure states which checkpoint to resume, how duplicates are prevented, how downstream effects are reconciled and how success is verified. Practice it before the first production partial failure.

Production reasoning

Ask four questions: Who owns this object? What dependency must become ready next? Which controller reports the blocking condition? What evidence would disprove my current theory? This prevents symptom-driven changes. Record the context, namespace, object generation, image digest and recent rollout before mutation; a recreated Pod may erase the evidence you needed.

Warning: Running is not the same as ready, healthy, durable or correct. Kubernetes status is layered. Confirm the application-level outcome as well as the object state.
Goal: Put this model into practice in the Kubernetes lab cronjob-concurrency-deadline. Open /labs/kubernetes, choose cronjob-concurrency-deadline, predict the failure path before changing anything, then use the simulator's check command to validate the finished state.

Deliberate practice

Before the lab, write the expected object relationship and the first three commands you will run. Afterward, explain why the fix converged and name one tempting change that would only mask the symptom. Repeat using an explicit namespace and a structured output format. This prediction-observation-explanation loop is what turns command familiarity into production judgment.

45-minute investigation

  1. Map (5 min): draw the owner-to-child chain and mark every namespace, selector, identity and dependency involved.
  2. Predict (5 min): write one expected status condition, one likely event and one log or metric signal before opening the lab.
  3. Observe (10 min): collect YAML, describe output and ordered events. Do not mutate state. Record which observation disproves your first theory.
  4. Repair (15 min): make the smallest declarative correction, watch the responsible controller converge, and verify the user-facing path rather than stopping at Running.
  5. Stress (10 min): change one relevant constraint - replica count, label, readiness, resource value or placement rule - predict the outcome, observe it, then restore the known-good declaration.
Tip: Keep a short incident note with symptom, evidence, hypothesis, change, result. Across four sections this produces a reusable runbook instead of a pile of remembered commands.