capabilities.work

Run a One-Off Task Pod and Collect Its Output

A 4-step recipe with Kubernetes. We can help you get access.

Recipe · 4 steps

Run a One-Off Task Pod and Collect Its Output

with Kubernetes

Create a run-to-completion pod, read its output, and delete it.

  1. Create the task pod with restartPolicy Never so a failed command is left in the Failed phase for inspection rather than being restarted forever.
  2. Read the pod until it leaves the Pending phase. Succeeded means the command exited zero; Failed means it exited non-zero.
  3. Read the container log as the task's output. This is the only place the command's stdout is retained once the pod is deleted.
  4. Delete the completed task pod so finished one-off runs do not accumulate. The log has already been collected by the previous step.
What you'll need to have ready
  • namespace: The namespace to run the task pod in.
  • taskName: Name for the one-off pod and its container.
  • image: Container image that carries the task.
  • command: Entrypoint command array, replacing the image ENTRYPOINT.
  • args: Arguments to the entrypoint, replacing the image CMD.
  • env: Environment variables to set in the task container.
  • serviceAccountName: Service account granting the task pod its API permissions.
  • tailLines: Number of log lines to collect from the end of the log.
For your developer

Arazzo workflow kubernetes:run-one-off-pod, 4 operations: createNamespacedPod → getNamespacedPod → readNamespacedPodLog → deleteNamespacedPod. Specs on APIs.io.

More recipes with Kubernetes

Attach a HorizontalPodAutoscaler to a DeploymentAudit Cluster Node CapacityPublish a ConfigMap and Roll Out Its ConsumersDeploy an Application into a New NamespaceExpose a Deployment Behind a ServiceApply Governance Labels to a NamespaceProvision a ClusterRole If MissingQuarantine a Failing Pod for InspectionRestart a Pod by Deleting ItRetire an Application from a Namespace