Define a Kubernetes job task

Theme: Configure
Who Is It For? Automation Engineer

What is it?

A Kubernetes job task tells OpCon what container image to run, what command and arguments to pass to it, how many pods to spin up, and what resource limits to apply. When OpCon runs the job, the connector submits these parameters to the Kubernetes cluster.

• Use this procedure when you need to schedule a containerized workload through OpCon
• Use this procedure when configuring resource constraints for jobs that run in a shared Kubernetes cluster

The Kubernetes Connector supports the following task type:

Task type	Description
Job	Runs a command within a container in a Kubernetes cluster

Define a Job task

Prerequisite: The Kubernetes agent must be defined and communicating before a job task can be submitted. See Agent definition.

To define a Kubernetes job task, complete the following steps:

1. In Solution Manager, select Library.

2. From the Administration menu, select Master Jobs.

3. Select +Add.

4. In the Schedule field, select the schedule name from the list.

5. In the Name field, enter a unique name for the job within the schedule.

6. In the Job Type field, select Kubernetes Job from the list.

7. In the Task Type field, select Job from the list.

8. Select the Task Details button.

9. In the Integration Selection section, select the Kubernetes agent previously defined.

10. In the Task Configuration section, complete the following fields:

    Field	Description	Required	Default
    Environment Variables	Multiple environment variables can be added by selecting the + AddItem button. Each variable has a Name and a Value. To read the value from a Kubernetes Secret instead, select the Value From option and enter the Key within the secret and the Name of the secret. When Value From is selected, the Value field is not required. A value cannot contain = or ;	No
    Image	The Docker Registry image that provides the runtime environment for the command	Yes
    Name Space	The Kubernetes namespace to run the job in	Yes	default
    Job Name	The name assigned to the Kubernetes job object	Yes
    Container Name	The name assigned to the container when the image runs	Yes
    Command	The command to run inside the container. Separate multiple values with a comma (,); a value cannot contain a comma	Yes
    Arguments	The arguments to pass to the command. Separate multiple values with a comma (,); a value cannot contain a comma	No
    Volume Mounts	Optional information about a volume that should be mounted for the task to process. Add definitions by selecting the + AddItem button. Enter the Name, the Mount Path, the Claim Name or Secret Name, and select the Read Only option if the volume is read only	No
    Resources	The CPU and memory requests and limits applied to each pod, set in the four fields that follow	Yes
    Request CPU	The initial CPU allocation guaranteed to each pod	Yes	250m
    Request Memory	The initial memory allocation guaranteed to each pod	Yes	512Mi
    Limit CPU	The maximum CPU each pod may consume	Yes	500m
    Limit Memory	The maximum memory each pod may consume	Yes	1Gi
    Pods to Complete	The number of pod runs the job needs. The job ends when the number of pods that have succeeded or failed reaches this value, and finishes successfully only if all of them succeeded	Yes	1
    Parallel Executions	The number of pods that may run concurrently	Yes	1
    Restart Policy	The restart policy to set for the job: Never or OnFailure	Yes
    BackOffLimits	The number of retries Kubernetes allows on failure. With Restart Policy set to Never, the first failed pod fails the job and the connector removes the Kubernetes job before it can retry. Retries happen with OnFailure, where Kubernetes restarts the container within the same pod	Yes	6

11. Select the Save button. The job is added to the schedule.

Killing a job

caution

Killing the job in OpCon does not delete the Kubernetes job. Its pods keep running in the cluster, and a later run with the same Job Name in that namespace fails until the job is deleted. Delete the job in the cluster, for example with kubectl delete job <name> -n <namespace>.

Job log

The job log contains the job settings, the environment variables with their values, and the output of each pod. Environment variables entered as plain values appear in the job log as entered, so use Value From for sensitive values.

FAQs

How do I pass multiple commands or arguments?
Separate multiple values with a comma (,). For example, to pass two arguments, enter --config,/etc/app/config.yaml in the Arguments field.

What is the difference between Pods to Complete and Parallel Executions?
Pods to Complete sets the total number of pod runs the job needs. Parallel Executions sets how many of those pods can run at the same time. For example, setting Pods to Complete to 4 and Parallel Executions to 2 runs two pods at a time until four have finished. The job finishes successfully only if all four succeeded.

What happens if a pod exceeds its CPU or memory limit?
Kubernetes terminates any pod that exceeds its memory limit. CPU throttling applies when a pod exceeds its CPU limit. Set limits high enough to avoid unexpected pod terminations.

Can I use OpCon global properties in task fields?
Yes. Global properties using the [[property_name]] token syntax are supported in all task configuration fields.

Glossary

Image — A packaged container image stored in a Docker Registry that defines the runtime environment and application code for the pod.

Namespace — A Kubernetes isolation boundary that logically separates workloads within a cluster. Jobs submitted to a namespace are only visible to and affected by resources within that namespace.

Pods to Complete — The number of pod runs the job needs. OpCon waits until this many pods have finished, and marks the job complete only if all of them succeeded.

Resource requests — The minimum CPU and memory that Kubernetes guarantees to a pod. The cluster scheduler only places a pod on a node that has at least this much capacity available.

Resource limits — The maximum CPU and memory a pod may consume. Kubernetes enforces limits by throttling CPU and terminating pods that exceed memory limits.