Kubernetes Jobs
Why we need kubernetes Jobs
As you learned in the previous chapters, a Pod created via a Deployment, StatefulSet, or DaemonSet, runs continuously. When the process running in one of the Pod’s containers terminates, the Kubelet restarts the container. The Pod never stops on its own, but only when you delete the Pod object. Although this is ideal for running web servers, databases, system services, and similar workloads, it’s not suitable for finite workloads that only need to perform a single task.
A finite workload doesn’t run continuously, but lets a task run to completion. In Kubernetes, you run this type of workload using the Job resource. However, a Job always runs its Pods immediately, so you can’t use it for scheduling tasks. For that, you need to wrap the Job in a CronJob object. This allows you to schedule the task to run at a specific time in the future or at regular intervals.
In this chapter you’ll learn about Jobs and CronJobs.
Introducing the Job resource
The Job resource resembles a Deployment in that it creates one or more Pods, but instead of ensuring that those Pods run indefinitely, it only ensures that a certain number of them complete successfully.
As you can see in the following figure, the simplest Job runs a single Pod to completion, whereas more complex Jobs run multiple Pods, either sequentially or concurrently. When all containers in a Pod terminate with success, the Pod is considered completed. When all the Pods have completed, the Job itself is also completed.
Three different Job examples. Each Job is completed once its Pods have completed successfully.
Creating Job resource
Displaying the detailed Job status
To see more details about the Job, use the kubectl describe command as follows:
Introducing the cronjob resource
When you create a Job object, it starts executing immediately. Although you can create the Job in a suspended state and later un-suspend it, you cannot configure it to run at a specific time. To achieve this, you can wrap the Job in a CronJob object.
In the CronJob object you specify a Job template and a schedule. According to this schedule, the CronJob controller creates a new Job object from the template. You can set the schedule to do this several times a day, at a specific time of day, or on specific days of the week or month. The controller will continue to create Jobs according to the schedule until you delete the CronJob object. The following figure illustrates how a CronJob works.
Figure 17.11 The operation of a CronJob
As you can see in the figure, each time the CronJob controller creates a Job, the Job controller subsequently creates the Pod(s), just like when you manually create the Job object. Let’s see this process in action.
Creating a CronJob
The following listing shows a CronJob manifest that runs a Job every minute.
Inspecting the CronJob status in detail
Apply the manifest file to create the CronJob. Use the kubectl get cj command to check the object:
$ kubectl get cj
NAME SCHEDULE SUSPEND ACTIVE LAST SCHEDULE AGE
aggregate-responses-every-minute * * * * * False 0 2s
kubectl get cj aggregate-responses-every-minute -o yaml
As you can see, the status section of a CronJob object shows a list with references to the currently running Jobs (field active), the last time the Job was scheduled (field lastScheduleTime), and the last time the Job completed successfully (field lastSuccessfulTime). From the last two fields you can deduce whether the last run was successful.
Inspecting Events associated with a CronJob
To see the full details of a CronJob and all Events associated with the object, use the kubectl describe command as follows: