Some Processes are Job Oriented

Some containerized workloads are job-specific & not like a web-server.
Job-oriented workloads perform a task and then are done.

Some Processes are Job Oriented

Jobs Can Run in Containers

Say we have a simple math problem to represent a job to do: 8 + 3.
Docker can run a container to solve just this problem:

docker run ubuntu expr 8 + 3  
11  

# Inspect the container status
docker container ls -a
CONTAINER ID    IMAGE     COMMAND       CREATED           STATUS                      PORTS     NAMES
bc8765118836   ubuntu     "expr 8 + 3"  13 seconds ago    Exited (0) 13 seconds ago             mystifying_noether

Here, the container turns on, runs the job, returns the results, and then sits in an "Exited" state. The container is no longer running.

Kubernetes Can Manage These Types of Containers

A Pod Definition file:

apiVersion: v1
kind: Pod
metadata:
  name: simple-job-pod
spec:
  containers:
  - name: simple-job-box
    image: ubuntu
    command: ["expr", "8", "+", "3"]

Run this in k8s:

kubectl create -f simple-pod.yaml

# wait a few seconds, check the pod status
kubectl get pods
NAME            READY     STATUS    RESTARTS      AGE
simple-job-pod  0/1       Completed 1             1d

Here, the pod turns on, runs the container, and then sits in the "Completed" state.

K8s Wants The Container Running

K8s notices that the pod is in the "Completed" stage.
K8s recreates the container in an attempt to keep the container running: K8s thinks it is broken & wants it to run.
K8s will continue to run and recreate the container until a limited number of re-starts is reached.

Adjust The Restart Policy of a Pod

The spec of a pod has a restartPolicy field that can be overwritten - this will adjust how the pod auto-restarts.
By default, this is set to Always.
Other options are OnFailure and Never.

apiVersion: v1
kind: Pod
metadata:
  name: simple-job-pod
spec:
  containers:
  - name: simple-job-box
    image: ubuntu
    command: ["expr", "8", "+", "3"]
  restartPolicy: Never

Leveraging K8s Jobs

Make sure all pods perform the tasks assigned to them and exits
Manage creating an un-specified number of pods: the number of pods may be defined by the workload, itself

Replicasets make sure a specified number of pods is always running.
Jobs, on the other hand, runs a set of pods to perform a task to completion.

Creating a Job

# simple-job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: simple-job
spec:
  # the spec has a template
  template:
    # the template contains all spec contents of the pod
    spec:
      containers:
      - name: simple-job-box
        image: ubuntu
        command: ["expr", "8", "+", "3"]
      restartPolicy: Never

Manage the job with kubectl

# create the job
kubectl create -f simple-job.yaml

# check it out!
kubectl get jobs
NAME        DESIRED     SUCCESSFUL      AGE
simple-job  1           1               12s


# See the PODS that the job created
kubectl get pods
NAME                READY     STATUS    RESTARTS      AGE
simple-job-pod-986  0/1       Completed 1             2m

Inspect job output and cleanup the job

# see the logs of the pod
kubectl logs simple-job-pod-986
11

# kill the job
kubectl delete job simple-job

More Job Complexities

Run Multiple Ccntainers

K8s allows for running a job many times, using many instances of the job pod. Here, run the job pod 3x, in sequence:

# simple-job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: simple-job
spec:
  # run till the job completes 3x
  completions: 3
  template:
    spec:
      containers:
      - name: simple-job-box
        image: ubuntu
        command: ["expr", "8", "+", "3"]
      restartPolicy: Never

Manage the job with kubectl

# create the job
kubectl create -f simple-job.yaml

# check it out, wait a little longer for the 3 jobs to run
kubectl get jobs
NAME        DESIRED     SUCCESSFUL      AGE
simple-job  3           3               30s


# See the PODS that the job created
kubectl get pods
NAME                READY     STATUS    RESTARTS      AGE
simple-job-pod-a9j  0/1       Completed 1             2m
simple-job-pod-o3c  0/1       Completed 1             2m
simple-job-pod-pe2  0/1       Completed 1             2m

Kubernetes Will Restart until the Completions are Successful

NOTE: the 2nd pod only gets created after the first pod finishes.
Here, an image randomly fails:

# simple-job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: simple-job
spec:
  completions: 3
  template:
    spec:
      containers:
      - name: simple-job-box
        image: kodekloud/random-error
      restartPolicy: Never

Manage the job with kubectl

# create the job
kubectl create -f simple-job.yaml

# check it out, wait a little longer for the 3 jobs to run
kubectl get jobs
NAME        DESIRED     SUCCESSFUL      AGE
simple-job  3           3               30s


# See the PODS that the job created
kubectl get pods
NAME                READY     STATUS    RESTARTS      AGE
simple-job-pod-a9j  0/1       Completed 1             2m
simple-job-pod-o3c  0/1       Error     1             2m
simple-job-pod-pe2  0/1       Completed 1             2m
simple-job-pod-qwe  0/1       Error     1             2m
simple-job-pod-asd  0/1       Error     1             2m
simple-job-pod-zxc  0/1       Completed 1             2m

Run Jobs In Parralel

Here, the Kubernetes Job will create 3 pods off the bat:

# simple-job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: simple-job
spec:
  completions: 3
  parallelism: 3
  template:
    spec:
      containers:
      - name: simple-job-box
        image: kodekloud/random-error
      restartPolicy: Never

Here, K8s will "know" how many pods to recreate until success. It doesn't recreate 3 pods every time.

CronJobs

CronJobs can be scheduled, like CronTabs in linux: like something that runs a report, gets some data, and sends an email.

# simple-job.yaml

apiVersion: batch/v1
kind: CronJob
metadata:
  name: simple-cron-job
# cronJob spec
spec:
  schedule: "*/1 * * * *"
  jobTemplate:
    # job spec
    spec:
      completions: 3
      parallelism: 3
      template:
        # pod spec
        spec:
          containers:
          - name: simple-job-box
            image: kodekloud/random-error
          restartPolicy: Never

Epic Imperative line:

kk create cronjob throw-dice-cron-job --image=kodekloud/throw-dice --schedule="30 21 * * *"
cronjob.batch/throw-dice-cron-job created

Backing a Job Off After Failed attempts

K8s docs on backoffLimits:
jobs can fail, outright, after a set number of retries with spec.backoffLimit.

apiVersion: batch/v1
kindL Job
metadata:
  name: whalejob
spec:
  # JOB DEETS
  completions: 5
  backoffLimit: 6
  # the job spec
  template:
    # the "pod" spec that the job runs
    spec:
      containers:
      - name: whalesay
        image: docker/whalesay
        command: ["cowsay this is a string"]

Tags:

Kubernetes