×
Community Blog Kubernetes Batch Jobs

Kubernetes Batch Jobs

This tutorial describes how to run 3 types of Kubernetes jobs and introduces the concept of parallelism in computing.

By Alwyn Botha, Alibaba Cloud Tech Share Author. Tech Share is Alibaba Cloud's incentive program to encourage the sharing of technical knowledge and best practices within the cloud community.

This tutorial describes how to run 3 types of Kubernetes jobs:

  • Jobs using one Pod that runs once
  • Parallel Jobs with a fixed completion count
  • Parallel Jobs with a work queue

Other Pods run continuously forever ( for example a web server or a database ).

All 3 types of jobs above have a fixed ( batch ) job to do. They finish it then their status becomes completed.

1) Simple Job Example

Create the following as your simple job example YAML spec file.

Note the kind is Job. The rest of the spec is the same as that of a normal Pod.

nano myJob.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'echo Job Pod is Running ; sleep 10']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0

A job does its work using Pods.

Pods do work using its containers.

Containers work using Docker images. Our example above uses the Alpine Linux docker image to do a job: sleep 10 seconds.

Even this simple example will teach us about Kubernetes jobs.

Create the Job

kubectl create -f myJob.yaml

job.batch/myjob created

Let's list all Pods by running kubectl get pods several times.

kubectl get pods
NAME          READY   STATUS    RESTARTS   AGE
myjob-hbhtl   1/1     Running   0          2s

NAME          READY   STATUS    RESTARTS   AGE
myjob-hbhtl   1/1     Running   0          4s

NAME          READY   STATUS    RESTARTS   AGE
myjob-hbhtl   1/1     Running   0          6s

NAME          READY   STATUS    RESTARTS   AGE
myjob-hbhtl   1/1     Running   0          9s

NAME          READY   STATUS      RESTARTS   AGE
myjob-hbhtl   0/1     Completed   0          12s

NAME          READY   STATUS      RESTARTS   AGE
myjob-hbhtl   0/1     Completed   0          16s

We see this Pod ran for around 10 seconds. Then its status turns to Completed .

In the READY column we see that after 10 seconds the Pod is no longer ready. It is complete.

Delete Job

kubectl delete -f myJob.yaml

job "myjob" deleted

We just used kubectl get pods to monitor the progress of a job.

This example will use kubectl get job to monitor progress.

kubectl create -f myJob.yaml
job.batch/myjob created

Repeatedly run kubectl get job

kubectl get job
NAME    COMPLETIONS   DURATION   AGE
myjob   0/1           3s         3s

NAME    COMPLETIONS   DURATION   AGE
myjob   0/1           5s         5s

NAME    COMPLETIONS   DURATION   AGE
myjob   0/1           7s         7s

NAME    COMPLETIONS   DURATION   AGE
myjob   0/1           9s         9s

NAME    COMPLETIONS   DURATION   AGE
myjob   1/1           12s        12s

Only the last line shows : COMPLETIONS 1/1.

Frankly not information-loaded output at all. Later with more complex jobs it will reveal its value.

With experience you will learn when to use which :

  • kubectl get pods
  • kubectl get job

Demo complete, delete ...

kubectl delete -f myJob.yaml
job.batch "myjob" deleted

2) Job backoffLimit: 2

If your job has an unrecoverable error you may want to prevent it from continuously trying to run.

backoffLimit specify the number of retries before a Job is marked as failed. Default: 6

We set it to 2 in the example so that we can see it in action within seconds.

nano myJob.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'echo Job Pod is Running ; exit 1']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
      
  backoffLimit: 2

Note the exit 1 error exit code. This container will start up and exit immediately with an error condition.

Create the Job

kubectl create -f myJob.yaml

job/myjob created

Monitor job progress several times:

kubectl get pods

NAME          READY   STATUS              RESTARTS   AGE
myjob-drzhx   0/1     ContainerCreating   0          2s
myjob-hwdc5   0/1     Error               0          3s

NAME          READY   STATUS   RESTARTS   AGE
myjob-drzhx   0/1     Error    0          5s
myjob-hwdc5   0/1     Error    0          6s

NAME          READY   STATUS   RESTARTS   AGE
myjob-drzhx   0/1     Error    0          9s
myjob-hwdc5   0/1     Error    0          10s

NAME          READY   STATUS   RESTARTS   AGE
myjob-drzhx   0/1     Error    0          14s
myjob-gks4f   0/1     Error    0          4s
myjob-hwdc5   0/1     Error    0          15s

backoffLimit: 2 ... job stops creating containers when third error occurs.

Describe detail about our job: ( only relevant fields shown )

kubectl describe job/myjob
Name:           myjob
Parallelism:    1
Completions:    1
Start Time:     Thu, 24 Jan 2019 07:50:07 +0200
Pods Statuses:  0 Running / 0 Succeeded / 3 Failed
Pod Template:
  Containers:
   myjob:
    Command:
      sh
      -c
      echo Job Pod is Running ; exit 1
  
Events:
  Type     Reason                Age   From            Message
  ----     ------                ----  ----            -------
  Normal   SuccessfulCreate      35s   job-controller  Created pod: myjob-hwdc5
  Normal   SuccessfulCreate      34s   job-controller  Created pod: myjob-drzhx
  Normal   SuccessfulCreate      24s   job-controller  Created pod: myjob-gks4f
  Warning  BackoffLimitExceeded  4s    job-controller  Job has reached the specified backoff limit

Most informative lines are:

Pods Statuses:  0 Running / 0 Succeeded / 3 Failed

Warning  BackoffLimitExceeded  4s    job-controller  Job has reached the specified backoff limit

We have not discussed Parallelism: 1 and Completions: 1 yet. It simply means one Pod must run in parallel and one Pod must complete for job to be considered complete.

Determine a suitable backoffLimit for each of your production batch jobs.

Delete Job

kubectl delete -f myJob.yaml

job "myjob" deleted

3) Job Completions: 4

Job completions specify how many Pods must successfully complete for job to be considered complete.

Our job Pod below has work: sleep 3 seconds

4 Pods must complete ... completions: 4

nano myJob.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'echo Job Pod is Running ; sleep 3']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
      
  backoffLimit: 2
  completions: 4

Create the Job

kubectl create -f myJob.yaml

job/myjob created

This time follow progress using kubectl get jobs myjob repeatedly.

kubectl get jobs myjob
NAME    COMPLETIONS   DURATION   AGE
myjob   0/4           2s         2s

NAME    COMPLETIONS   DURATION   AGE
myjob   0/4           4s         4s

NAME    COMPLETIONS   DURATION   AGE
myjob   1/4           6s         6s

NAME    COMPLETIONS   DURATION   AGE
myjob   1/4           9s         9s

NAME    COMPLETIONS   DURATION   AGE
myjob   2/4           12s        12s

NAME    COMPLETIONS   DURATION   AGE
myjob   3/4           14s        14s

NAME    COMPLETIONS   DURATION   AGE
myjob   3/4           16s        16s

NAME    COMPLETIONS   DURATION   AGE
myjob   4/4           18s        19s

We can clearly see the 4 Pods each sleeping 3 seconds successfully and one-by-one, not all simultaneously .

Delete Job

kubectl delete -f myJob.yaml

job "myjob" deleted

Same job but another demo: this time we monitor progress using kubectl get po

Create the Job

kubectl create -f myJob.yaml

job/myjob created

Monitor progress:

kubectl get po
NAME          READY   STATUS              RESTARTS   AGE
myjob-74f5q   0/1     ContainerCreating   0          1s

NAME          READY   STATUS    RESTARTS   AGE
myjob-74f5q   1/1     Running   0          4s

NAME          READY   STATUS      RESTARTS   AGE
myjob-74f5q   0/1     Completed   0          7s
myjob-7gf7x   1/1     Running     0          2s

NAME          READY   STATUS              RESTARTS   AGE
myjob-74f5q   0/1     Completed           0          9s
myjob-7gf7x   0/1     Completed           0          4s
myjob-pfssq   0/1     ContainerCreating   0          0s

NAME          READY   STATUS              RESTARTS   AGE
myjob-74f5q   0/1     Completed           0          11s
myjob-7gf7x   0/1     Completed           0          6s
myjob-pfssq   0/1     ContainerCreating   0          2s

NAME          READY   STATUS      RESTARTS   AGE
myjob-74f5q   0/1     Completed   0          13s
myjob-7gf7x   0/1     Completed   0          8s
myjob-pfssq   1/1     Running     0          4s

NAME          READY   STATUS      RESTARTS   AGE
myjob-74f5q   0/1     Completed   0          17s
myjob-7gf7x   0/1     Completed   0          12s
myjob-pbmbt   1/1     Running     0          3s
myjob-pfssq   0/1     Completed   0          8s

NAME          READY   STATUS      RESTARTS   AGE
myjob-74f5q   0/1     Completed   0          19s
myjob-7gf7x   0/1     Completed   0          14s
myjob-pbmbt   0/1     Completed   0          5s
myjob-pfssq   0/1     Completed   0          10s

Note that only 1 Pod runs at a time. ( Default parallelism is one )

Delete Job

kubectl delete -f myJob.yaml

job "myjob" deleted

4) Job Parallelism: 2

I am running this on a 4 core server.

If you are running these exercises on an at-least 2 core server the following demo will work.

parallelism: 2 below specifies we want to run 2 Pods simultaneously.

nano myJob.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'echo Job Pod is Running ; sleep 3']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
      
  backoffLimit: 2
  completions: 4
  parallelism: 2

Create the Job

kubectl create -f myJob.yaml

job/myjob created

Monitor progress.

kubectl get pods
NAME          READY   STATUS              RESTARTS   AGE
myjob-11111   0/1     ContainerCreating   0          2s
myjob-22222   0/1     ContainerCreating   0          2s

NAME          READY   STATUS    RESTARTS   AGE
myjob-11111   1/1     Running   0          4s
myjob-22222   1/1     Running   0          4s

NAME          READY   STATUS      RESTARTS   AGE
myjob-33333   1/1     Running     0          1s
myjob-44444   1/1     Running     0          1s
myjob-11111   0/1     Completed   0          7s
myjob-22222   0/1     Completed   0          7s

NAME          READY   STATUS      RESTARTS   AGE
myjob-33333   1/1     Running     0          3s
myjob-44444   1/1     Running     0          3s
myjob-11111   0/1     Completed   0          9s
myjob-22222   0/1     Completed   0          9s

NAME          READY   STATUS      RESTARTS   AGE
myjob-33333   0/1     Completed   0          5s
myjob-44444   0/1     Completed   0          5s
myjob-11111   0/1     Completed   0          11s
myjob-22222   0/1     Completed   0          11s  

We needed completions: 4 Therefore 2 Pods were run in parallel twice to get the completions done.

Next we monitor progress using kubectl get jobs myjob

Delete Job

kubectl delete -f myJob.yaml

job "myjob" deleted

Same job created again.

kubectl create -f myJob.yaml
job.batch/myjob created

Monitor:

kubectl get jobs myjob
NAME    COMPLETIONS   DURATION   AGE
myjob   0/4           2s         2s

NAME    COMPLETIONS   DURATION   AGE
myjob   0/4           5s         5s

NAME    COMPLETIONS   DURATION   AGE
myjob   2/4           9s         9s

NAME    COMPLETIONS   DURATION   AGE
myjob   4/4           9s         11s

As expected COMPLETIONS get done in multiples of 2.

5) Job Parallelism: 4

If you are using at least 4 CPU server you can run this example :

We need 4 completions - all 4 run in parallel.

nano myJob.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'echo Job Pod is Running ; sleep 3']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
      
  backoffLimit: 2
  completions: 4
  parallelism: 4

Create the Job

kubectl create -f myJob.yaml

job/myjob created

Monitor:

kubectl get pods
NAME          READY   STATUS    RESTARTS   AGE
myjob-2j2fz   1/1     Running   0          2s
myjob-hkzmh   1/1     Running   0          2s
myjob-ngxps   1/1     Running   0          2s
myjob-skrbd   1/1     Running   0          2s

NAME          READY   STATUS    RESTARTS   AGE
myjob-2j2fz   1/1     Running   0          4s
myjob-hkzmh   1/1     Running   0          4s
myjob-ngxps   1/1     Running   0          4s
myjob-skrbd   1/1     Running   0          4s

NAME          READY   STATUS      RESTARTS   AGE
myjob-2j2fz   0/1     Completed   0          6s
myjob-hkzmh   0/1     Completed   0          6s
myjob-ngxps   0/1     Completed   0          6s
myjob-skrbd   0/1     Completed   0          6s

As expected 4 Pods running simultaneously.

Not shown - kubectl get jobs myjob ... what do you expect the output to look like?

Delete Job

kubectl delete -f myJob.yaml

job "myjob" deleted

6) Job activeDeadlineSeconds

activeDeadlineSeconds specifies total runtime for job as a whole.

From https://kubernetes.io/docs/concepts/workloads/controllers/jobs-run-to-completion/#job-termination-and-cleanup

Once a Job reaches activeDeadlineSeconds, all of its Pods are terminated and the Job status will become type: Failed with reason: DeadlineExceeded.

This exercise will demo this DeadlineExceeded

Note last line below: absurd low activeDeadlineSeconds: 3

nano myJob.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'echo Job Pod is Running ; sleep 4']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
      
  backoffLimit: 2
  completions: 4
  parallelism: 4
  activeDeadlineSeconds: 3

Create the Job

kubectl create -f myJob.yaml

job/myjob created

Repeatedly monitor progress:

kubectl get pods
NAME          READY   STATUS    RESTARTS   AGE
myjob-7xlch   1/1     Running   0          2s
myjob-8b9wl   1/1     Running   0          2s
myjob-qvpc7   1/1     Running   0          2s
myjob-rnhrg   1/1     Running   0          2s

kubectl get pods
No resources found.

Note after 3 seconds Pods no longer exist. We will see why below.

Check job status:

kubectl get jobs myjob

NAME    COMPLETIONS   DURATION   AGE
myjob   0/4           20s        20s

Disappointingly NO indication that we have a problem job.

Describe details about job:

kubectl describe job/myjob
Name:                     myjob
Namespace:                default
Parallelism:              4
Completions:              4
Start Time:               Thu, 24 Jan 2019 08:25:53 +0200
Active Deadline Seconds:  3s
Pods Statuses:            0 Running / 0 Succeeded / 4 Failed
Events:
  Type     Reason            Age   From            Message
  ----     ------            ----  ----            -------
  Normal   SuccessfulCreate  8s    job-controller  Created pod: myjob-5rcwp
  Normal   SuccessfulCreate  8s    job-controller  Created pod: myjob-nbdz4
  Normal   SuccessfulCreate  8s    job-controller  Created pod: myjob-4jb28
  Normal   SuccessfulCreate  7s    job-controller  Created pod: myjob-vb765
  
  Normal   SuccessfulDelete  4s    job-controller  Deleted pod: myjob-vb765
  Normal   SuccessfulDelete  4s    job-controller  Deleted pod: myjob-4jb28
  Normal   SuccessfulDelete  4s    job-controller  Deleted pod: myjob-5rcwp
  Normal   SuccessfulDelete  4s    job-controller  Deleted pod: myjob-nbdz4
  
  Warning  DeadlineExceeded  4s    job-controller  Job was active longer than specified deadline

Much more helpful: 0 Succeeded / 4 Failed

Also informative: Warning - - DeadlineExceeded - - Job was active longer than specified deadline

Now we see why our Pods disappeared ... they were deleted.

I do not understand the logic:

  • when Pods complete successfully they continue to exist so you can inspect the SUCCESS job logs
  • when Pods complete UNsuccessfully they get deleted so you CANNOT inspect the FAILED job logs

Let's attempt to investigate the failed logs:

kubectl logs myjob-5rcwp
Error from server (NotFound): pods "myjob-5rcwp" not found

NOTE: use activeDeadlineSeconds only when you have successfully resolved this missing logs issue.

Delete Job

kubectl delete -f myJob.yaml

job "myjob" deleted

7) Parallel Jobs with a Work Queue - Setup

Kubernetes documentation contains 2 complex examples of parallel jobs

More than 80% of those guides focus on setting up the work queue functionality.

This tutorial focus on learning about parallel jobs and queues using my very simple bash implementation.

We need a directory for our bash work queue functionality:

mkdir /usr/share/jobdemo

Our work queue:

nano /usr/share/jobdemo/workqueue
1
2
3
4
5
6
7
8
9
10
11
12

We will need this backup later ( Running Pods will delete lines from our workqueue file )

cp /usr/share/jobdemo/workqueue /usr/share/jobdemo/workqueue-backup

Work queue processing script:

nano /usr/share/jobdemo/jobscript.sh

#!/bin/bash
for counter in 1 2 3 4 5 6 7 8 9 10 11 12
do
if [ -s /usr/share/jobdemo/workqueue ]
then 
   echo " did some work "
   sed -i '1d' /usr/share/jobdemo/workqueue
   sleep 1
else
   echo " no more work left "
   exit 0
fi
done
exit 0

Program explanation ( focus on getting minimal work queue just barely working ) :

  • for loop loops 12 times
  • if -s /usr/share/jobdemo/jobqueue ... if workqueue not empty
  • echo " did some work " if there are lines left in workqueue
  • remove first line from workqueue file ; sleep 1 second
  • if no more lines in workqueue file then exit with return code 0 = success

Basically every Pod will delete whatever first line it finds from workqueue file and echo that they did some work.

When a Pod finds workqueue empty it just exits with 0 code which means success.

Seeing this in action several times will make it more clear.

We need to place the workqueue file and the jobscript on a persistent volume. Now all Pods will use the same workqueue file. Every Pod will take work from the SAME queue and delete the line ( work ) it took from the queue.

nano myPersistent-Volume.yaml

kind: PersistentVolume
apiVersion: v1
metadata:
  name: my-persistent-volume
  labels:
    type: local
spec:
  storageClassName: pv-demo 
  capacity:
    storage: 10Mi
  accessModes:
    - ReadWriteOnce
  hostPath:
    path: "/usr/share/jobdemo"    

Create a 10Mi Persistent Volume pointing to the location / path of our 2 workqueue objects.

Reference : https://kubernetes.io/docs/concepts/storage/volumes/

nano myPersistent-VolumeClaim.yaml

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: my-persistent-volumeclaim
spec:
  storageClassName: pv-demo 
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 10Mi

Claim usage of storageClassName: pv-demo -> pointing to the Persistent Volume.

Reference : https://kubernetes.io/docs/concepts/storage/persistent-volumes/

kubectl create -f myPersistent-Volume.yaml

kubectl create -f myPersistent-VolumeClaim.yaml

8) Parallel Jobs with a Work Queue - Simplest Example

This example uses one Pod to read and process a workqueue until it is empty.

Create myWorkqueue-Job.yaml

nano myWorkqueue-Job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'source /usr/share/jobdemo/jobscript.sh']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
  
    volumeMounts:
      - mountPath: "/usr/share/jobdemo"
        name: my-persistent-volumeclaim-name

  volumes:
    - name: my-persistent-volumeclaim-name
      persistentVolumeClaim:
       claimName: my-persistent-volumeclaim  
       
  restartPolicy: Never
  terminationGracePeriodSeconds: 0
  parallelism: 1

This spec mounts our persistent volume and the command runs our jobscript.sh

Create the Job.


kubectl create -f myWorkqueue-Job.yaml
job.batch/myjob created

Repeatedly run kubectl get pods ... monitor progress for parallelism: 1

kubectl get pods
NAME          READY   STATUS    RESTARTS   AGE
myjob-x9ghz   1/1     Running   0          3s

NAME          READY   STATUS    RESTARTS   AGE
myjob-x9ghz   1/1     Running   0          6s

NAME          READY   STATUS    RESTARTS   AGE
myjob-x9ghz   1/1     Running   0          8s

NAME          READY   STATUS    RESTARTS   AGE
myjob-x9ghz   1/1     Running   0          10s

NAME          READY   STATUS    RESTARTS   AGE
myjob-x9ghz   1/1     Running   0          12s

NAME          READY   STATUS      RESTARTS   AGE
myjob-x9ghz   0/1     Completed   0          15s

Our single Pod took 12 seconds to delete the workqueue lines 1 by 1.

We can see this in the Pod log.

kubectl logs pod/myjob-x9ghz
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work 

Delete job.


kubectl delete -f myWorkqueue-Job.yaml
job.batch "myjob" deleted

9) Parallel Jobs with a Work Queue : Parallelism: 2

Note last line in spec: we are now going to run 2 Pods in parallel.

nano myWorkqueue-Job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'source /usr/share/jobdemo/jobscript.sh']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
  
    volumeMounts:
      - mountPath: "/usr/share/jobdemo"
        name: my-persistent-volumeclaim-name

  volumes:
    - name: my-persistent-volumeclaim-name
      persistentVolumeClaim:
       claimName: my-persistent-volumeclaim  
       
  restartPolicy: Never
  terminationGracePeriodSeconds: 0
  parallelism: 2

We need to put the deleted lines back into workqueue file.

cp /usr/share/jobdemo/workqueue-backup /usr/share/jobdemo/workqueue
kubectl create -f myWorkqueue-Job.yaml
job.batch/myjob created

Repeatedly run kubectl get pods ... monitor parallelism: 2

kubectl get pods
NAME          READY   STATUS              RESTARTS   AGE
myjob-9nkj5   0/1     ContainerCreating   0          2s
myjob-gdrj8   0/1     ContainerCreating   0          2s

NAME          READY   STATUS    RESTARTS   AGE
myjob-9nkj5   1/1     Running   0          4s
myjob-gdrj8   1/1     Running   0          4s

NAME          READY   STATUS    RESTARTS   AGE
myjob-9nkj5   1/1     Running   0          6s
myjob-gdrj8   1/1     Running   0          6s

NAME          READY   STATUS    RESTARTS   AGE
myjob-9nkj5   1/1     Running   0          8s
myjob-gdrj8   1/1     Running   0          8s

NAME          READY   STATUS      RESTARTS   AGE
myjob-9nkj5   0/1     Completed   0          11s
myjob-gdrj8   0/1     Completed   0          11s

As expected 2 Pods were running parallel all the time.

If we investigate the logs of both our Pods we can see that each did around half the work.

Both Pods then exited with exit 0 when they found 'no more work left' ( workqueue empty ).

kubectl logs pod/myjob-9nkj5
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 no more work left

kubectl logs pod/myjob-gdrj8
 did some work
 did some work
 did some work
 did some work
 did some work
 did some work
 no more work left

Get job overview.

kubectl get jobs

NAME    COMPLETIONS   DURATION   AGE
myjob   2/1 of 2      9s         98s

We can see 2 Pods simultaneously took 9s versus 12s for just one Pod.

Delete job.

kubectl delete -f myWorkqueue-Job.yaml
job.batch "myjob" deleted

10) Parallel Jobs with a Work Queue : Parallelism: 4

Note last line in spec: we are now going to run 4 Pods in parallel.

nano myWorkqueue-Job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'source /usr/share/jobdemo/jobscript.sh']
    
      restartPolicy: Never
      terminationGracePeriodSeconds: 0
  
    volumeMounts:
      - mountPath: "/usr/share/jobdemo"
        name: my-persistent-volumeclaim-name

  volumes:
    - name: my-persistent-volumeclaim-name
      persistentVolumeClaim:
       claimName: my-persistent-volumeclaim  
       
  restartPolicy: Never
  terminationGracePeriodSeconds: 0
  parallelism: 4

We need to put the deleted lines back into workqueue file.

cp /usr/share/jobdemo/workqueue-backup /usr/share/jobdemo/workqueue

Create the job.

kubectl create -f myWorkqueue-Job.yaml
job.batch/myjob created

Monitor:

kubectl get pods

NAME          READY   STATUS              RESTARTS   AGE
myjob-8l6bc   0/1     ContainerCreating   0          1s
myjob-krmcc   0/1     ContainerCreating   0          1s
myjob-lxd9w   0/1     ContainerCreating   0          1s
myjob-ntgf7   0/1     ContainerCreating   0          1s

NAME          READY   STATUS    RESTARTS   AGE
myjob-8l6bc   1/1     Running   0          4s
myjob-krmcc   1/1     Running   0          4s
myjob-lxd9w   1/1     Running   0          4s
myjob-ntgf7   1/1     Running   0          4s

NAME          READY   STATUS      RESTARTS   AGE
myjob-8l6bc   0/1     Completed   0          6s
myjob-krmcc   0/1     Completed   0          6s
myjob-lxd9w   1/1     Running     0          6s
myjob-ntgf7   0/1     Completed   0          6s

NAME          READY   STATUS      RESTARTS   AGE
myjob-8l6bc   0/1     Completed   0          8s
myjob-krmcc   0/1     Completed   0          8s
myjob-lxd9w   0/1     Completed   0          8s
myjob-ntgf7   0/1     Completed   0          8s

4 Pods starting up simultaneously.

4 Pods running parallel as expected.

kubectl get jobs

NAME    COMPLETIONS   DURATION   AGE
myjob   4/1 of 4      6s         18s

4 parallel Pods faster than 2. ( Overhead of ContainerCreating prevents it from being twice as fast ).

Describe detail about our job: ( only relevant fields shown )

kubectl describe job/myjob

Name:           myjob
Parallelism:    4
Start Time:     Thu, 24 Jan 2019 12:58:11 +0200
Completed At:   Thu, 24 Jan 2019 12:58:17 +0200
Duration:       6s
Pods Statuses:  0 Running / 4 Succeeded / 0 Failed
Events:
  Type    Reason            Age    From            Message
  ----    ------            ----   ----            -------
  Normal  SuccessfulCreate  2m17s  job-controller  Created pod: myjob-8l6bc
  Normal  SuccessfulCreate  2m17s  job-controller  Created pod: myjob-lxd9w
  Normal  SuccessfulCreate  2m17s  job-controller  Created pod: myjob-krmcc
  Normal  SuccessfulCreate  2m17s  job-controller  Created pod: myjob-ntgf7

4 Succeeded / 0 Failed and only success lines in events at the bottom.

This is example of a perfectly done job.

These 2 outputs below mean the same thing ( success ... note all the 4s for our parallelism: 4 job )

kubectl get jobs

NAME    COMPLETIONS   DURATION   AGE
myjob   4/1 of 4      6s         18s

4 lines of 4 Completed Pods with zero RESTARTS.

kubectl get pods

NAME          READY   STATUS      RESTARTS   AGE
myjob-8l6bc   0/1     Completed   0          8s
myjob-krmcc   0/1     Completed   0          8s
myjob-lxd9w   0/1     Completed   0          8s
myjob-ntgf7   0/1     Completed   0          8s

Delete job.

kubectl delete -f myWorkqueue-Job.yaml
job.batch "myjob" deleted

11) Parallel Jobs with a Work Queue : Parallelism: 8

Note last line in spec: we are now going to run 8 Pods in parallel. ( On a node with only 4 CPU cores )

nano myWorkqueue-Job.yaml

apiVersion: batch/v1
kind: Job
metadata:
  name: myjob
spec:
  template:
    spec:      
      containers:
      - name: myjob
        image: alpine
        imagePullPolicy: IfNotPresent
        
        command: ['sh', '-c', 'source /usr/share/jobdemo/jobscript.sh']

        volumeMounts:
          - mountPath: "/usr/share/jobdemo"
            name: my-persistent-volumeclaim-name

      restartPolicy: Never
      terminationGracePeriodSeconds: 0
  
      volumes:
        - name: my-persistent-volumeclaim-name
          persistentVolumeClaim:
            claimName: my-persistent-volumeclaim  
            
  terminationGracePeriodSeconds: 0
  parallelism: 8

We need to put the deleted lines back into workqueue file.

cp /usr/share/jobdemo/workqueue-backup /usr/share/jobdemo/workqueue

Create job.

kubectl create -f myWorkqueue-Job.yaml
job.batch/myjob created

Monitor:

kubectl get pods
NAME          READY   STATUS              RESTARTS   AGE
myjob-88g75   0/1     ContainerCreating   0          2s
myjob-8wj9w   0/1     ContainerCreating   0          2s
myjob-br9bb   0/1     ContainerCreating   0          2s
myjob-pfnth   0/1     Pending             0          2s
myjob-r9p46   0/1     ContainerCreating   0          2s
myjob-scjwx   0/1     ContainerCreating   0          2s
myjob-zm92q   0/1     ContainerCreating   0          2s
myjob-zt8k6   0/1     ContainerCreating   0          2s

NAME          READY   STATUS              RESTARTS   AGE
myjob-88g75   0/1     ContainerCreating   0          6s
myjob-8wj9w   0/1     Completed           0          6s
myjob-br9bb   0/1     ContainerCreating   0          6s
myjob-pfnth   0/1     ContainerCreating   0          6s
myjob-r9p46   0/1     ContainerCreating   0          6s
myjob-scjwx   0/1     Completed           0          6s
myjob-zm92q   0/1     Completed           0          6s
myjob-zt8k6   0/1     Completed           0          6s

NAME          READY   STATUS      RESTARTS   AGE
myjob-88g75   0/1     Completed   0          10s
myjob-8wj9w   0/1     Completed   0          10s
myjob-br9bb   0/1     Completed   0          10s
myjob-pfnth   0/1     Completed   0          10s
myjob-r9p46   0/1     Completed   0          10s
myjob-scjwx   0/1     Completed   0          10s
myjob-zm92q   0/1     Completed   0          10s
myjob-zt8k6   0/1     Completed   0          10s

Determine total job runtime.

kubectl get jobs

NAME    COMPLETIONS   DURATION   AGE
myjob   8/1 of 8      8s         19s

Using 4 Pods took 9 seconds and using 8 Pods took 8 seconds.

It does not make sense to run more CPU-intensive workload Pods in parallel than CPU cores on server. Jobs will context switch too much.

https://en.wikipedia.org/wiki/Context_switch

Notice how we used a very simple basic bash script to emulate a work queue.

Even that VERY simple script enabled us to learn a great deal about parallel jobs processing ONE shared, SIMULATED work queue.

Cleanup - delete job.

kubectl delete -f myWorkqueue-Job.yaml
job.batch "myjob" deleted

12) Different Patterns for Parallel Computation

Some different ways to manage parallel jobs are discussed at https://kubernetes.io/docs/concepts/workloads/controllers/jobs-run-to-completion/#job-patterns

1 0 0
Share on

Alibaba Clouder

1,904 posts | 408 followers

You may also like

Comments

5004576456507455 December 16, 2019 at 12:39 am

All the work queue items not working,Getting volume mount errorParallel Jobs with a Work Queue - Simplest ExampleParallel Jobs with a Work Queue : Parallelism: 8

Alibaba Clouder

1,904 posts | 408 followers

Related Products

  • Container Registry

    A secure image hosting platform providing containerized image lifecycle management

    Learn More
  • Container Service

    A high-performance container manage service that provides containerized application lifecycle management

    Learn More
  • Container Service for Kubernetes

    Alibaba Cloud Container Service for Kubernetes is a fully managed cloud container management service that supports native Kubernetes and integrates with other Alibaba Cloud products.

    Learn More
  • ECI(Elastic Container Instance)

    Elastic Container Instance (ECI) is an agile and secure serverless container instance service. You can easily run containers without managing servers. Also you only pay for the resources that have been consumed by the containers. ECI helps you focus on your business applications instead of managing infrastructure.

    Learn More