I have many one-shot dockers to generate an output file in my project, they are small but cost many resources, so k8s is the best choice. In this scenario, I start pods by kubectl run --restart=Never without any replication controllers. As a result, the output file is not guaranteed, and I have to handle them by myself including:

• remove succeeded pods.
• restart failed pods.

To remove succeeded pods, there seems no "--rm" argument in k8s as in the docker, nor the methods to pass into docker command, so kubectl delete pods --all should be the only choice. What's worse, all the pods status are Error or ExitCode:0, but fortunately, the output file generated successfully.

To restart failed pods, I really have no idea on that right now.

So, does it look silly, or is there any better solutions?

Thanks in advance!