[Feature]: Kube Play: add support for Job

[ygalblum]

Feature request description

Currently, podman supports pods and deployments. While pod is the base element, K8S users will usually not deploy pods. Instead, they will deploy deployments for long lasting workloads and jobs for single shot operations.
When using systemd, the behavior of jobs seems to be a better fit as it will not have Podman monitor the pods but instead pass the monitoring and hence the handling of the restart policy to systemd.
For example, the supported valued for restartPolicy for containers in a Job are only Never or OnFailure. So, a default restartPolicy of Never, with parallelism and completions set to 1 will result in propagating the exit status of the pod to systemd, letting it decide whether or not the service should be restarted (depending on the restart policy of the systemd unit)

Suggest potential solution

Passing a job to podman kube play will result in running the containers listed in the Pod specification

Have you considered any alternatives?

As explained, the user may use deployments or pods directly. But, it will not provide the same experience

Additional context

While the trigger for this feature request is Quadlet and systemd, podman will directly benefit from this added support as well

[rhatdan]

Makes sense to me.

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[ygalblum]

/stale remove

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[rhatdan]

@umohnani8 PTAL

[fsdrw08]

any updates?

[umohnani8]

A design doc has been created for this and it is on the to do list, targeting this for 4.7.

[mrgleeco]

link to design doc / discussions?

[adelton]

@umohnani8 Could you please show the link to the design doc?

[fsdrw08]

any updates?

[umohnani8]

Yup, I have a WIP in PR - will clean it up and open a PR by next week.

[jaimehrubiks]

This makes a lot of sense. I believe podman is getting a lot of attention in HPC, I myself find it very flexible and simple for rootless use cases, where developers test their own containers, and in many cases they then send them to kubernetes. For HPC, jobs are the most common kubernetes yaml.

[rhatdan]

@umohnani8 Still have the PR? Maybe someone else can take it over?