Multiple Retry Strategies for a Single Pod?

[xiki-tempula]

Hi Argo Community,

I'm wondering if it's possible to configure multiple distinct retry strategies for a single pod within an Argo Workflow.

We have a couple of different failure scenarios we'd like to handle with specific retry behaviors:

Network-related errors: For transient network issues, we'd like to retry the pod up to a maximum of 10 attempts. If it still fails after these 10 retries, the step should be marked as failed.
Pod preemption (e.g., "pod deleted" error): In cases where a pod is preempted (often seen with spot instances or cluster autoscaling), we want to retry the pod indefinitely, perhaps once every hour, without a maximum attempt limit, until it succeeds.
Is it possible to define such conditional or multiple retry strategies for one pod, or is a pod limited to a single retry strategy configuration?

Thanks in advance for any insights or suggestions!

[tczhao]

No, we don't have any feature like this at the moment.
Perhaps we can go to a design proposal and introduce retryStrategies