Hi everyone wave In Prefect1 we were able to pass a Kubernet

Question

Hi everyone wave In Prefect1 we were able to pass a KubernetesRun while creating a flow with Python client For example we can provide a custom template via KubernetesRun to specify resources add tolerations if we want to schedule the flow on a gpu node etc How can we achieve same in Prefect2 Thanks

Deceivious · Answer

<https docs prefect io api ref prefect infrastructure prefect infrastructure KubernetesJob> Unsure if this is the solution you are looking for but the customization parameters might be the answer

Christopher Boyd · Answer

You can define it using a job template and building from yaml

Christopher Boyd · Answer

Here s an older thread for someone asking something similar I think <https prefect community slack com archives CL09KU1K7 p1673616433349529>

Christopher Boyd · Answer

and should help what that might look like for 2 x

Tolga Karahan · Answer

So you suggest creating a new infra block from a yaml each time configuration is changed It is frequently the case for us because users specifying some configurations from the UI and run flows

Christopher Boyd · Answer

You can create a baseline and pass in customizations

Christopher Boyd · Answer

If changes are being made though you would need to modify the deployment customizations for those users requesting changes

Christopher Boyd · Answer

e g if user a wants 5 cpus and user b wants 10 cpus those would be 2 separate deployments

Tolga Karahan · Answer

but they want to run same flow with different resources in this sense it doesn t make sense to me to create multiple deployments for same flow logic due to config changes

Tolga Karahan · Answer

maybe we can overwrite an infra block with each new request and use it with same flow deployment

Tolga Karahan · Answer

I played with the tutorial you sent It seems we are able to override already applied deployment with new configuration so I can create a temporary block out of base configuration by putting user specified values inside and apply the deployment with new infra block