Flyte enables production-grade orchestration for machine learning workflows and data processing created to accelerate local workflows to production.

Flyte

Good morning<@UNR3C6Y4T> <@UNZB4NW3S>. We recently updated to flytekit 1.3.2 and flyte 1.3.0. I’m seeing an issue where a task’s specific memory requests/limits and custom image is not being used. This is happening for a task where we set `task_config=Pod()` as we needed to customize some of the k8s spec. Usually for simpler tasks we set `requests` and `limits` for the `task` decorator
It seems like the way we should be setting these values may have changed based on <https://github.com/flyteorg/flyte/issues/2703|this>? I’m reviewing <https://docs.flyte.org/en/latest/deployment/configuration/general.html#configuring-custom-k8s-resources|this documentation now>

Hey <@U026D9YKXTL> so sorry about this. Using the configuration you mentioned - ie. existing `PodTask` should still be fully supported. We implemented the new `PodTemplate` to allow finer-grained configuration accross task types - for example `ContainerTask`. This is supposed to be fully backwards compatible, if it's not, it is a problem. Can you provide an example that reproduces this issue?

Right now the task definition is very specific to our environment. Let me try and reproduce on a separate example

False alarm it was operator error on our part!

:sweat_smile: that's relieving! Hopefully debugging wasn't too difficult!

And if you want to chat more about the PodTemplate stuff, would be happy to. We're working on some more extensions <https://github.com/flyteorg/flyte/issues/3123|here>. Basically, you can define a `PodTemplate` in the task decorator which can also support `ContainerTask` in addition to different plugins (ex. Dask, Ray, etc). Ultimately, this will achieve the same thing as the `task_config=Pod` (ie. `PodTask`) that you're currently using, but will apply to more task types offering more configurability.

<@U029U35LRDJ> I am trying to apply k8s labels using `pod_template`  for `ContainerTask`  but the labels doesn't reflect. I have tried `flytekit==1.4.2`  and task def looks like below, Can you suggest if I am missing something here :
```calculate_ellipse_area_shell = ContainerTask(
    name="ellipse-area-metadata-shell",
    input_data_dir="/var/inputs",
    output_data_dir="/var/outputs",
    inputs=kwtypes(a=float, b=float),
    outputs=kwtypes(area=float, metadata=str),
    image="<http://ghcr.io/flyteorg/rawcontainers-shell:v2|ghcr.io/flyteorg/rawcontainers-shell:v2>",
    command=[
        "./calculate-ellipse-area.sh",
        "{{.inputs.a}}",
        "{{.inputs.b}}",
        "/var/outputs",
    ],
    pod_template=PodTemplate(labels=labels),
)```

<@U04MG1WLA4T> what version of flytepropeller are you running? <https://github.com/flyteorg/flyteplugins/pull/325|This PR> added support for this functionality into flyteplugins, which made it's way into <https://github.com/flyteorg/flytepropeller/releases/tag/v1.1.71|flytepropeller in the v1.1.71 release>. This flytepropller version seems to be included in <https://github.com/flyteorg/flyte/pull/3563|Flyte v1.5>.