Hey. I am trying to run flytectl on my Macbook Air M3. This is an apple silicon architecture (aka ARM64). I have a docker image that I have built which is built for platform=linux/amd64 that I want to use for my workflow/tasks. I have specified this in the the

pyflyte register

command using the

--image

tag, but I am not having any luck forcing the flytectl cluster to use the AMD64 architecture. I started the demo specifically with this env variable

flytectl demo start --env DOCKER_DEFAULT_PLATFORM=linux/amd64

(which theoretically should force docker to try to pull amd64 images?). I am getting errors in my workflow that it can't find the container

[1/1] currentAttempt done. Last Error: USER::Grace period [3m0s] exceeded|containers with unready status: [afqwjjtbs6n4ltk4pd9l-n0-0]|Back-off pulling image "localhost:30000/my-image:latest"

Below is what I used to build/push/install the workflow:

docker build -t my-image:latest --platform=linux/amd64 .
docker tag detekt-challenge:latest localhost:30000/my-image:latest
docker push localhost:30000/my-image:latest
pyflyte register --project my-project --image localhost:30000/my-image:latest src

At this point I am not sure if this is an architecture issue or something else not working. I did verify I can pull the image back to my local machine (so the docker registry is running). How can I test this locally with

flytectl

? Thanks 🙂

DOCKER_DEFAULT_PLATFORM=linux/amd64

I think this is for emulation. To build locally on M1 I have used ImageSpec and the

platform="linux/arm64"

argument

I could have sworn I tried to use this, and it still didn’t work. I’ll give it another attempt in a bit

@glamorous-pizza-44029 any change u manage to fix this issue somehow?

Hey @fierce-monitor-77717 No, I never got this to work. I pointed out above where the issue is, but was not confident enough in how the Flyte repository works to submit a PR for it. Theoretically, there just needs to be the extra params passed to the docker API when trying to pull the images. My "workaround" was to actually just make my Dockerfile for my Flyte task actually work on ARM and not only AMD64 (which is what I was originally trying to do). Sorry I can't be of more help

Hiya 👋 , having a similar issue with this too, on an M4 Pro MBP and it breaks the Getting Started guide for me. Looking at the Flyte Demo container logs:

E0509 16:11:11.973535      75 remote_image.go:180] "PullImage from image service failed" err="rpc error: code = NotFound desc = failed to pull and unpack image \"localhost:30000/say-hello-image:Y0RqVPnX2_221ovInaIIqw\": no match for platform in manifest: not found" image="localhost:30000/say-hello-image:Y0RqVPnX2_221ovInaIIqw"

But I can run the image (via Docker Desktop with Rosetta enabled), albeit with a warning:

$ docker run --rm localhost:30000/say-hello-image:Y0RqVPnX2_221ovInaIIqw echo 'hello'
WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested
hello

Slightly modified `hello_world.py`:

# Hello World

import flytekit as fl

image_spec = fl.ImageSpec(
    # The name of the image. This image will be used by the `say_hello`` task.
    name="say-hello-image",
    # Lock file with dependencies to be installed in the image.
    requirements="uv.lock",
    # Image registry to to which this image will be pushed.
    # Set the Environment variable FLYTE_IMAGE_REGISTRY to the URL of your registry.
    # The image will be built on your local machine, so enure that your Docker is running.
    # Ensure that pushed image is accessible to your Flyte cluster, so that it can pull the image
    # when it spins up the task container.
    registry="localhost:30000",
)


@fl.task(container_image=image_spec)
def say_hello(name: str) -> str:
    return f"Hello, {name}!"


@fl.workflow
def hello_world_wf(name: str = "world") -> str:
    greeting = say_hello(name=name)
    return greeting

Adding

platform="linux/arm64"

to the

ImageSpec

works 🎉 (thank you for that), it might be useful to add that to the docs. It'd also be useful to point out the bundled image registry on

localhost:30000

in the getting started guide as well, as that's great for local testing but not advertised in the guide.