Flow run infrastructure exited with non-zero status code 255

[hongbo-miao]

I deployed Prefect successfully. Here are my deployment codes:

• Prefect Server
• Prefect Worker

Now I am trying to run a Prefect workflow. Here are my codes:

main.py

import platform
from prefect import flow, get_run_logger

@flow
def print_platform() -> None:
    logger = get_run_logger()
    logger.info(f"Node: {platform.node()}")
    logger.info(f"Platform: {platform.platform()}")
    logger.info(f"Python Version: {platform.python_version()}")


if __name__ == "__main__":
    print_platform()

deploy.py

import asyncio
from main import print_platform

async def deploy() -> None:
    await print_platform.deploy(
        name="hm-development-print-platform",
        work_pool_name="hm-work-pool",
        image="ghcr.io/hongbo-miao/hm-prefect-print-platform:latest",
        push=True,
    )

if __name__ == "__main__":
    asyncio.run(deploy())

pyproject.toml

[project]
name = "prefect-print-platform"
version = "1.0.0"
requires-python = "~=3.11.0"
dependencies = [
  "prefect-client==3.2.7",
]

[dependency-groups]
dev = [
  "poethepoet==0.32.2",
  "pytest-cov==6.0.0",
  "pytest==8.3.4",
]

[tool.uv]
package = false

[tool.poe.tasks]
deploy = "python src/deploy.py"

When I run uv run poe deploy, I can see

Poe => python src/deploy.py
Successfully built image 
'ghcr.io/hongbo-miao/hm-prefect-print-platform:latest'
Successfully pushed image 
'ghcr.io/hongbo-miao/hm-prefect-print-platform:latest'
Successfully created/updated all deployments!

                            Deployments                             
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━┳━━━━━━━━━┓
┃ Name                                         ┃ Status  ┃ Details ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━╇━━━━━━━━━┩
│ print-platform/hm-development-print-platform │ applied │         │
└──────────────────────────────────────────────┴─────────┴─────────┘

To schedule a run for this deployment, use the following command:

        $ prefect deployment run 
'print-platform/hm-development-print-platform'


You can also run your flow via the Prefect UI: https://prefect.hongbomiao.com/deployments/deployment/6790e4ac-d9fa-4c0c-8724-9de10660dc18

However, the job failed with

Flow run infrastructure exited with non-zero status code 255.

Any help would be appreciate, thanks! ☺️

[desertaxle]

I don't think I've ever seen a 255 exit code before! Can you share any logs from the pod your flow was running in? I'm hoping it will have a clue as to what went wrong.

[hongbo-miao]

Hi @desertaxle the pod failed immediately, I got error

exec /usr/bin/tini: exec format error

Ah it is because of the "exec format error" typically occurs when trying to run a binary that was compiled for a different CPU architecture than the one you're running on. This commonly happens with Docker containers when there is an architecture mismatch.

Because

    await print_platform.deploy(
        name="hm-development-print-platform",
        work_pool_name="hm-work-pool",
        image="ghcr.io/hongbo-miao/hm-prefect-print-platform:latest",
        push=True,
    )

built the image on my macOS which is arm64 while my Kubernetes cluster is using amd64. I feel push=True is kind of handy when doing development.

Hmm in this case, actually any other better suggestions for testing the workflow?

Also, does it make sense to support something using QEMU, so that we can build image for different platforms?

    await print_platform.deploy(
        name="hm-development-print-platform",
        work_pool_name="hm-work-pool",
        image="ghcr.io/hongbo-miao/hm-prefect-print-platform:latest",
        platforms=["amd64", "arm64"]  # <--
        push=True,
    )