Hi All, Thank you for being that generous with your time and work. I’m migrating from prefect 2 to prefect 3 and I need to switch to the new deployment mechanism. Unfortunately it doesn’t work as I expect. The old deployment was:

Hi
Deployment.build_from_flow(
    name="...",
    flow=my_flow,
    parameters={..},
    storage=RemoteFileSystem.load('minio'),
    path='...',
    version=os.getenv("GIT_COMMIT_SHA"),
    infrastructure=DockerContainer(
        auto_remove=True,
        image='prefect-default-worker:2.20.9',
        image_pull_policy='NEVER',
        network_mode='bridge',
    ),
).apply()

Where ‘minio’ is a Remote File System block containing Minio connection data, prefect-default-worker:2.20.9 is an image based on prefecthq/prefect:2.20.9-python3.10 with all the flow’s dependencies installed. I have another container prefect-agent:2.20.9 with

entrypoint: ["prefect", "agent", "start", "-q", "default"]

Reading the migration guide I get that I need to 1) Replace ‘infrastructure’ with ‘work_pool_name’ and 2) Replace ‘storage’ with a call to ‘flow.from_source’ 3) replace

entrypoint: ["prefect", "agent", "start", "-q", "default"]

with

entrypoint: ["prefect", "worker", "start", "-p", "default"]

The new deployment script looks like this:

my_flow.from_source(
    source=RemoteFileSystem.load("minio"),
    entrypoint='path to the file with @flow definition'
).deploy(
    name="same",
    work_pool_name='default',
    parameters={same},
    version=os.getenv("GIT_COMMIT_SHA"),
    build=False,
    push=False,
    job_variables=dict(pull_policy="NEVER", network_mode="bridge", auto_remove=True)
)

Before the migration, the script uploads the flow’s python file (and the whole folder) to minio and then create/update the deployment. Now it doesn’t seem to do the upload to Minio. The script throws an error: FileNotFoundError: [Errno 2] No such file or directory: ‘/var/folders/l9/q51p6spx7sb6j3gqdq5q7jhr0000gn/T/tmpi8gx9s0t/remote-file-system-minio/…’ If I manually put the files in Minio manually It seems to load the flow. I can see that a ScriptError is raised loading the flow’s python file. I’ll share the error after I clarify this issue. So I wonder if I can keep the previous behaviour, or I should upload the flow’s definition to Minio manually.

Hi Tsvetelin, thanks for your message. It's possible to push code to remote storage using the push step of the prefect.yaml (an alternative way to deploy your flow which uses the CLI). You're correct that the .deploy() method does not push your code to remote storage on your behalf. You'd need to write some code that pushes your flow to minio in your deployment script.

Thank you, Bianca. This one does the job

s3_bucket = RemoteFileSystem.load("minio")
s3_bucket.put_directory(to_path='test_flow', ignore_file='.prefectignore')