Interesting question, I have the following flow state function (simplified for example)

def flow_updater(flow, current_state, next_state):
    if isinstance(next_state, Failed):
        error_information = {k.name: v.message for k, v in next_state.result.items() if type(v) is Failed}
    else:
         error_information = None
    report(error_information)

When I run this flow locally using

prefect run -p path/to/flow.py

with a simulated error, I get the expected error information, however when I run it on our deployment (not cloud, backend in AWS). I get error_information equal to

{}

(i.e. no task results have a result of type

Failed

). Is there some other type that gets used when running connected to server? (I am using

type(v) is Failed

instead of

isinstance

cause I don want to report

TriggerFailed

, etc)

hi people, i have a problem after update to prefect 2.6.0 and my script stop working the message of error says .fn() for call my function outside the flow

after adding call my functions the following error presented when execute File "/home/cpfl0dev/.local/lib/python3.8/site-packages/prefect/logging/loggers.py", line 110, in get_run_logger raise MissingContextError("There is no active flow or task run context.") prefect.exceptions.MissingContextError: There is no active flow or task run context.

in version 2.4.2 working

Has anyone ever tried running prefect flows through GCP lambda functions? I’m exploring how to support a mapped run of 1000s of tasks in parallel, and am not sure whether to go the Google Kubernetes Engine route (which Prefect has an agent for) or lamda functions. Would appreciate to anyone’s experience.

What should be the expected behaviour if I do the following: 1. Run a flow with caching enabled on the tasks 2. Delete the cache files 3. Rerun the flow This would be using the built in

task_input_hash

as the

cache_key_fn

? On step 3 I’m seeing the flow fail that it can’t find the cache file…

Hi again, This is probably very basic python question, but I'm having trouble that Prefect can't find my module. I already have module

database

and a few files in it and all those work fine, but now that I created a new one, the run on the Prefect is saying that it can't find it. I'm probably just missing something obvious, but because of my lack of knowledge of both Python and Prefect I'd really appreciate help on this. Thank you in advance!

Noticing a change in behaviour for results since upgrading to 2.6. A task that would return an empty list now returns

None

. Wondering if this is intentional. Minimal example in the thread. Also found out I can’t downgrade from 2.6 to 2.5 because of what looks like some database migrations? Running

prefect orion database reset -y

on a database with 2.5 after upgrading to 2.6 and it fails.

Hey, how are people dealing with Prefect 2.0 CI/CD and agents with deployments/schedules/params, are there any good github action examples floating around?

Hi, I'm having an issue where I was trying to set up a profile with a workspace. I'm able to do this locally. And for some reason I'm stuck in this loop of trying to delete my profile for this account. But everytime i try and delete the profile the profile keeps adding '\\\' and i cant seem to call the profile to delete it. Is there a work around where i can delete the profiles?

Hey everyone, im trying to run orion on my local dev machine and keep running into Its Windows 11, Python 3.9 I have the same configuration on another windows machine and its working fine. Any ideas what might be causing this?

Error: Got unexpected extra argument (prefect.orion.api.server:create_app)

Hi, does anyone know if the prefect-dask lib should be able to handle the operator.KubeCluster from dask_kubernetes?

Hey everybody, I have just started with prefect graphql interface, but I cannot find how I should correctly do the pagination with it? Does anybody knows how to paginate response in prefect graphql?

Hello all. Any idea why the Prefect Flows are freezing when running them in ECS with Prefect Cloud ? They freeze and in the cloud they appear as they are not finished, but they never actually do anything after a certain period. And the ECS machines I use are the largest one possible. ?

Hi all, I started a fresh new installation with this Docker composition template https://github.com/fraibacas/prefect-orion pointing to the latest build. It is working as expected but the Blocks page is empty. Then using this API call

curl -X 'POST' \
  '<http://prefect-server:4200/api/block_types/install_system_block_types>' \
  -H 'accept: application/json' \
  -H 'x-prefect-api-version: 2.6.1' \
  -d ''

I’m getting few of them available from that page, the Remote File System one doesn’t come up though. Is this expected with the latest version?

Hi, I’m trying to hook up a min.io tenant running on a k8s as a storage block. Is that supported at the moment? Seems the aws s3-bucket doesn’t have get/put-directory?

Hi! I think I found a bug (it happens since 2.6.1 version), when a flow has consecutive

.map

calls, if one task fails in the first map, then all dependent tasks stay in NotReady state. Example flow in the comments

I'm dipping my toe back into prefect after last playing with 1.x. I'm trying to get my head around the deployment model. Previously I had a prefect agent running persistently in ECS, which would spin up new containers on demand to handle prefect jobs. I used "docker" storage to do that, which appears to not be a option currently. I can't tell if that's because it's not needed to achieve what I want to do... I'm open to using other storage options like S3, but I guess that's difficult if my flow code has external dependencies? I'm also not much of a python dev so I don't have much of an intuition on how bundling works in the python world!

Hello everybody, I am trying to make the orion db persistent by linking a volume to orion server, but the db still ephemeral and each time the container is crashed, data is missed. here is the manifest file I am using, looking forward to helping me to figure out what is the issue

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: prefect
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 3Gi
---
apiVersion: apps/v1
kind: Deployment
metadata:
  namespace: prefect
  name: orion
spec:
  selector:
    matchLabels:
      app: orion
  replicas: 1  
  template:
    metadata:
      labels:
        app: orion
    spec:
      containers:
      - name: api
        image: prefecthq/prefect:2.6.0-python3.10
        command: ["prefect", "orion", "start", "--host", "0.0.0.0", "--log-level", "WARNING"]
        ports:
        - containerPort: 4200
        env:
          - name: PREFECT_API_URL
            value: '<https://mydomain/api>'
        volumeMounts:
          - mountPath: ~/.prefect
            name: prefect
      restartPolicy: Always
      volumes:
        - name: prefect
          persistentVolumeClaim:
            claimName: prefect

Hi Everyone, is there a way to run an agent to run queues as local process using a virtual env? The agent seems to be picking up the deafult python on the machine it is running on

I'm getting a 404 not found error when trying to add a collaborator to the workspace. Anyone else experiencing this?

S.O.S. what is happening

File "/usr/local/lib/python3.9/site-packages/prefect/utilities/dispatch.py", line 186, in lookup_type

raise KeyError(

KeyError: "No class found for dispatch key 'ecs-task' in registry for type 'Block'."

Can someone suggest how to run an agent in remote server?

Need tech advice… Strange Error on a flow that used to work… I get

Finished task run for task with final state: 'TimedOut'

after 3 second the task is running even though the task is set to

@task(timeout=600, **params)

and it was running fine till friday. it also runs fine when using

flow.run()

I am using https://cloud.prefect.io/ any suggestions? I deleted the flow an reregister it, but it did not help

Hey, is Prefect Cloud 2.0 down? I've logged into the UI and can't see anything. The API/client still seems to be working though.

Can I run Prefect without the UI? I am looking at running some local testing using serverless fargate, but currently we would not be allowed to connect to the Prefect Cloud Service now are we allowed to run servers. Unless I can get away with running the UI inside fargate which may be an option?

Hello, Prefect is not validating the

cache

when running it with celery using this command.

celery -A app.tasks.celery_app worker --max-tasks-per-child=1 --loglevel=ERROR

Everything works fine when I use this command

celery -A app.tasks.celery_app worker --loglevel=ERROR

Thanks for the ability to not persist results! Now I can let prefect run without wiping out its storage every day :)

Hi all, Is it possible to use module name instead of full file name in orion deployments when using “native” storage? In prefect1 that was possible to specify just module nam, e.g.

flow.storage = Module('samples.flow_orion')

Now i’m kind of forced to specify exact full filename, even though I have file in python path. That’s very inconvenient for cases when developers have no control over the final image layout. That’s how I create deployments now:

entrypoint = 'samples.flow_orion:entrypoint' # this is just module, available on path
flow = prefect.utilities.importtools.import_object(entrypoint)

# here is path/entrypoint i need to specify path. could I use module name instead?
deployment = Deployment.build_from_flow(
    name=flow.name,
    flow=flow,
    work_queue_name="kubernetes",
    skip_upload=True,
    infrastructure=KubernetesJob(
        image=image,
        namespace='prefect2',
        job=k8s_template_orion(flow)
    ),
    path='/app/lib/python3.9/site-packages/samples/', # TODO 
    entrypoint='flow_orion.py:entrypoint'
)
deployment.apply()

Did i miss how to use module name instead of path?

When using String blocks - the value that is extracted from the block is String(value='thisisthevalue') - is there a built in way to only get the value of the actual string - thisisthevalue ?