hi, we're using prefect server as our backend, and using a local agent on our infrastructure to execute flows. we're noticing that for some flows, only the first level of the dag (ie, the tasks with no upstream requirements) get executed, and then nothing happens... any clues as to what might be happening?

Hello everyone, is there such a way to use Dask Temporary Cluster, created by Prefect, inside a task in order to execute any workload on that Dask cluster?

Hey folks! I’ve been struggling a bit with understanding how Prefect connects with the AWS ecosystem. Namely ECR, ECS, and Fargate. I’d be willing to pay for a few hours of tutoring. Please DM me your rates :)

hi! I have a Task that is querying a slow API (10K times). I want to checkpoint this task each iteration (disk is okay), but struggling with how to template the

location

since the task argument is what makes the task unique, but that’s not available to the context object as far as I can tell. Am I missing something obvious for this use case?

hi Prefect, I noticed that the

prefect/cli/server.py

shows user can specify different ports to start the Prefect server services (i.e. graphql_port='xxxx', server_port='xxxx', etc.). But when I used the flow.register(project_name) in my script, the

prefect/core/flow.py

seems like using default 4200 port to register the flow and had no param to adjust the registering port in

register()

function, so it raised HTTPConnection error. Once I changed line 1666 - flow.py

client = prefect.Client()

with

client = prefect.Client(api_server='<http://localhost>:MODIFIED_PORT/')

and it works. Is this an issue here? Thank you 🙂

Hello everyone, I have a question. Is it possible to rerun only a subset of the mapped tasks of a flow ? Like only the ones which have failed.

Hello, I try to understand how applicable Prefect as a deployment tool. In a simplest form what I'm trying to achieve is this: 1. Read IP list of an unknown length (2, 3, 42, etc) 2. For each IP run a number of tasks sequentially (forbidden to run in parallel): 2.1 stop IP 2.2 wait new IP 2.3 init new IP Additional requirements: * Some task should be retryable (done) * Some task should be manually recoverable (done) Here's what I came up with:

import random

import prefect
from prefect import task, Flow
from prefect import Parameter
from prefect.engine.signals import FAIL, LOOP, PAUSE

@task(max_retries=3, retry_delay=datetime.timedelta(seconds=5))
def get_ips() -> list:
    if random.random() > 0.8:
        raise FAIL("Failed to get IPs")
    count = random.randint(2, 6)
    ips = [f"IP-{i}" for i in range(1, count)]
    print(ips)
    return ips

@task
def seq_reduce(func, init: any, lst: list) -> any:
    loop_payload = prefect.context.get("task_loop_result", {})

    index = loop_payload.get("index", 0)
    acc = loop_payload.get("acc", init)

    if index >= len(lst):
        return acc

    acc = func(lst[index], acc)
    raise LOOP(message=f"acc={acc} index={index}", result=dict(acc=acc, index=index+1))

def seq_map(func, lst: list) -> list:
    def aux(x, acc: list) -> list:
        acc.append(func(x))
        return acc
    return seq_reduce(aux, [], lst)

@task
def stop_ip(ip: str) -> None:
    print("stop_ip: " + ip)
    pass

@task
def wait_for_ip(ip: str) -> str:
    print("wait_for_ip: " + ip)
    return ip + "-new"

@task
def init_ip(ip: str) -> None:
    print("init_ip: " + ip)
    if random.random() > 0.5:
        raise PAUSE("Initing IP failed. Waiting for human input")
    return "Done"

with Flow("process_ip") as process_ip_flow:
    old_ip = Parameter("old_ip")
    stop_ip(old_ip)
    new_ip = wait_for_ip(old_ip)
    init_ip(new_ip)

def process_ip(old_ip):
    state = process_ip_flow.run(old_ip=old_ip)
    #process_ip_flow.visualize(flow_state=state)
    return state.result[new_ip].result
    
with Flow("process_ips") as process_ips_flow:
    old_ips = get_ips()
    new_ips = seq_map(process_ip, old_ips)

if __name__ == "__main__":
    state = process_ips_flow.run()
    #process_ips_flow.visualize(flow_state=state)
    print(state.result[new_ips].result)

Is the above what Prefect is designed for? Are there some other approaches?

Hi I want to try Prefect I decide setup VM with win 10 for that maybe is any system requirements for the prefect (Core, RAM, etc.)?

Hey all! One of the useful things we can do in airflow is to set a start date of a task far back in the past and it will run through all the “missed” executions until its up to date. Great for polling down a ton of historical data. Is there functionality within Prefect for this? Ive been trying to see if theres some way of configuring Clocks/schedules, etc, but on the face value Im not sure if this will have the wanted behaviour.

Hello all, I am working on setting up Slack Notifications with Prefect and I am running into some issues the following code does not work but if I replace the request.post(my_secrect) call with the Slack webhook address directly it prints the is_failed() message as expected. What am I doing wrong with the configuration of the Secrets?

def post_to_slack(task, old_state, new_state):
    my_secret = PrefectSecret("SLACK_WEBHOOK_URL").run()
    if new_state.is_retrying():
        msg = "Task {0} failed and is retrying at {1}".format(task, new_state.start_time)

        # replace URL with your Slack webhook URL
        <http://requests.post|requests.post>(my_secret, json={"text": msg})
    elif new_state.is_failed():
        msg = "Task {0} failed".format(task)

     # replace URL with your Slack webhook URL
        <http://requests.post|requests.post>(my_secret, json={"text": msg})

    return new_state

I specifically receive this message if I leave the code as it is above:

Exception raised while calling state handlers: ClientError('400 Client Error: Bad Request for url: <http://host.docker.internal:4200/graphql>\n\nThe following error messages were provided by the GraphQL server:\n\n    GRAPHQL_VALIDATION_FAILED: Cannot query field "secret_value" on type "Query".\n\nThe GraphQL query was:\n\n    query($name: String!) {\n                secret_value(name: $name)\n    }\n\nThe passed variables were:\n\n    {"name": "SLACK_WEBHOOK_URL"}\n')
Traceback (most recent call last):
  File "/usr/local/lib/python3.8/site-packages/prefect/client/secrets.py", line 137, in get
    value = secrets[self.name]
KeyError: 'SLACK_WEBHOOK_URL'

Thank you in advance!

hi, is there a quick way to check the health and status of the lazarus process/service? perhaps a REST endpoint we could query or a command line utility we can invoke?

Hi All, I am trying to determine what AWS IAM policies are required for running flows as ECS tasks. If my flow storage is in S3 where would I assign the appropriate role so that the ECSAgent can read them? These are the places I can think of: • on the server running

prefect agent ecs start

with any of the aws cli ways • setting

task_role_arn

on either agent start or

ECSRun

• setting

execution_role_arn

on either agent start or

ECSRun

• In the task container Any insight on the differences of these would also be appreciated.

Hey all, just wondering if there are any good tips / links for Prefect project structure / how to save your tasks + flows for easy source control, etc

Hi all, I am having issues with Prefect Cloud at the moment… getting an error trying to add a coworker to the team. It is saying that we have no additional space for users, when the account dashboard says that we do. Can someone message me offline to resolve?

So is the

Prefect Agent

responsible for passing output of one task as input of another task in Prefect flow DAG since the Prefect server manages no code or input/output data. i'm trying to wrap my heads around what is the responsibility of the Agent. Maybe an architecture diagram might help.

Hello all, fairly new to Prefect here, trying to set up a flow-of-flows following the docs here: https://docs.prefect.io/core/idioms/flow-to-flow.html Ultimately I'd like to kick off only one flow using a schedule and chain a bunch of dependent flows to run, I tried the attached code but the runner is getting stuck on "Flow a" and not proceeding. Could someone take a look and tell me what I'm doing wrong?

python a_flow.py
python b_flow.py
python parentflow.py

[2021-02-02 15:19:53-0500] INFO - prefect.FlowRunner | Beginning Flow run for 'parent-flow'
[2021-02-02 15:19:53-0500] INFO - prefect.TaskRunner | Task 'Flow a': Starting task run...

Q: Can the same workflow be submitted concurrently?. i.e. I want to run 2 instances of the same workflow at same time from a Prefect server. So I assume that some UUID must be generated under the hood by Prefect to allow this? I know in a

Argo Workflow

, some random string is generated and appended to a workflow to disambiquate 2 workflows having the same name.

Hi folks we came across a bug when chaining more than one decorator with the prefect task decorator

Hi - I'd like to confirm that this is indeed a bug, and not something on my end. Thank in advance https://github.com/PrefectHQ/prefect/issues/4054

hi, i noticed that when a flow that is registered with the Prefect Server gets picked up by an agent to run it, the

CloudTaskRunner

seems to use a

LocalResult

to persist the results of every task... can someone point me to where this result object gets created?

another question: is it possible to have lazarus kick in sooner? right now we're seeing it kick in to resubmit flow runs at around the 15-17 minute mark, is there some config value we can set to bring this down a bit?

Hello team, we changed base for our worker container from

prefecthq/prefect:all_extras-0.13.9

to

prefecthq/prefect:0.14.5-python3.8

and now a lot of tasks a marked as failed because of

State Message:
No heartbeat detected from the remote task; marking the run as failed.

Did anyone else come across a similar usecase ? It feels to me like a common problem.

hi folks! I'm currently looking for a solution to store a csv file generated with python to a snowflake stage. Does anybody have experence with this? thanks!

Is it possible to use S3 storage if your flow depends on code in external modules? I keep running into this error:

Failed to load and execute Flow's environment: StorageError('An error occurred while unpickling the flow:\n  ModuleNotFoundError("No module named \'project\'")\nThis may be due to a missing Python module in your current environment. Please ensure you have all required flow dependencies installed.')

Using ECSAgent and a custom image in ECR. I have no problem importing from the external module inside the container or in the environment where my agent is launched or where my flow is registered from.

Hi all! Im trying to get a prefect server deployed into a swarm, and cant use port 8080 for the UI. I notice in the docs that we can configure prefect (https://docs.prefect.io/core/concepts/configuration.html) but I cant find a list of the actual variables availables to customise. Is this online somewhere else?

Also, it seems like the official prefect image is configured to display the ascii prefer splash screen (below) and then immediately terminate? The docs say to just use docker-compose, but point to the generic docker-compose website and not an example compose file. Is there an example compose file which has the server configured to keep the server online?

_____ _____ ______ ______ ______ _____ _______
| __ \| __ \| ____| ____| ____/ ____|__ __|
| |__) | |__) | |__ | |__ | |__ | | | |
| ___/| _ /| __| | __| | __|| | | |
| | | | \ \| |____| | | |___| |____ | |
|_| |_| \_\______|_| |______\_____| |_|

Thanks for using Prefect!!!

This is the official docker image for Prefect Core, intended for executing
Prefect Flows. For more information, please see the docs:
<https://docs.prefect.io/core/getting_started/installation.html#docker>

Apologies for the chat spam, Im just wondering what else I am supposed to do after running

prefect server start

locally (which launches fine and gets me to the Welcome to PREFECT SERVER ascii splash screen). Navigating to localhost:8080 shows me connect to the server (and pull down the favicon+title) but the web page is empty (completely white). I can see its got content and scripts in the html, but theres this error and nothing is every made visible in the UI:

Following up from yesterday.. situation is as follows: I could not add a new member to Cloud; they were previously a member of our team, but since they could not sign in (forgot password) I tried to remove them and add them back in. Now, an invite is in my team members inbox, when they click to join - it takes them to the sign in page.. understandable. Now, we cannot sign in because they do not remember the password and when you click forgot password, this email comes through… asking for sign-in help brings you back to the forgot password screen… just keeps going in a cycle. How can I get my team members password reset, given they are now a pending invite to our team?

Hi all! I am running the server locally and initially when opening the UI Id be able to see the toolbar but nothing else. After some mindless googling, I found an instruction from here https://github.com/flavienbwk/prefect-docker-compose says I need to run

prefect backend server
prefect server create-tenant --name default --slug default

And doing this does seem to fix the problem. Is there a way I can specify in the config options or somewhere in the docker-compose that I’d like to ensure this tenant exists on server startup?