@Marvin how can I override timeout_seconds task argument when calling the task, so that the user can override or define timeout_seconds when using a predefined task something like @task(timeout_seconds = timeout_seconds) def mytask(otherargs, timeout_seconds:int=10)) sleep(11) @flow def myflow(): mytask(timeout_seconds=12) mytask() # should timeout

It seems that the search did not directly return the specific way to override the

timeout_seconds

argument when calling a task. However, based on my general understanding of how Prefect works, one way you could achieve this is by setting the

timeout_seconds

within the task decorator options and then it should apply the timeout when the task runs. However, since we need a specific example conforming to your request, let's determine whether you can pass the timeout dynamically when calling the task from the flow without having to define it upfront in the task decorator. I'll continue to search for a more definitive answer. While I gather more precise information, typically, task timeouts are set as arguments in the task definition and not typically overridden at call time. Therefore, what you want to achieve isn't directly supported by the classic use of decorators and might require another approach. One such approach might be to use context managers inside your task to enforce a timeout, which could be passed as an argument from the flow level. Nonetheless, let's try to confirm this with the proper documentation or examples. I will look through code examples to find if there is a way to dynamically set

timeout_seconds

for a task when calling it from a flow. Please hold on for a moment. The code example provided does not specifically address the use of

timeout_seconds

dynamically within a task. Typically, a task's timeout is specified statically in the task decorator, and the dynamic specification of timeouts is not the usual behavior in workflows. If the dynamic assignment of timeouts is critical for your requirement, you might consider an approach using Python's standard library, such as managing timeouts with

concurrent.futures

within the task definition, so the timeout value can be passed as an argument. Remember, this is more of a custom workaround than a built-in Prefect feature. For official guidance, it would be beneficial to consult Prefect's documentation or community forums for any recent updates or patterns that others have used to address this scenario. It might be fruitful to look for a similar issue on Prefect's GitHub repository to see if this has been discussed or addressed in any issue or PR. Shall we try that?

Yeah please check

@Marvin

with_options

?

The search returned information about AWS ECS worker and job configuration, but it appears not to include documentation on how to override

timeout_seconds

using

with_options

. However,

with_options

is indeed a method provided in Prefect for setting task options dynamically when calling a task within a flow, including timeout settings. For example, if you were using Prefect, you might be able to use the

with_options

method to set the

timeout_seconds

for a specific task call like this:

python
from prefect import task, flow

@task
def my_task(some_arg):
    # Your task implementation here
    pass

@flow
def my_flow():
    my_task.with_options(timeout_seconds=12)(some_arg='value')
    my_task(some_arg='another value')  # Without option override, will use default timeout

In this example,

my_task.with_options(timeout_seconds=12)(some_arg='value')

would call the

my_task

with a

timeout_seconds

of 12, overriding any default timeout set in the decorator. It's key to remember that you must call

with_options

before calling the task (as in

my_task.with_options(...)

) to override the task options for that specific run. For the most up-to-date usage, refer to the Prefect documentation on task options or the exact section on Configuring tasks with with_options. (I'd share a direct link, but my circuits seem to be malfunctioning again.) Please note that the links are for general references and might require navigation to the specific section.

@Nate 🙌 https://github.com/PrefectHQ/prefect/blob/6df6feb4087ed8960aa638bbc3be9ea03de83c25/src/prefect/tasks.py#L367

👍

Yep indeed - I'll comment on it

🙏