Yes. In Prefect 3.x you can make task retries conditional with the

retry_condition_fn

option on a task. This function is called whenever the task raises an exception; if it returns True, Prefect will retry (subject to remaining attempts). If it returns False, Prefect will stop retrying and keep the failure. There isn’t a built-in “do not retry” exception type. The common pattern is to define your own exception type(s) and make your

retry_condition_fn

return False for those. Example 1: stop retrying for a custom non-retryable error

from prefect import task

class NonRetryableError(Exception):
    pass

def should_retry(exc: BaseException) -> bool:
    # Don’t retry for explicitly non-retryable errors
    if isinstance(exc, NonRetryableError):
        return False
    # Retry everything else
    return True

@task(
    retries=5,
    retry_delay_seconds=[5, 15, 30, 60, 120],
    retry_condition_fn=should_retry,
)
def do_work(x: int) -> int:
    if x < 0:
        # This will fail the task immediately with no further retries
        raise NonRetryableError("Bad input; retrying will not help.")
    # Simulate a transient error that should be retried
    raise RuntimeError("Temporary issue; please retry.")

Example 2: stop retrying for 4xx HTTP errors but retry others

from prefect import task

def retry_on_http_error(exc: BaseException) -> bool:
    # Example for httpx; adapt for your HTTP client
    try:
        import httpx
        if isinstance(exc, httpx.HTTPStatusError):
            status = exc.response.status_code
            # Do not retry on 4xx, do retry on 5xx
            return status >= 500
    except Exception:
        pass
    # Retry by default for other exceptions
    return True

@task(retries=3, retry_condition_fn=retry_on_http_error)
def call_api(...):
    ...

Key points - Set

retries

to the max attempts you want and use

retry_condition_fn

to dynamically permit/deny each retry based on the raised exception. - Raising a specific error alone will not skip retries; you must pair it with logic in

retry_condition_fn

. - You can define

retry_condition_fn

in the task decorator (as above). If you need to change it at call time, you can wrap your function with

task(...)

and construct variants with different options. If you want, share a snippet of your task and the errors you’d like to treat as non-retryable, and I can help tailor the condition function.