How to get the final exception from a task when using SmartRetryMiddleware

[NoisyCake]

Hi! 👋

I'm using SmartRetryMiddleware with retry_on_error=True. I don't really unserstand the thing:
My goal is to catch the final exception in the calling code, but after the final retry, the original exception is not propagated back. Should I use task.wait_result()?

@broker.task(retry_on_error=True)
async def send_broadcast_message(*args):
    await do_the_thing(*args)

[abderrahimchadli]

Yes—wait on the task handle returned by .kiq(), then explicitly raise the error from the returned result. In the current implementation, wait_result() returns a TaskiqResult; it does not itself re-raise the task's exception.

from taskiq.exceptions import TaskiqResultTimeoutError

task = await send_broadcast_message.kiq(*args)
try:
    result = await task.wait_result(timeout=300)
except TaskiqResultTimeoutError:
    # No final result arrived within this client's wait limit.
    # This does not cancel the worker or prove retries are exhausted.
    raise
else:
    result.raise_for_error()  # raises the error stored in the result
    value = result.return_value

Configure the same persistent result backend on the producer and workers. With SmartRetryMiddleware(no_result_on_retry=True) (the default), retrying attempts use NoResultError, so the receiver skips saving those intermediate results. Once the retry limit is reached, the final failure can be saved and retrieved. Setting no_result_on_retry=False can expose an intermediate failure before retries finish.

Allow enough timeout for execution plus retry delays. Exception reconstruction also depends on the serializer and exception type; don't assume the original worker traceback is preserved.

Source: wait_result, raise_for_error, retry handling, and receiver result saving. If this still doesn't return the final failure, which Taskiq version and result backend are you using?