SmartRetryMiddleware retries immediately without scheduler, but too late with scheduler (delay not respected)

[NoisyCake]

I’m using SmartRetryMiddleware to handle retries for failed tasks:

broker.with_middlewares(
    SmartRetryMiddleware(
        default_retry_count=3,
        default_delay=2,
        use_jitter=True,
        use_delay_exponent=True,
        max_delay_exponent=10,
        schedule_source=redis_source
    )
)

I noticed that when I don’t use a scheduler, retries happen immediately, one after another — effectively ignoring the default_delay parameter.

However, when I do use scheduler, retries are only executed after the scheduler runs its

Getting previous time schedules

cycle — which happens once per minute by default

redis_source = ListRedisScheduleSource(
    url=redis_url
)

Can you please explain how to solve this problem?

[s3rius]

It's because generally scheduler gathers tasks once every minute. You can increase how often schedules are being pulled from sources by specifying --update-interval as an option of taskiq scheduler command.