Lessons Learned From Running Apache Airflow at Scale - Deterministically-random round crontab generator

random_airflow_schedule.py

from hashlib import md5
from random import randint, seed

def compute_schedule(dag_id, interval):

    # create deterministic randomness by seeding PRNG with a hash of the table name:
    seed(md5(dag_id.encode()).hexdigest())

    if interval.endswith("h"):
        val = int(interval[:-1])
        if 24 % val != 0:
            raise ValueError("Must use a number which evenly divides 24.")
        offset = randint(0, val-1)
        minutes = randint(0, 59)
        return f"{minutes} {offset}/{val} * * *"

    elif interval.endswith("m"):
        val = int(interval[:-1])
        if 60 % val != 0:
            raise ValueError("Minutes must use a number which evenly divides 60.")
        offset = randint(0, val-1)
        return f"{offset}/{val} * * * *"

    elif interval == "1d":
        return f"{randint(0, 59)} {randint(0, 23)} * * *"

    raise ValueError("Interval must be (1, 2, 3, 4, 5, 6, 10, 12, 15, 20, 30)m, (1, 2, 3, 4, 6, 12)h or 1d")