yeah, your assumption is right, it will start one scheduler for each forked process and it will run the same job multiple times.
If you are running this service on multiple servers, you could keep only one active and the others as idle,
we achieved it using uwsgi legion.

            if os.environ.get("scheduler_lock") == "1":
                scheduler.start()
                logging.debug("Schedular Started ,-------------")
                os.environ["scheduler_lock"] = os.environ.get("scheduler_lock") + "1"

before start the program ,you had better to export scheduler_lock="1"

Sorry for the late update
@everhopingandwaiting yea that is one way of doing it, but I am doing it differently
Running scheduler as a rpc server outside the app, so its a different process now....so whenever there is a scheduling request, it gets connected to the rpc and process the request there....

@viniciuschiele closing this

@viniciuschiele @viniciuschiele fcntl mscvrt

Why did you add this code? Unable to dynamically modify timing tasks?

import atexit
import fcntl
from apscheduler.events import EVENT_JOB_ERROR, EVENT_JOB_MISSED, EVENT_JOB_EXECUTED
def scheduler_init(app):

f = open("scheduler.lock", "wb")
try:
    fcntl.flock(f, fcntl.LOCK_EX | fcntl.LOCK_NB)


    scheduler.add_listener(job_listener, EVENT_JOB_ERROR | \
                       EVENT_JOB_MISSED | \
                       EVENT_JOB_EXECUTED)
    scheduler.init_app(app)
    scheduler.start()
except:
    pass
def unlock():
    fcntl.flock(f, fcntl.LOCK_UN)
    f.close()
atexit.register(unlock)