Avoid scheduled pipeline to be triggered immediately after creation in azureml (Python SDK v2)

Hello Jaume Amores,

I hope this has been helpful! We appreciate hearing from you and would love to help others who may have the same question. Accepting answers helps increase visibility of this question for other members of the Microsoft Q&A community. Thank you for helping to improve Microsoft Q&A!

Hi @Anshika Varshney . I tried what you suggested but using the RecurrenceTrigger and JobScheduler class, as follows:

import datetime as dt

from azure.ai.ml.entities import (
    RecurrenceTrigger,
    RecurrencePattern,
    JobSchedule,
)
from azure.ai.ml.constants import TimeZone

# --------------------------------------------------
# parameters that can change
# --------------------------------------------------
start_time="2025-12-01 09:00"
frequency="day"
interval=7
schedule_name="my_schedule"
hours=9
minutes=0

# --------------------------------------------------
# code
# --------------------------------------------------

# convert start_time to the format required by Azure ML if needed
td = dt.datetime.strptime(start_time, "%Y-%m-%d %H:%M") - dt.datetime.utcnow()
start_time = dt.datetime.utcnow() + td

schedule = RecurrencePattern(hours=hours, minutes=minutes)
recurrence_trigger = RecurrenceTrigger(
    frequency=frequency,
    interval=interval,
    schedule=schedule,
    start_time=start_time,
    time_zone=TimeZone.UTC,
)

job_schedule = JobSchedule(
    name=schedule_name, trigger=recurrence_trigger, create_job=job_object # reference to the job object created before
)

job_schedule = ml_client.schedules.begin_create_or_update( # ml_client is the Azure ML client object
schedule=job_schedule
).result()

This effectively creates a schedule that has as start time the one indicated. However, the first job of such schedule runs immediately after creation. It doesn't wait until the start time. If we check the settings of the schedule, we can see that the start time is the one desired, the 12th of December at 9:00, but it didn't wait until then. Is there anything wrong with this?

Hello Jaume Amores,

thanks for sharing the details and code snippet. what you’re seeing is actually the expected behavior in azure machine learning scheduling. when you create a job schedule with a recurrence trigger, the service validates the schedule and immediately starts the first run, even if the start_time is set for the future. this happens because the start_time defines when the recurrence pattern begins, but the first job run is triggered automatically as part of the schedule creation process. it’s designed this way to confirm that the job configuration is valid and executable.

to avoid the immediate run, you can create the schedule with the status set to “disabled” and then enable it later when you’re ready for it to start. another option is to create the schedule through the rest api or cli and include the “no_initial_run” property, which prevents the first run from being triggered automatically.

for example, using the cli command

az ml schedule create --file schedule.yml --set properties.no_initial_run=true

will help achieve this behavior. note that the current python sdk does not provide a direct parameter for skipping the first run.

for more details, you can refer to the official documentation on recurrencetrigger class reference.

hope this helps! please click accept answer if this resolves your query.

Thanks @Anshika Varshney Just the last question: how do i create it with status disabled in python, and if I do that and then enable it programmatically at a future date, will the first job run immediately or will wait until start_time?

Hello Jaume Amores,

Great follow-up question! yes, you can create a schedule with status set to “disabled” using the python sdk. this ensures the job doesn’t run immediately after creation. once you’re ready, you can enable it programmatically, and the schedule will start following its defined recurrence pattern.

to create it in a disabled state, you can do something like this:

from azure.ai.ml.entities import JobSchedule

job_schedule = JobSchedule(
    name="my_schedule",
    trigger=recurrence_trigger,  # your RecurrenceTrigger object
    create_job=job_object,       # reference to the job object
    status="Disabled"            # key part to prevent immediate run ) 

later, when you want to enable it, just update the status to “Enabled”:

job_schedule.status = "Enabled"
ml_client.schedules.begin_create_or_update(schedule=job_schedule).result()

when re-enabled, the schedule will respect your configured start_time and recurrence. it won’t trigger immediately unless the current time has already passed the start_time.

for more details, check out the official documentation:

how to schedule jobs

azure.ai.ml JobSchedule reference

hope this helps! please click accept answer if this resolves your query.

Hello Jaume Amores,

Did you get any chance to review the above response.

Thank you!

Thanks @Anshika Varshney . I tried the code you indicated. If I indicate status="Disabled", it gives an error:

File "c:\azure\ai\ml\entities\_schedule\schedule.py", line 227, in __init__
    super().__init__(
  File "c:\azure\ai\ml\entities\_schedule\schedule.py", line 75, in __init__
    super().__init__(name=name, description=description, tags=tags, properties=properties, **kwargs)
  File "c:\azure\ai\ml\entities\_resource.py", line 67, in __init__
    super().__init__(**kwargs)
TypeError: object.__init__() takes exactly one argument (the instance to initialize)

The same happens if I indicate "Enabled". However, if I don't indicate this parameter it runs fine.

Another thing is that, in the links that you provide, the second one gives a 404 not found error. The first one is found but doesn't indicate the status as one of parameters.

The error occurs because the current Azure ML Python SDK does not support passing status as a parameter in the JobSchedule constructor. That’s why adding status="Disabled" or "Enabled" throws a TypeError.
It happens because The JobSchedule entity in the SDK only accepts specific arguments like name, trigger, and create_job. also status is not an exposed parameter in the constructor, so Python raises an error when you include it.

Currently, the SDK does not allow setting the status during creation. The correct approach is:

Create the schedule without status.

job_schedule = JobSchedule(
    name="my_schedule",
    trigger=recurrence_trigger,
    create_job=job_object
)

ml_client.schedules.begin_create_or_update(schedule=job_schedule).result()

Then update the status using Azure CLI or REST API:

az ml schedule update --name my_schedule --set status=Disabled

This way, you can disable the schedule after creation and enable it later when needed.

Thankyou!