Mitch
06/15/2023, 3:04 PMMitch
06/15/2023, 3:06 PMfrom x import myflow
and then calling myflow(params)
vs doing run_deployment(name="myflow/mydeployment")
?Ryan Peden
06/15/2023, 3:36 PMprefect profile create local
and then prefect profile use local
.
The downside is that if your testing on blocks saved in Prefect Cloud, they won't work.
My guess with the debugger shutdown is that it's terminating the container without giving the flow a chance to shut down cleanly. If the Prefect engine can't stop the flow cleanly and make the API call to update the flow state to crashed/cancelled, the bank end won't know if the flow crashed or is just a long-running flow that hasn't completed yet.Ryan Peden
06/15/2023, 3:39 PMfrom x import myflow
and then myflow(params)
will run the flow as a subflow in-process in the same Python interpreter as your main flow is running.
In contrast, run_deployment(name="myflow/mydeployment")
submits the run to your flow run infrastructure, where an agent or worker will pick it up and run it independently. At the very least, it'll be in its own process, and depending on your infrastructure setup it might even run in a different container/VM/server than the flow from which you called run_deployment
.Ryan Peden
06/15/2023, 3:41 PM