So ended up renaming the flow name and version
p
So ended up renaming the flow name and version
k
Hey @Prateek Gupta, Prefect is indeed mainly for batch jobs. What commits are you referring to? A bit unlcear what you are saying. Could you give me more details?
p
We have triggers in GitHub master branch which will register the flows again and schedule them
k
Ah ok. That sounds good. And the task name has issues? Do you have a specific error?
p
So for a small change in code .. we made a commit which triggered the registration process.. and it required changing the version number
I just started yesterday and there may be some gaps what my team mate told me
k
Ah ok so in general, code changes will need a re-registration of the Flow to sync with what Prefect has stored (metadata). Are you using Github/Gitlab/Git storage for the Flow? In some cases, you don’t need to re-register. It’s mainly needed if the structure of the DAG changes. You would need to register and bump up the version number. In the future though, we will move to having users supply the version so you’d be able to supply the Git hash as the version
Is the bump in version number a problem for you?
p
No..bumping version is not a problem..we do not have devops .. so we are doing it for the first time .. so it took us a lot of time to figure out
@Kevin Kho I am watching your ML tutorial on YouTube. Flow.register is supposed to be called on the server itself?
How is the flow appearing in the UI web server?
k
flow.register registers a script to your backend (cloud or server). If you do
prefect backend server
, it will be on the server. Otherwise it will be on cloud. Only registered flows triggered with the "Run" button or on a schedule will appear in the UI.
flow.run()
flows will not