https://kedro.org/ logo
Powered by
#plugins-integrations
Title
# plugins-integrations
l

Linda Sun

05/01/2023, 3:17 PM
Hi all! I am using kedro-azureml plugin to run kedro pipeline in azureml. In the final step, I used 
kedro azureml run --pipeline=xxx -s xxxxx to run the pipeline in azure ml
, and met this error:
Copy code
azure.core.exceptions.ResourceExistsError: (UserError) The run has been submitted somewhere else, please change the run id and try again. Existing pipline run id: *******
Code: UserError
Message: The run has been submitted somewhere else, please change the run id and try again. Existing pipline run id: *******
I’ve used azure cli to list all jobs that I have in the workspace, and deleted the job with the run id it is complaining about. But still met the same error 😶. Would love to get some insights from the group: 1). any clue on how to solve the error? 2)Is there any way that I can set job name or run id in kedro project or with the kedro-azureml cmd (In azureml.yml seems users can only set experiment name)? Thank you!!
m

marrrcin

05/01/2023, 4:08 PM
Are you using docker flow or code flow?
l

Linda Sun

05/01/2023, 4:13 PM
I used kedro-docker. Here is my process: use kedro-docker build image ==> docker push to ACR ==> kedro azureml run (Same as the youtube tutorial) here is code_directory in my azureml.yml: 
code_directory: ~
I also tried 
code_directory: "."
and I’m getting same error
Hi @marrrcin, let me know if you have any thoughts or if you need me to provide further information. Thank you 🙂!
m

marrrcin

05/02/2023, 10:39 PM
I will get back to you on Thursday, I'm on off right now
🙏 1
Hi again @Linda Sun, I would like to help but this is the first time I’m seeing this issue. Our code base always goes through end-to-end tests, which involve runnig the actual pipeline (from Spaceflights starter) in Azure ML and as of latest release (
0.4.0
) it works smoothly. I suspect that there is some issue with your local configuration. Could you provide more details on how your project is configured, where do you invoke the 
kedro azureml run
etc? As for second point:
2)Is there any way that I can set job name or run id in kedro project or with the kedro-azureml cmd (In azureml.yml seems users can only set experiment name)?
The pipeline name is automatically taken from Kedro’s pipeline name: https://github.com/getindata/kedro-azureml/blob/2e5836b72256d7455d8525c7769a68d4c844ccf7/kedro_azureml/generator.py#L90 Run ID for the Azure job is assigned automatically by the Azure ML SDK (it’s not our code).
l

Linda Sun

05/05/2023, 1:55 PM
Thank you for the detailed explanation, @marrrcin! For the first point, I run 
kedro azureml run
in local ternimal (VS code). I think the code structure follows standard kedro project. I actually tried the method of uploading code (docker image only contains requirements), as I might modify code and then kedro azureml run the same pipeline again. Sometimes I got the error too, but it is less frequent than using docker flow. Let me know if you need to know more details. Thank you for the information on the second point!
m

marrrcin

05/05/2023, 1:57 PM
Maybe you can create a new project on a side, with new virtual env, to verify whether this happens on fresh projects - this will help to exclude some issues. Just yesterday I’ve revamped the docs, so the quickstart is more up to date.
Maybe if you could share some code that causes the issue / provide a steps to reproduce, then we will be able to fix it
2 Views
Powered by