Using kfp to submit pipelines fails

General Discussions microk8s
1

Hey Guys,
I installed microk8s and is working. I can upload and run a pipeline using the UI. Now I installed kfp python api (issues exist with the cli too) and I’m trying to submit a pipeline using the cli. Here is the api for the client function:

def __init__(self, host=None, client_id=None, namespace='kubeflow', other_client_id=None, other_client_secret=None, existing_token=None, cookies=None):

I’m only passing the host 10.64.140.43.xip.io, but it fails with

AttributeError: 'NoneType' object has no attribute 'id'

Which suggests that the host and/or other default parameters are wrong. Unfortunately there is no doc whatsoever anywhere. My guess is I should somehow provide a user pass too. Can someone provide some concrete steps to filling the parameters so thet I can submit a pipeline programmatically?

Thanks

2

Try this: client = kfp.Client(host=‘pipelines-api.kubeflow.svc.cluster.local:8888’).
This helped me resolve the HTTPConnection error and AttributeError: ‘NoneType’ object has no attribute ‘id’

3

Thanks, but it did not work either, I got connection error:

urllib3.exceptions.MaxRetryError: HTTPConnectionPool(host='pipelines-api.kubeflow.svc.cluster.local', port=8888): Max retries exceeded with url: /apis/v1beta1/experiments (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7f5d95a485c0>: Failed to establish a new connection: [Errno -3] Temporary failure in name resolution',))
4

Try this from the given link:https://ubuntu.com/blog/data-science-workflows-on-kubernetes-with-kubeflow-pipelines-part-2
To ensure access to the packages needed through your Jupyter notebook instance, begin by installing Kubeflow Pipelines SDK (kfp) in the current userspace:

!pip install -q kfp --upgrade --user

5

Thanks,
That link does not look right,
Also, I’m not running from jupyter, I’m running directly the local host (python pipeline.py) on microk8s machine.

6

@hamidi did you solve this issue I am stuck on similar issue through my .py file

7

sadly, not.

8

Try this: https://growingdata.com.au/kubeflow-running-and-deploying-pipelines-via-the-sdk/ also track issue here: https://github.com/kubeflow/examples/issues/827