Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[bug] After upgrade to kfp 2.0(latest), not able to run old pipelines created using kfp v1. "Start" button disabled #10153

Closed
mChowdhury-91 opened this issue Oct 25, 2023 · 11 comments · Fixed by #10276
Assignees
Labels

Comments

@mChowdhury-91
Copy link

mChowdhury-91 commented Oct 25, 2023

Environment

  • How do you deploy Kubeflow Pipelines (KFP)? - Standalone deployment
  • KFP version:
    kfp 2.2.0
    kfp-pipeline-spec 0.2.2
    kfp-server-api 2.0.2

Steps to reproduce

Recently we have upgraded kubeflow pipelines to support KFP V2, but noticed that backward compatibility is breaking. The pipelines that were created earlier using V1 is no more executable.
The "Start" button is disabled as the error says:"A pipeline must be selected" and under Run parameters section the parameters are not visible "This pipeline has no parameters".

Expected result

The v1 executions should work as before, without having to re-upload a new yaml version.
Note: If we upload the a new version with same yaml content, it works(shows the run parameters)

Impacted by this bug? Give it a 👍.

@jlyaoyuli jlyaoyuli self-assigned this Oct 26, 2023
@mChowdhury-91
Copy link
Author

@jlyaoyuli I did some more analysis on the above issue and here are my findings:
After upgrade, The PipelineSpec column is added new to pipeline_versions table which holds the data of pipeline yaml. Any new version upload, creates a new entry in the PipelineSpec column and the pipeline works.
Also, if we manually edit the database column to upload a yaml entry to old pipelines, they seem to work too..

Before upgrade, the data for pipeline yaml was being read/fetched from minio directly.
The backward compatibility seems to be broken.

In my opinion, this is the possible reason why old pipelines created before upgrade aren't working. Let me know your thoughts and analysis and how this can be fixed.

@jlyaoyuli
Copy link
Contributor

Hello @mChowdhury-91 , thanks for your update! Would you mind providing the pipeline you used? V1 pipeline works well from my side.

@mChowdhury-91
Copy link
Author

@jlyaoyuli None of the pipelines which were uploaded before the upgrade works, but as I mentioned, if we upload a new version of the same pipeline, it works well(V1 as well)

@thesuperzapper
Copy link
Member

@mChowdhury-91 can you share a minimal pipeline that does not work, because I have a number of pipelines that do successfully work.

@subasathees
Copy link

@thesuperzapper @mChowdhury-91 @jlyaoyuli , I am also facing the same issue. None of the default pipeline running after upgrade to kubeflow v1.8.
I am using hello-world pipeline and it givens below error.
Error:
image

Hello world pipeline reference: https://www.kubeflow.org/docs/components/pipelines/v2/hello-world/

@Davidnet
Copy link
Member

@subasathees Hi that is not an error, is a warning, that at the moment, the component runs in a python3.7 enviroment and that in the future it will use 3.8.

@DnPlas
Copy link

DnPlas commented Nov 20, 2023

Hi folks, I am actually hitting a similar issue to what @mChowdhury-91 reports. After upgrading pipelines from 2.0.0-alpha7 (kubeflow 1.7) to 2.0.3, I cannot longer create runs from my already uploaded pipelines as the "Start" button is greyed out, check the image:
image

Please note that this is not the case when creating a run using the older Python SDK; so kfp==1.8.22 and a v1 code works just fine.

Also note that cloning an existing run that was created using the Python SDK also works, but trying to clone a run from an uploaded v1 pipeline does not work.

This is what I have observed so far:

SDK Backend Compiled pipeline Create runs from UI Create runs using SDK
1.8.22 2.0.3 v1 No Yes
2.4.0 2.0.3 v2 Yes Yes
2.4.0 2.0.3 v1 No No

I wonder if this should be considered an issue for releasing kubeflow 1.8.1, do you mind confirming @chensun ?

@chensun
Copy link
Member

chensun commented Nov 21, 2023

@jlyaoyuli is going to look into it.

@DnPlas
Copy link

DnPlas commented Nov 29, 2023

Hey @chensun @jlyaoyuli - are there any updates on this one? We think it would be a good idea to include it in a potential patch release of 1.8, but we need confirmation this is going to be worked on before EOW.

@jlyaoyuli
Copy link
Contributor

Thank you everyone for all of the discussion. This issue should be fixed after #10276 is merged.

@chensun
Copy link
Member

chensun commented Dec 8, 2023

Hey @chensun @jlyaoyuli - are there any updates on this one? We think it would be a good idea to include it in a potential patch release of 1.8, but we need confirmation this is going to be worked on before EOW.

@DnPlas we released KFP 2.0.5 today, which contains this fix and another frequent ask: #10229

Please feel free to patch Kubeflow 1.8 release if these two issues would have high impact.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
Status: Closed
Development

Successfully merging a pull request may close this issue.

7 participants