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

Client.wait_for_run_completion and "Terminate" in the UI #3576

Closed
RunOrVeith opened this issue Apr 21, 2020 · 4 comments
Closed

Client.wait_for_run_completion and "Terminate" in the UI #3576

RunOrVeith opened this issue Apr 21, 2020 · 4 comments
Labels
area/backend area/sdk kind/bug priority/p1 status/triaged Whether the issue has been explicitly triaged

Comments

@RunOrVeith
Copy link

What steps did you take:

Start a pipeline via python code, and wait for it's execution:

client = Client()
# fill this with whatever works for you, just start a random pipeline from python that does not return immediately
run = client.run_pipeline(experiment_id="test",
                                        job_name="test job",
                                        pipeline_id="123456")   
 # we don't care how long it takes, just wait
completed_run = client.wait_for_run_completion(run_id=run.id, timeout=float("inf"))  

Once it runs, click "Terminate" inside the kubeflow pipelines UI on the top right.
For me, I hit terminate while the run was stuck on "ImagePullBackoff" (since some permissions in our docker registry changed), so before anything was actually being run, but while the step was in "Pending State".

What happened:

The code is now blocked forever, and does not receive knowledge of the termination.

What did you expect to happen:

The code returns with a message that the run was terminated, or crashes with the same message. Either is fine.

I am not sure if this is the same bug as #1992 and #2588, since this also applies (the run is still in pending state after hitting terminate), nevertheless I wanted to state that it does not trigger "wait_for_run_completion" either.

Environment:

KFP version: Build commit: ee207f2

KFP SDK version: 0.4.0

/kind bug
/area backend
/area sdk

@Bobgy
Copy link
Contributor

Bobgy commented May 7, 2020

@RunOrVeith do you know if this still exists in latest KFP version?
looks like your deployment is very old (Sep.25th 2019)

@Bobgy Bobgy added status/triaged Whether the issue has been explicitly triaged priority/p1 labels May 7, 2020
@Bobgy Bobgy added this to To do in KFP Post 1.0 Backlog via automation May 7, 2020
@RunOrVeith
Copy link
Author

@Bobgy Not sure, we upgraded our deployment last weekend to ca58b22, but I can't manually trigger the ImagePullbackError to check.
If you tell me how to do that I can try it again.

@Bobgy
Copy link
Contributor

Bobgy commented May 7, 2020

@RunOrVeith You can follow doc in https://www.kubeflow.org/docs/pipelines/sdk/build-component/#create-a-python-function-to-wrap-your-component to create a container op with invalid image

@RunOrVeith
Copy link
Author

I just tried it again (or rather I just got stuck in the pending state again), and it seems to work now.
I'll close this issue.

KFP Post 1.0 Backlog automation moved this from To do to Done May 29, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
area/backend area/sdk kind/bug priority/p1 status/triaged Whether the issue has been explicitly triaged
Projects
No open projects
Development

No branches or pull requests

3 participants