Hello! we are still evaluating metaflow and like what we see so far. We are on GCP (GKE) with Argo Workflows. Most of our tutorials runs went great, but 08 is not working as we would hope. Everything seems to be running fine until the end where it errors out with that message:

moviestatsflow-sjrqn-877611774:
      id: moviestatsflow-sjrqn-877611774
      name: moviestatsflow-sjrqn.end
      displayName: end
      type: Skipped
      templateName: end
      templateScope: local/
      phase: Error
      boundaryID: moviestatsflow-sjrqn
      message: >-
        unable to evaluate expression '(join.Succeeded || join.Skipped ||
        join.Daemoned)': unable to evaluate expression '(join.Succeeded ||
        join.Skipped || join.Daemoned)': type func(...interface {}) (interface
        {}, error)[string] is undefined (1:7)
         | (join.Succeeded || join.Skipped || join.Daemoned)
         | ......^
      startedAt: '2024-01-19T19:27:32Z'
      finishedAt: '2024-01-19T19:27:32Z'
      nodeFlag: {}

Any idea on what we could check to get passed this?

what is the version of argo-workflows that you are running?

3.5.2

Hey there @hundreds-midnight-75494. Did you manage to fix this issue? I'm in the same boat as yours, been trying to find the root cause for such error for few days already.

No resolution from my side

do you run into the same issue with Argo 3.4?

Hey there @square-wire-39606. Thank you very much 🙏 Lowering the version down to v3.4.0 the issue. Do you happen to know which 3.5* change breaks this? I'll go over the release notes to see if I can find something

not yet, but we are investigating

I can confirm that the flow worked after downgrading to 3.4.14

looking into this, will get to testing once I have an upgraded argo deployment to fiddle with. In the mean time, are either of you using the

notify-on-error/notify-on-success

feature when deploying the flows? At first glance the error might be related to the lifecycle hooks that we use, which get added in case either of the notifies are enabled.

I am literally using the tutorials example for this, not sure tbh, my role is more around the deployment and configuration of the environments themselves, I only tested the basic flows to validate the environment I created worked with the tutorials and basic flows

whoosh, finally found out the issue here: https://github.com/argoproj/argo-workflows/issues/12037 the quick workaround is to have the step in the flow not be named

join

(or any other of the problematic words listed in the issue)

oh wow! thank you for that!

Thanks all for the help !!

The issue above seems to have a fix in 3.5.5, hopefully that solves it 🙂