bug: Argo integration - labels on original workflows crd are not synced

kubestellar / galaxy

Additional modules, tools and documentation to facilitate KubeStellar integration with other community projects

Apache License 2.0

0 stars 3 forks source link

bug: Argo integration - labels on original workflows crd are not synced #21

Open nicolas-vivot opened 3 months ago

nicolas-vivot commented 3 months ago

Describe the bug

Running a few given examples, after a workflow is completed, the correct status is not reflected on the UI.

The main reason is the labels on the original workflows from the main cluster are not updated.

Consequences:

incorrect / misleading information
filters on the UI are broken (using the phase)

Labels on the main cluster:

Labels on the target execution cluster (cluster1 in this case):

Steps To Reproduce

Run the any sample workflows coming with the demonstration
Checkout in Argo UI after completion

Expected Behavior

Information should be synced.

Additional Context

No response

nicolas-vivot commented 3 months ago

Other than this bug, i found this initiative quite interesting, but i wonder how serious this is ?

is it a use case officially supported by Argo Workflow ?
will it be in the future, are there any plans or is this project "just" a showcase ?

pdettori commented 1 month ago

Hi @nicolas-vivot thank you for your feedback, this is quite useful. At this time there is no direct support in KubeStellar to sync back labels from the execution clusters to the main cluster, so I think that potential solutions at this time may be based on an additional controller on the main cluster that updates those labels by observing the status in the Argo workflow object.

As for your questions:

is it a use case officially supported by Argo Workflow ?

AFAIK, no, but our KubeStellar community would like to gauge the interest of the Argo community in this Multi-Cluster use case.

will it be in the future, are there any plans or is this project "just" a showcase ?

At this time, this integration is just a showcase of capabilities for KubeStellar, but this may change if there is enough interest in the community for this use case.

scottrigby commented 1 month ago

At this time, this integration is just a showcase of capabilities for KubeStellar, but this may change if there is enough interest in the community for this use case.

@pdettori where's the best place for community to express interest in this use case? Would you want to create a poll in https://github.com/kubestellar/kubestellar/discussions and point people there for feedback?

pdettori commented 1 month ago

@scottrigby yes, that could be a good starting point, thx