Yuan Tang
There's also discussion around donating Spark-on-K8s project to Kubeflow (no open issue yet since we are still waiting for governance update). I personally think that project is similar to MPI...
Seems like the issue might be that `/root/manager` was not built successfully or not copied successfully to the next phase https://github.com/kubeflow/xgboost-operator/blob/78f8cf50bb943247e038a8feb5a9f7e47d810d65/Dockerfile#L19. Are you able to build the Docker image successfully?
Yes, I think MPI controller is doing it correctly.
I am a little concerned about the fact that these are still experimental. We may expect the removal of them at any point of time.
> @terrytangyuan I am testing the new implementation right now, looks like there is a RunContainerError, should we update the image version in our deployment config file from v0.1.0 to...
Hmm this is XGBoost operator though. We should probably aim to provide better support and experience for XGBoost users. I don’t think LightGBM would need to be supported here.
Related issues https://github.com/argoproj/argo-workflows/issues/7646, https://github.com/argoproj/argo-workflows/issues/9906, https://github.com/argoproj/argo-workflows/issues/10107, https://github.com/argoproj/argo-workflows/issues/10124, https://github.com/argoproj/argo-workflows/issues/11250
> Actually correct me if I am wrong about Postgres| MySql not being mandatory, if it is mandatory, makes more sense to just use that. It's only required when persistence/workflow...
Thanks for letting us know. We only use it in Java for now.