Gerard Casas Saez

Would be nice to have this in sourcegraph!

@charlesccychen @zhitaoli Seems MLMD is reusing context and registering 2 output artifacts for the same component run instead of overwriting it on the 2nd run (at least that's my intuition)...

To be clear, we are using TFX existing wrapper for Airflow, we dont have any custom code. This fails on current release 0.24.1. The first attempt is succesful, but probably...

Here's a similar issue we ran when an ExampleGen was retried by Airflow. Example artifacts pbtxt in MLMD: ``` id: 3495 type_id: 5 uri: "gs://some-bucket/BigQueryExampleGen/examples/3312" custom_properties { key: "name" value...

Any updates on this? My guess at this point is that on retry artifacts are not overwritten. Aka if I re-run a component with the same run_id and same pipeline_id,...

Steps to repro: - Create adhoc component that fails randomly (ex: assert random(0,1)==1) and gets restarted by Airflow to be completed) - Get a downstream component that has a data...

Out of curiosity @nmelche how do you add retries on KubeflowDagRunner? Using https://github.com/kubeflow/pipelines/blob/master/samples/core/retry/retry.py?

CC @rcrowe-google

@rcrowe-google any chance to tag this to Twitter ?

We have been using views as an alternative yes. But it makes for versioning difficult as one needs to update those separately. Regarding uploading it works, however the issue is...