The human-centric platform for production ML & AI

Outerbounds

I'm curious if there is any documentation available for setting up an on-premise datastore/metadata with Metaflow using MinIO? I'm thinking about experimenting with a PoC setup like the one described at this issue which includes Kubernetes as well: 
<https://github.com/Netflix/metaflow/issues/682>
Nothing production-level at the moment ... entirely PoC.

We don't have any official docs but it should work. After you set up Minio you'd need to make Metaflow use a custom S3 API endpoint (served by Minio), you can do this by setting `METAFLOW_S3_ENDPOINT_URL` and likely you'd need `METAFLOW_S3_VERIFY_CERTIFICATE` too (<https://github.com/Netflix/metaflow/blob/master/metaflow/metaflow_config.py#L88-L89|here> )

You'll also need to configure S3 credentials for steps running on K8S, you can do this if you store creds in K8S secret and then use it in kubernetes decorator like this: `@kubernetes(secrets=["my_minio_secret"])`

Thanks so much <@U01P6EGHFDW>! I think this is enough to get me started. I appreciate it!