Second conda question: I am distributing internal libraries as conda packages, through an internal conda channel. This channel is protected using BasicHTTPAuth, so it takes credentials to pull a package. What I found on the doc is the ability to use the conda syntax

channel::package

to specify a channel, however it implies this info is written in plain text in the workflow file. In my case it would be

<https://login:password@private_channel::package|https://login:password@private_channel::package>

, which would require exposing secrets, which you can understand is a problem to me. Is there a way to leverage a secret there? Or do you know is it is possible to make a custom value for the

CONDA_CHANNELS

env var available to conda/mamba when creating the environment? This env var has the highest priority to specify channels in the conda ecosystem, so if I can have it be a secret and be available at the right time, there's a chance it could work.

If you use the bleeding edge version of the decorators, you can set this here: https://github.com/Netflix/metaflow-nflx-extensions/blob/main/metaflow_extensions/netflix_ext/config/mfextinit_netflixext.py#L115

@elegant-lamp-40213 correct - you can just set CONDA_CHANNELS env var outside of metaflow and metaflow will automatically respect it

I am using Bitnami's SealedSecret in my K8s cluster. So it is treated like standard kubernetes secrets and managed by a controller at the cluster level.

We don’t pull down packages from the conda channel inside the kubernetes pod when you are running a step with @kubernetes

Ok, I believe I tested that. I will test again, then. Thank you for pointing the direction 🙂

You can put me.

I can only add maintainers through a GitHub handle, emails can't be used in this case.

Ah. Ok so me. Romain-Intel is my handle.

Fyi @dry-beach-38304,

metaflow-netflixext

is now available through conda-forge. This package is created from the sources published with the Pypi package, so nothing to change on your release process. One of the conda-forge bots will detect new versions released and open a PR to reflect changes on the conda recipe. I also enabled an automerge feature, so in case where the new recipe successfully builds on the new version branch, the PR would be merged by the bot. This may fail if dependencies need to be updated.

Thank you!!