The human-centric platform for production ML & AI

Outerbounds

<https://outerbounds-community.slack.com/archives/C02116BBNTU/p1676013974756079?thread_ts=1675989837.483879&amp;cid=C02116BBNTU>   <@U02175E3M0F>, am curious if it’s possible for you to share how do you manage the Metaflow datastore (I.E S3 bucket), Do you have all the clients using the one s3 bucket ?

<@U02559V97KJ>, users use their own s3 buckets.  They register their access credentials as k8s secrets in their k8s namespace and use
```--with=kubernetes:secrets=my-s3-secret```

ohh I forgot there is even a config parameter which makes cli simpler:
<https://github.com/Netflix/metaflow/blob/056b556a2b59e95751928e8fbefdd0dac545b2ca/metaflow/metaflow_config.py#L250>

oh, thanks for the input that's helpful.  I'm exploring around this area of managing "clients" through isolating various resources.
i was going down the path of having multiple nodegroups per client and each nodegroups is tied to an IAM with specific access to S3.

Lots of researching in around kubernetes ( :exploding_head: )