Docs for fully on-prem
The doc page for Supported Infrastructure Components says:
You can choose to deploy Metaflow on: ... Any Kubernetes cluster including on-premise deployments.
But that link points to AWS installation docs.
I see #682 about this topic, and the discussion there suggests that, yes, Metaflow can run on fully on-prem infra. But I am having doubts because there are no docs on this subject, and the one mention I can find about this points to an AWS doc.
Where can I learn more about the steps involved for doing an installation onto an on-prem k8s cluster? And are there any limitations if I were to do so?
In case the details matter for my particular case: I have a 10-node on-prem RKE2 cluster, containing a mixture of GPUs and CPU-only nodes and a Ceph storage cluster. Looking for options for MLOps framework to run on this.
+1 to this — We are in a similar situation and would really appreciate clearer guidance on fully on-prem installations. The mention of “any Kubernetes cluster” is encouraging, but it’s a bit confusing when the only linked docs are AWS-specific. Thanks!
I've created PR #2521 to address this documentation gap. The PR includes:
- Comprehensive guide for on-premises Kubernetes deployment
- MinIO setup for S3-compatible storage
- Example flow that works on any K8s cluster
- Specific instructions for RKE2 and Ceph (as mentioned in your use case)
@tpanza @stepanek-petr - Would appreciate your feedback on whether this addresses your needs!