MLflow Tracking Server Deployment Architecture on EKS

MLflow Tracking Server Deployment Architecture on EKS An architecture diagram generated by Archify. SDK / UI client · Client SDK · Web UI · Architecture component SDK / UI client Client SDK · Web UI Protected entry · TLS / ingress policy · Architecture component Protected entry TLS / ingress policy Kubernetes Service · Load-balances across Pods · EKS cluster Kubernetes Service Load-balances across Pods Tracking Server Pod · mlflow server · replica 1 · EKS cluster Tracking Server Pod mlflow server · replica 1 ServiceAccount · S3 workload identity · EKS cluster ServiceAccount S3 workload identity Tracking Server Pod · mlflow server · replica 2 · EKS cluster Tracking Server Pod mlflow server · replica 2 PostgreSQL · metadata / auth DBs · Architecture component PostgreSQL metadata / auth DBs S3 Bucket · Artifact store · Architecture component S3 Bucket Artifact store HTTPS grants S3 access DB login + TLS artifacts EKS cluster Legend Backend Database Cloud Security External

Verify shared state

  • • Tracking/auth DBs, session secrets, enabled queues/caches
  • • Proposed replica layout; validate shared state before scaling

Permission boundaries

  • • Configure S3 IAM and DB login/TLS separately
  • • ClusterIP, TLS, and host checks do not replace user authorization