8 — MLflow Integration

The operator auto-discovers MLflow on RHOAI/ODH clusters and enables experiment tracking for every FeatureStore deployment. When the MLflow operator is present and healthy, Feast pods receive MLflow configuration automatically — no manual YAML editing required.

---

Auto-discovery

The operator lists all MLflow CRs (mlflow.opendatahub.io/v1) in the cluster and uses the first one with an Available=True or Ready=True condition. When found, it populates tracking_uri from status.address.url and ui_url from status.url.

If the MLflow CR does not report conditions (older operator versions), auto-discovery will not activate. Set trackingUri explicitly in that case.

No MLflow? The FeatureStore stays Ready. Non-MLflow FeatureViews and all other Feast services are completely unaffected.

---

FeatureStore CR configuration

Auto-enabled (default when MLflow is present)

No spec.mlflow needed. The operator auto-enables when an Available MLflow CR is detected:

apiVersion: feast.dev/v1
kind: FeatureStore
metadata:
  name: my-store
spec:
  feastProject: my_project
  services:
    onlineStore: {}
    registry: {}
    ui: {}

Explicit configuration

Override defaults or enable additional features:

apiVersion: feast.dev/v1
kind: FeatureStore
metadata:
  name: my-store
spec:
  feastProject: my_project
  services:
    onlineStore: {}
    registry: {}
    ui: {}
  mlflow:
    enabled: true
    trackingUri: "https://custom-mlflow.example.com:8443"
    uiUrl: "https://dashboard.example.com/mlflow"
    trackingAuth: "kubernetes-namespaced"
    autoLog: true
    autoLogEntityDf: true
    entityDfMaxRows: 50000
    logOperations: true
    opsExperimentSuffix: "-feast-ops"

Opt-out

Disable MLflow even when the MLflow operator is present:

spec:
  mlflow:
    enabled: false

---

Field reference

Field	Type	Default	Description
enabled	bool	auto-detected	Master switch for MLflow integration
trackingUri	string	auto-discovered	MLflow tracking server URI (in-cluster, from status.address.url)
uiUrl	string	auto-discovered	Browser-reachable MLflow URL for Feast UI lineage links (from status.url)
trackingAuth	*string	"kubernetes-namespaced"	Auth method for Feast pods calling MLflow
autoLog	*bool	true	Auto-log feature metadata on every retrieval
autoLogEntityDf	*bool	false	Save entity DataFrame as artifact
entityDfMaxRows	*int32	100000	Skip artifact for large DataFrames
logOperations	*bool	false	Log feast apply / materialize to ops experiment
opsExperimentSuffix	*string	"-feast-ops"	Ops experiment name suffix
extraConfig	map[string]string	—	Additional YAML fields (coerced to native types)

---

Authentication

The operator injects MLFLOW_TRACKING_AUTH into all Feast pod containers. The MLflow Python client's auth plugin system uses this env var to attach credentials to tracking server requests.

trackingAuth value	Behavior
"kubernetes-namespaced" (default)	SA token + X-MLFLOW-WORKSPACE: <namespace> header. Multi-tenant on RHOAI.
"kubernetes"	SA token only. Single-tenant setups.
"basic"	HTTP Basic auth via MLFLOW_TRACKING_USERNAME / MLFLOW_TRACKING_PASSWORD env vars.
"bearer"	Static bearer token from MLFLOW_TRACKING_TOKEN env var.
"" (empty string)	No auth header. Local dev or unprotected MLflow.

No Kubernetes RoleBinding is needed for MLflow tracking API access. The MLflow server validates the SA token directly via TokenReview.

---

Tracking URI resolution order

1. Explicit trackingUri in the FeatureStore CR
2. Auto-discovered from MLflow CR status.address.url (first Available/Ready CR)
3. MLFLOW_TRACKING_URI environment variable (on workbench pods, injected by the MLflow operator)
4. MLflow default (./mlruns)

---

UI URL resolution order

Used for browser hyperlinks in Feast UI lineage panels:

1. Explicit uiUrl in the FeatureStore CR
2. MLFLOW_UI_URL environment variable
3. Auto-discovered from MLflow CR status.url (external gateway route)
4. Falls back to trackingUri (works for local dev)

---

Graceful degradation

Scenario	Behavior
MLflow operator not installed	No mlflow block in YAML; FeatureStore stays Ready
MLflow CR exists but not Ready	Discovery returns empty; MLflow stays off
Tracking URI becomes unreachable	SDK logs a warning; feature retrieval is not blocked
spec.mlflow.enabled: false	MLflow integration explicitly disabled

---

Workbench usage

In a RHOAI workbench notebook connected to the FeatureStore:

from feast import FeatureStore

store = FeatureStore(...)  # from mounted client config

with store.mlflow.start_run(run_name="training"):
    df = store.get_historical_features(
        entity_df=entity_df,
        features=["driver_stats:conv_rate", "driver_stats:acc_rate"],
    ).to_df()
    model = train(df)
    store.mlflow.log_model(model, "model")

Dependency: the Feast image must include feast[mlflow] (mlflow or mlflow-skinny).

---

RBAC permissions

The operator needs get, list, watch on mlflows in the mlflow.opendatahub.io API group. This is included in the default operator ClusterRole.

- apiGroups:
    - mlflow.opendatahub.io
  resources:
    - mlflows
  verbs:
    - get
    - list
    - watch

---

See also

• API field reference — MlflowConfig
• MLflow DataSource reference
• Guide 5 — Security (RBAC / OIDC auth)
• Guide 7 — OpenLineage & Materialization