Skip to main content
Once Data Fabric writes traces to your bucket, Arize AX roles no longer decide who can read them. Arize AX roles control who can read, create, update, and delete connectors in a space, while your cloud provider controls access to the data itself. Anyone with read access to the bucket can read everything written to it, so keep your cloud policies in step with your Arize AX roles or a member could read traces from a space they cannot open in Arize. Each bucket must also carry the arize-ingestion-key tag, label, or metadata value from the connector setup dialog. See Data Fabric.

One prefix per space

Each connector is scoped to one Arize AX space. Give each space its own bucket or prefix, and scope IAM policies to it.

Reuse identity provider groups

SAML SSO role mappings map identity provider groups to Arize AX spaces and roles. The same groups can grant cloud storage access, so one membership controls both.
  1. Define a group per team in your identity provider, for example team-alpha.
  2. Map the group to the team’s space with a SAML role mapping.
  3. Map the same group to a cloud IAM policy that reads only your-bucket/team-alpha/*.

Prefix-scoped IAM policies

The Arize arize-data-fabric role or service account needs write access across all prefixes. These policies are for people and analytics tools that read only their team’s prefix.
With AWS IAM Identity Center, create a permission set per team prefix and assign it to the matching identity provider group.

Query-time controls

Warehouses can restrict a user who already has bucket read access.
  • AWS Lake Formation enforces column-level and row-level permissions for Athena and Redshift Spectrum.
  • BigQuery IAM controls access to external tables independently of the GCS bucket.
  • Snowflake RBAC restricts which roles can query each external table.