Skip to main content
Data Fabric keeps a copy of your tracing project data in a cloud storage bucket you own, so you can query traces in your warehouse alongside the rest of your data. Arize AX writes to Google Cloud Storage, Amazon S3, or Azure Blob Storage as Apache Iceberg or Delta Lake tables, which BigQuery, Databricks Unity Catalog, and Snowflake can read. Data Fabric is available on Enterprise accounts.

How sync works

  • You need write access to the target bucket and at least one tracing project in your space.
  • A connector points at one file path in your bucket and syncs any number of projects.
  • Sync runs every 60 minutes. The interval cannot be changed in the UI. The first sync includes all historical data.
  • Project syncs can be paused, resumed, or deleted. Deleting a synced project stops its sync and leaves written data in your storage.
  • Tables land at <path>/<namespace>/<project_name>/. The current Iceberg metadata file is <namespace>/<project_name>/metadata/latest.metadata.json.

Create a connector

  1. Go to Settings → Space Settings → Data Fabric in your space and click New Connector.
  2. Enter a Connector Name, then Select Project(s) to sync.
  3. Choose the Table Format: Apache Iceberg or Delta Lake.
  4. Choose the Select Data Storage provider, then enter the File Path to Write Data To.
Enter the bucket and prefix, for example my-data-bucket/arize-sync/production.

Set up permissions

Each bucket needs a tag, label, or metadata value set to the ingestion key shown in the setup dialog. Arize uses it to confirm that you own the bucket. The key name is arize-ingestion-key on Google Cloud Storage and Amazon S3, and arize_ingestion_key on Azure.
  1. Add a bucket label with key arize-ingestion-key and the value from the setup dialog.
  2. Create the custom IAM role from the setup dialog:
  1. Grant the role to the Arize service account on your bucket:
Click Validate, then Start Job. The first sync starts immediately. Who can read the synced data is covered in Access control.

Set up query tables

After the first sync finishes, create one external table per synced project.