Databricks Release Notes Dashboard

Databricks Asset Bundle that ingests Databricks release notes for AWS, Azure, and GCP daily, which power the Databricks Release Notes Tracker dashboard.

Architecture

RSS Feeds (AWS, Azure, GCP) ─► Notebook (MERGE) ─► Delta Table ─► Dashboard

Resource	Description
Job	Daily Databricks Release Notes ETL, runs at 06:00 SGT
Table	<catalog>.<schema>.release_notes
Dashboard	Databricks Release Notes Tracker

Bundle structure

dbx-release-notes-dashboard/
├── databricks.yml                        # Bundle config, variables, targets
├── resources/
│   ├── release_notes_job.yml             # Job: schedule, timeout
│   └── release_notes_dashboard.yml       # Dashboard: warehouse, catalog/schema
└── src/
    ├── release_notes_etl.ipynb           # ETL notebook
    └── release_notes_tracker.lvdash.json # Dashboard definition

Variables

All variables without defaults are required and must be set at deploy time.

Variable	Default	Description
catalog	(required)	Unity Catalog catalog
schema	databricks_release_notes	Unity Catalog schema (created automatically if it doesn't exist)
warehouse_id	(required)	SQL warehouse for the dashboard

Quickstart

# Validate
databricks bundle validate

# Deploy to dev (prefixed, schedule paused)
databricks bundle deploy -t dev \
  --var catalog=my_catalog \
  --var warehouse_id=abc123def456

# Deploy to prod
databricks bundle deploy -t prod \
  --var catalog=prod_catalog \
  --var warehouse_id=abc123def456

# Trigger a run
databricks bundle run -t dev release_notes_etl

Optional: SQL alert for new release notes

You can create a Databricks SQL alert to get notified when new release notes are ingested:

SELECT count(*) AS new_notes
FROM <catalog>.<schema>.release_notes
WHERE ingested_at >= current_timestamp() - INTERVAL 1 DAY

Set the alert to trigger when new_notes > 0 and configure your preferred notification destination (email, Slack, webhook). The query can be further customized focus on specific release note events.

Key features

• Content-hash MERGE: only updates rows whose content actually changed; reports inserted / updated / untouched counts each run
• HTTP retry: exponential backoff on 429 / 5xx errors
• Dev / Prod targets: mode: development auto-prefixes job name and pauses schedule
• Dashboard parameterisation: dataset_catalog and dataset_schema swap per target
• Parameterised notebook: catalog and schema passed as widget parameters from the job
• Auto-provisioning: target schema is created automatically if it doesn't exist