Alerts & Reports#

Superset's Alerts & Reports feature delivers scheduled notifications with rendered chart or dashboard content (PNG screenshots, PDFs, CSV attachments, or inline data tables) via email or Slack.

• Reports run on a cron schedule and always send the content.
• Alerts run a SQL query against a database; the notification fires only when the query result meets a configured threshold (e.g., >, !=, not null). Alert logic lives in superset/commands/report/alert.py.

The feature is disabled by default and must be enabled with the ALERT_REPORTS feature flag 1. Notification channels are Email, Slack (v1), and Slack v2 (ALERT_REPORT_SLACK_V2 feature flag) 2.

Architecture#

The beat schedule runs reports.scheduler every minute. scheduler() fetches all active ReportSchedule records, evaluates their cron expressions via cron_schedule_window(), and calls execute.apply_async() with the computed ETA and optional Celery time limits. Each execute task invokes AsyncExecuteReportScheduleCommand 3, which handles rendering and delivery.

Key files:

Required Components#

Alerts & Reports requires three processes running alongside the web server:

1. Celery worker — executes reports.execute tasks (renders screenshots, sends notifications).
2. Celery beat — runs the reports.scheduler task every minute to queue work.
3. Redis (or RabbitMQ) — production message broker; the default SQLite broker in CeleryConfig is for development only.

In the Docker Compose dev environment, these are the superset-worker and superset-worker-beat services 4. The default CeleryConfig imports superset.tasks.scheduler (which registers all report tasks) plus superset.tasks.thumbnails and superset.tasks.cache 5.

The beat schedule also includes reports.prune_log (daily at midnight) to clean up old ReportExecutionLog records per each schedule's ALERT_REPORTS_DEFAULT_RETENTION policy 6.

Configuration Reference#

All defaults are in superset/config.py. Override via superset_config.py.

Feature Flags#

Alerts & Reports Settings#

SMTP Settings#

Configured at 18:

Slack Settings#

Timeout Summary#

WebDriver / Screenshot Generation#

Screenshots are taken by a headless browser launched inside the Celery worker process. Implementation is in superset/utils/webdriver.py 8.

Key webdriver config keys 8:

The dashboard is loaded in ?standalone=3 (report mode) to hide navigation chrome. Authentication is provided by MachineAuthProvider.get_auth_cookies(), which opens a Flask test_request_context, logs in the executor user, and returns session cookies. The provider class is overridable via MACHINE_AUTH_PROVIDER_CLASS 19.

Screenshots are stored in THUMBNAIL_CACHE_CONFIG. This must be set to a real cache backend (e.g., Redis) — the default is NullCache, which will cause screenshots to fail silently 22.

For CSV reports, the worker makes a server-to-server HTTP call to Superset's ChartDataRestApi (not Selenium). See Report Scheduler Internal HTTP Requests for the full call flow.

Networking and Proxy Considerations#

When Superset runs behind a TLS-terminating reverse proxy, several settings must be coordinated to keep Celery worker calls working 23:

All web server and worker instances must share identical SECRET_KEY and SQLALCHEMY_DATABASE_URI values — mismatches cause silent session decryption failures 25.