UBA Superset BI Deployment Guide

This document is aimed at operations engineers and data analysts, describing how to deploy Apache Superset and connect it to GoWind UBA's OLAP engine (Doris by default), building visual dashboards on the fact tables.

Read System Architecture first. UBA and Superset are loosely coupled: Superset runs as an independent BI container, connecting directly to the Doris FE via the MySQL protocol to query the fact tables of the gw_uba database.

1. Deployment Architecture

• UBA writes analytics data into Doris's gw_uba database (events_fact / sessions_fact / risk_events / users_dim etc.).
• Superset connects to Doris FE's port 9030 via the pydoris (or pymysql) driver, modeling these tables as Datasets.

2. One-Click Docker Deployment

Superset is preconfigured in docker-compose.yaml (apache/superset:latest, port 8088). The container automatically completes initialization at startup:

1. Switches the apt sources to Aliyun mirrors and installs build dependencies (gcc python3-dev default-libmysqlclient-dev pkg-config).
2. Enters Superset's real venv /app/.venv and installs pymysql and pydoris (the Doris driver).
3. Runs superset db upgrade → superset fab create-admin (admin/admin) → superset init.
4. Starts run-server.sh.

cd backend
./scripts/docker/full_deploy.sh   # full-stack deployment (includes superset)
# or dependencies only: ./scripts/docker/libs_only.sh (the libs compose also includes superset)

Visit: http://localhost:8088/login/, default account admin / admin.

⚠️ For production, make sure to: change the admin password, replace SUPERSET_SECRET_KEY, and restrict access origins.

3. Connecting to UBA's Doris

In the Superset UI → Settings → Database Connectors → add Apache Doris, fill in the connection string (pointing to the Doris FE query port 9030, database gw_uba):

Key points:

• Superset runs inside a container; use host.docker.internal to reach Doris on the host (or doris-fe:9030 within the same compose network).
• The database name is uniformly gw_uba (consistent with UBA Core's data.yaml doris.database).
• Enable options such as "Allow SQL Lab data upload" as needed when connecting.

Source reference: backend/docs/deploy_superset.md.

4. Creating Datasets & Dashboards

1. Add Datasets

In Superset → Data → Datasets, add UBA's fact/dimension tables as datasets:

2. Configure Field Metadata

For each Dataset, annotate field types, whether it is a time dimension (e.g. event_time / event_date), and whether it is aggregatable on the Columns page, so charts recognize them correctly.

3. Build Charts & Dashboards

Build charts on the Datasets (line trends, funnel steps, retention heatmaps, etc.) and combine them into dashboards. Typical dashboards:

• Operations overview: DAU/MAU trends, channel distribution (channel), platform distribution (platform).
• Conversion funnel: based on event_name sequences from events_fact.
• Risk monitoring: risk type/level distributions and trends from risk_events.
• User profiles: region/level/VIP distributions from users_dim.

The UBA admin console ships with 25 analysis models built in (covering general Web/APP and game scenarios; see the Product Introduction for details); Superset is for more flexible custom BI and reports. The two complement each other.

5. Direct Queries in SQL Lab

You can query directly without creating Datasets. In Superset → SQL Lab → SQL Editor, pick the Doris connection and write SQL directly:

-- DAU per channel over the last 7 days
SELECT channel, event_date, count(DISTINCT user_id) AS dau
FROM gw_uba.events_fact
WHERE event_date >= DATE_SUB(CURDATE(), INTERVAL 7 DAY)
GROUP BY channel, event_date
ORDER BY event_date, dau DESC;

For the dual-engine SQL dialect differences (Doris DATE_FORMAT vs ClickHouse toStartOfHour), see the OLAP Query Cookbook.

6. Common Issues

7. Related Documentation

• System Architecture
• Docker Deployment
• Configuration Reference
• OLAP Query Cookbook
• Data Analyst Getting Started