01 · Public app · Personal lab · 2025
Multi-DB Studio
One Java workbench for six database engines.
I use it to practice diagnostics, maintenance and moving a table between engines that do not share a type system. A reviewer can open the public app and try the explorer, the console and a connection test without me in the room.
What you can do in it
Pick a part of the application. The note says what it is for, what you should see, and one thing worth trying.
What it does
Connections
Proves the workbench can reach an engine before anyone trusts a query result.
- On the screen
- A connection entry per engine, and a test that reports whether that engine answered.
- Try this
- Open the app and run the connection test on PostgreSQL, then on a second engine. The two PostgreSQL deployments are separate: one on K3s, one on MicroK8s.
Maintenance by engine
The command changes with the engine
Pick an engine. This is the operation Multi-DB Studio is built to expose, not a live reading from the lab.
PostgreSQL
VACUUM ANALYZE
Reclaims space from dead tuples and refreshes the planner’s statistics. The studio exposes this on the PostgreSQL deployments, one on K3s and one on MicroK8s.
What the application is
Multi-DB Studio is a Java 17 workbench. PostgreSQL runs twice, once on K3s and once on MicroK8s. MySQL 8.4, SQL Server 2022, MongoDB 8 and Apache Geode sit beside them. The public site is published through a Cloudflare Tunnel. The Kubernetes manifests carry health probes and resource limits, and a scheduled job backs up MongoDB onto persistent storage.
At The Methodical Group I already ran SQL Server, PostgreSQL and MySQL side by side. This lab asks a narrower question: can I operate those engines, plus a document store and an in-memory grid, from one place, and can I explain what changes when a row moves between them?
How the pieces connect
A browser reaches the workbench through the tunnel. The workbench talks to each engine with its own driver. PostgreSQL is intentionally two clusters, so a mistake on K3s does not pretend to be a result from MicroK8s. Geode is not behind JDBC in the same way as the SQL engines: its screen is OQL and region operations.
What this does not prove
This is my lab, not a client system. There is no published before-and-after for a production workload, and the source repository is private because older repos on that GitHub account contain secrets. The useful test is whether the operations on the screen are the ones a DBA actually reaches for.