Floemdata
Data & ETL

Data Warehouse Backup and Restore Basics

A backup proves nothing until you restore it. Backups, restore tests, RPO and RTO for a small-business data warehouse, and what Floemdata does and does not cover.

Floemdata · · 4 min read
A solid L-shaped block, its dashed mirror copy, and an arrow pointing from the copy back to the original

The disk failed and the warehouse file cannot be read. "We have a backup," you say, but you have never restored it. Until a backup has been restored and its figures checked, it is an assumption.

Two terms: RPO and RTO

  • RPO (recovery point objective): how much data can you afford to lose? If the backup runs once a day, the worst case is a day of data.
  • RTO (recovery time objective): how long can the system stay down? Is it acceptable for the reports not to arrive for a day?

Write both down as numbers, together with the business owner. "As fast as possible" is not an objective.

What to look at in a backup

  • What is backed up: the warehouse files, the configuration files, credentials (encrypted and kept separately) and the pipeline's scheduled task definitions.
  • Where to: not the same disk as the warehouse. A separate device or a separate location.
  • How: copying a database file that is in use can produce a broken copy. File-based databases such as SQLite have their own backup interface (SQLite Online Backup API, accessed 2026-11-09). Taking the copy while the pipeline is idle is another safe route.
  • How many versions: if you keep only the latest backup, you have no sound copy to go back to when you notice the data was corrupted earlier.

The restore test

  • Every three months, restore the backup onto a separate computer.
  • Open a report from the restored warehouse; compare the row count and total amount with the live system.
  • Write down how long it took. That is your real RTO.
  • Put the steps in a document that someone else can follow on recovery day.

Does rebuilding from the source replace a backup?

Partly. If the source system is still available and returns history unchanged, the warehouse can be rebuilt; but the earlier state of records that were later changed or deleted in the source is gone. A warehouse that keeps a raw layer carries a smaller version of this risk.

Where Floemdata stands

According to our project documents, the warehouse is kept in SQLite database files (for example Gold: merkez.gold.db), and the Bronze layer keeps raw versions so data can be reprocessed. We did not find a feature in the code that backs up or restores; taking the backup is a task on the customer side. We have not tried rebuilding the warehouse from Bronze as a restore scenario.

To start, copy your warehouse files to a separate computer this week and try to open them. Questions are welcome at info@floemdata.com or via Floemdata.

← All guides