What Is a Rollback?

A rollback is the act of reverting your software to a previous known-good version after a deployment goes wrong.

No matter how good your testing is, some bad releases reach production. A rollback is your escape hatch: instead of scrambling to fix the new version live, you quickly restore the last version that worked, then investigate calmly.

What it is

A rollback returns a system to a prior, working state, usually by redeploying the previous build artifact or switching traffic back to it. The key idea is recovery by reversion: you do not debug under fire, you restore known-good behavior first and diagnose afterward.

How it works

Because earlier releases are versioned, stored artifacts, rolling back can be as simple as redeploying the previous one. With blue-green deployment, the old version is still running, so a rollback is just flipping traffic back, near-instant. Database changes complicate things, which is why migrations are designed to be backward compatible.

An example

A new release ships and error rates immediately climb. Rather than trying to hotfix it live, the on-call engineer rolls back to the previous artifact with one command. Within a minute the system is healthy again on the old version, and the team investigates the bug without the pressure of an active outage.

Why fast rollback matters

The faster you can roll back, the less a bad release costs, and the more confidently you can deploy in the first place. A reliable, quick rollback path is what makes frequent deployment safe: if any release can be undone in seconds, shipping often stops being scary.

What makes rollback reliable

• Keep previous artifacts versioned and ready to redeploy.
• Design database migrations to be backward compatible.
• Use deployment strategies that keep the old version available.
• Practice rollback so it works when you need it.

Key takeaways

• A rollback restores a previous known-good version after a bad deploy.
• Versioned artifacts and blue-green deploys make it fast.
• Fast rollback is what makes frequent deployment safe.

Related guides