FOR SRE AND PLATFORM

Rollback isn’ta rebuild.

Production is a pointer to a build that is already deployed. Rolling back is moving the pointer back. No rebuild, no re-upload. Friday deploys are fine.

Book a demo Start free →

Friday, 14:00. Error rate, checkout.

time to recover: 4 min

One move. #1289 was still deployed.

14:02 → #1290Release introduces an errorIllustrative release

14:06 → back to #1289One move. #1289 was still deployed.Previously deployed version

Illustrative incident. Recovery time depends on how fast your team detects the problem.

No rebuildthe last good build is still live.

No re-uploadnothing to push again.

Rolesdecide who can move production.

REBUILD VS. REPOINT

The fastest rollback is one you never build.

Rollback by rebuild

Find the last good commit
Revert, wait for CI
Build it again, hope it matches
Upload and deploy, under pressure

recovery = full pipelineMinutes to hours

Rollback on Zephyr

Point production at the last good build
There is no step 2

recovery = one moveSeconds

Repoint, don’t rebuild

Every build stays deployed. Rollback moves the production tag to a build you already ran.

Ship what you tested

Release promotes the exact build you reviewed, not a rebuild of the same commit.

Guard production

Protected environments and role-based permissions decide who can release and who can roll back.

HOW IT WORKS

Deploy is not release.

01Every build deploys

Each build is uploaded once and kept live at its own URL.

02Release moves a tag

production points at one build. The change takes effect without a rebuild.

03Rollback moves it back

Same move, opposite direction. The old build never left.

BOOK A DEMO

Watch us break prodand roll it back live.

30 minutes with an engineer. We ship a bad build on purpose and recover it on the call.

Release and rollback without a rebuild
Protected environments and roles
How it fits your incident process