Back Up Before Deploying Code

Back Up Before Deploying Code | Generated by AI

Home 2025.12

Question: When deploying new code or packages to production servers, is it a good practice to back up the old files/packages before replacing them?

Answer:

Yes, it is almost always considered a best practice to preserve the old version during deployment so you can roll back quickly and safely if something goes wrong.

Common safe deployment strategies that include keeping/backing up the old version:

Blue-green deployments – You keep the old (blue) environment fully intact while bringing up the new (green) one, then switch traffic. Instant rollback by switching back.
Rolling deployments with versioned directories – Deploy new code to /app/v2025.12.05-01, update symlink /app/current → new version. Old directory remains on disk until you delete it manually after verification.
Immutable artifacts + atomic switches – Build a new container image or package, deploy it alongside the old one, then update the load balancer or service manager to point to the new one. Old image/version stays available.
Simple backup before overwrite – Even in basic setups (scp/ftp/rsync), copy the current application directory to .bak or a timestamped folder first:
```
cp -r /var/www/myapp /var/www/myapp.bak.$(date +%Y%m%d-%H%M%S)
```
or
```
tar czf /backups/myapp-$(date +%Y%m%d-%H%M%S).tgz /var/www/myapp
```

Why you should almost never deploy by blindly overwriting without a rollback path:

A single syntax error or missing dependency can take the entire service down.
Configuration mistakes or incompatible database migrations are common.
Zero-downtime rollbacks are only possible if the old code is still present and runnable.
Compliance/audit requirements in many organizations explicitly require proven rollback capability.

Exceptions where you might skip an explicit backup (but still have rollback):

You are using Kubernetes with proper Deployment rollback (kubectl rollout undo).
You are on platforms like Heroku, Vercel, Netlify, or Cloud Run that keep previous releases automatically.
Git-based deployments (e.g., GitHub Actions deploying via git pull on the server) where you can simply git checkout the previous commit.

Even in those cases, most mature teams still keep at least a few previous artifacts/images for a few days.

Bottom line: Never deploy by permanently deleting or overwriting the currently running version without having a fast, tested way to get back to it. Keeping a backup or old version directory is cheap insurance.

References:

Back

x-ai/grok-4.1-fast

Donate