Taking the site down to deploy belongs to an earlier era. With a few standard techniques, you can ship updates while users keep working and nobody notices a thing.
Roll out gradually
Bring up new instances alongside the old ones, shift traffic over once they are healthy, then retire the old ones. If something is wrong, you shift back without an outage.
Make changes backward compatible
- Deploy database changes in steps that work with old and new code.
- Add new fields before you remove old ones.
- Keep each deploy small so problems are easy to trace.
Automate the safety checks
Health checks and automatic rollback turn a risky release into a routine one. The pipeline should be able to undo a bad deploy without a human in the loop.
DevOps Services Nepaldeploymentreliabilityuptime
Abishek Bimali
Founder & Engineer
Abishek founded SiteCraft Innovation and leads its engineering. He writes about building web and mobile products that hold up in production, for teams in Nepal and abroad.
