Zero-Downtime Deployment Strategies

Modern systems require deployments that don't cause service interruptions or user impact.

Rolling Deployment

Gradually replace old instances with new ones.

Start:  [v1][v1][v1][v1]
Step 1: [v2][v1][v1][v1]
Step 2: [v2][v2][v1][v1]
Step 3: [v2][v2][v2][v1]
Done:   [v2][v2][v2][v2]

Pros: Simple, resource-efficient Cons: Mixed versions during rollout

Blue-Green Deployment

Maintain two identical environments; switch traffic instantly.

Blue (v1): Live traffic → 100%
Green (v2): Deploy here → Test → Switch LB

After switch:
Blue (v1): 0% (standby for rollback)
Green (v2): 100%

Pros: Instant rollback Cons: 2x infrastructure cost

Canary Release

Route small percentage of traffic to new version.

v1: 95% of users
v2: 5% of users (canary)
↓
Monitor errors, latency, business metrics
↓
If healthy: gradually increase to 100%
If issues: rollback canary

Feature Flags

Deploy code without activating features.

if (featureFlag.isEnabled('new_checkout', user)) {
  renderNewCheckout();
} else {
  renderOldCheckout();
}

Tools: LaunchDarkly, Unleash, Flagsmith.

Database Migrations

Zero-downtime DB changes:

Step 1: Add new column (nullable, backward compatible)
Step 2: Deploy new code (writes to both old and new)
Step 3: Backfill data
Step 4: Deploy code reading new column
Step 5: Remove old column

Conclusion

Canary + feature flags is the modern standard. Separate code deployment from feature activation for safe, gradual rollouts.