Deployment Knowledge Base

Quick reference for deployment strategies, zero-downtime patterns, and release management.

Deployment Strategies

Strategy Comparison

Strategy	Downtime	Rollback Speed	Risk	Resource Usage
Recreate	Yes	Slow	High	1x
Rolling	No	Medium	Medium	1.25x
Blue-Green	No	Instant	Low	2x
Canary	No	Fast	Very Low	1.1x
A/B Testing	No	Fast	Low	Variable

┌─────────────────────────────────────────────────────────────────┐
│                    DEPLOYMENT STRATEGIES                        │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  RECREATE         ROLLING          BLUE-GREEN       CANARY      │
│  ┌───┐           ┌───┬───┐        ┌───┐ ┌───┐     ┌───┐        │
│  │ v1│           │v1 │v1 │        │ v1│ │ v2│     │v1 │ 90%    │
│  └───┘           └───┴───┘        └───┘ └───┘     └───┘        │
│    ↓               ↓  ↓              ↕             ┌───┐        │
│  ┌───┐           ┌───┬───┐        Traffic         │v2 │ 10%    │
│  │ v2│           │v2 │v1 │        Switch          └───┘        │
│  └───┘           └───┴───┘                                      │
│                    ↓  ↓                                         │
│                  ┌───┬───┐                                      │
│                  │v2 │v2 │                                      │
│                  └───┴───┘                                      │
└─────────────────────────────────────────────────────────────────┘

Blue-Green Deployment

Overview

Two identical environments (Blue = current, Green = new). Traffic switches instantly.

# Environment structure
environments:
  blue:
    url: blue.example.com
    version: v1.2.3
    active: true

  green:
    url: green.example.com
    version: v1.2.4
    active: false

# Load balancer config
upstream backend {
    server blue.example.com weight=100;
    server green.example.com weight=0;
}

Deployment Steps

#!/bin/bash
# blue-green-deploy.sh

ACTIVE=$(get_active_environment)
INACTIVE=$(get_inactive_environment)

# 1. Deploy to inactive environment
deploy_to_environment $INACTIVE $VERSION

# 2. Run health checks
if ! health_check $INACTIVE; then
    echo "Health check failed, aborting"
    exit 1
fi

# 3. Run smoke tests
if ! smoke_tests $INACTIVE; then
    echo "Smoke tests failed, aborting"
    exit 1
fi

# 4. Switch traffic
switch_traffic_to $INACTIVE

# 5. Verify
if ! verify_deployment $INACTIVE; then
    echo "Verification failed, rolling back"
    switch_traffic_to $ACTIVE
    exit 1
fi

# 6. Mark as active
set_active_environment $INACTIVE

Rollback

# Instant rollback - just switch traffic back
switch_traffic_to $PREVIOUS_ACTIVE

Canary Deployment

Traffic Distribution

# Canary stages
stages:
  - name: canary-5
    traffic: 5%
    duration: 10m

  - name: canary-25
    traffic: 25%
    duration: 30m

  - name: canary-50
    traffic: 50%
    duration: 1h

  - name: full-rollout
    traffic: 100%

Implementation

# nginx canary config
upstream backend {
    server stable.example.com weight=95;
    server canary.example.com weight=5;
}

# Or with cookie-based routing
map $cookie_canary $backend {
    "true"  canary.example.com;
    default stable.example.com;
}

Canary Analysis

# Automated canary analysis
analysis:
  metrics:
    - name: error_rate
      threshold: 1%
      comparison: less_than

    - name: latency_p99
      threshold: 500ms
      comparison: less_than

    - name: success_rate
      threshold: 99%
      comparison: greater_than

  duration: 10m
  interval: 1m

  on_failure: rollback
  on_success: promote

Rolling Deployment

Configuration

# Kubernetes-style rolling update
deployment:
  replicas: 4
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 1
      maxSurge: 1

Sequence

Time →
Pod 1: [v1][v1][v1][v2][v2][v2][v2]
Pod 2: [v1][v1][v1][v1][v2][v2][v2]
Pod 3: [v1][v1][v1][v1][v1][v2][v2]
Pod 4: [v1][v1][v1][v1][v1][v1][v2]

Zero-Downtime Checklist

Database Migrations

// WRONG: Destructive migration
Schema::dropColumn('users', 'old_field');

// RIGHT: Backward-compatible migration
// Step 1: Add new column (deploy #1)
Schema::addColumn('users', 'new_field');

// Step 2: Migrate data (deploy #2)
DB::statement('UPDATE users SET new_field = old_field');

// Step 3: Switch code to use new_field (deploy #3)
// Step 4: Drop old column (deploy #4, weeks later)
Schema::dropColumn('users', 'old_field');

Migration Strategies

Change	Strategy
Add column	Add with default, deploy, backfill
Remove column	Stop using, deploy, wait, remove
Rename column	Add new, migrate, switch, remove old
Change type	Add new column, migrate, switch
Add index	Online DDL, low-traffic window

Health Checks

// Readiness probe - can accept traffic?
public function ready(): JsonResponse
{
    return response()->json([
        'database' => $this->checkDatabase(),
        'cache' => $this->checkCache(),
        'queue' => $this->checkQueue(),
    ]);
}

// Liveness probe - is the app running?
public function live(): JsonResponse
{
    return response()->json(['status' => 'ok']);
}

Feature Flags

Implementation Patterns

// Simple feature flag
if (Feature::enabled('new-checkout')) {
    return $this->newCheckout();
}
return $this->oldCheckout();

// User-based rollout
if (Feature::enabledForUser('new-checkout', $user)) {
    return $this->newCheckout();
}

// Percentage rollout
if (Feature::enabledForPercentage('new-checkout', 10)) {
    return $this->newCheckout();
}

Feature Flag Service

interface FeatureFlagService
{
    public function isEnabled(string $feature): bool;
    public function isEnabledForUser(string $feature, User $user): bool;
    public function isEnabledForPercentage(string $feature, int $percent): bool;
    public function getVariant(string $feature): string;
}

Configuration

# features.yaml
features:
  new-checkout:
    enabled: true
    rollout:
      type: percentage
      value: 25
    users:
      - user-123  # Beta testers
      - user-456

  dark-mode:
    enabled: true
    rollout:
      type: user_attribute
      attribute: plan
      values: [premium, enterprise]

Best Practices

Feature Flag Lifecycle:
┌──────────┐    ┌──────────┐    ┌──────────┐    ┌──────────┐
│  Create  │───▶│  Rollout │───▶│ Full On  │───▶│  Remove  │
│   Flag   │    │  0-100%  │    │   100%   │    │   Flag   │
└──────────┘    └──────────┘    └──────────┘    └──────────┘
     1 day         1-2 weeks       2 weeks         Sprint

Rollback Procedures

Automated Rollback Triggers

rollback:
  triggers:
    - metric: error_rate
      threshold: 5%
      window: 5m

    - metric: latency_p95
      threshold: 2s
      window: 5m

    - metric: health_check_failures
      threshold: 3
      window: 1m

  actions:
    - switch_traffic_to_previous
    - notify_oncall
    - create_incident

Manual Rollback

#!/bin/bash
# rollback.sh

# 1. Get previous version
PREVIOUS=$(get_previous_version)

# 2. Switch traffic immediately (blue-green)
switch_traffic_to $PREVIOUS_ENV

# Or redeploy previous version (rolling)
deploy_version $PREVIOUS

# 3. Verify
health_check_all

# 4. Notify
notify_team "Rolled back to $PREVIOUS"

Database Rollback

// Always have down() migration
public function down(): void
{
    Schema::table('users', function (Blueprint $table) {
        $table->dropColumn('new_field');
    });
}

Environment Configuration

Environment Matrix

environments:
  development:
    replicas: 1
    resources: minimal
    auto_deploy: true

  staging:
    replicas: 2
    resources: medium
    auto_deploy: true
    feature_flags: all_enabled

  production:
    replicas: 4+
    resources: full
    auto_deploy: false
    requires_approval: true
    deployment_window: "Mon-Thu 09:00-16:00"

Secrets Management

# DO NOT: Hardcode secrets
database_password: "secret123"

# DO: Use environment variables
database_password: ${DATABASE_PASSWORD}

# DO: Use secret managers
database_password:
  vault: production/database
  key: password

Deployment Checklist

Pre-Deployment

• All tests passing
• Code review approved
• Database migrations tested
• Rollback plan documented
• Monitoring alerts configured
• Stakeholders notified

During Deployment

• Health checks passing
• No error spike in metrics
• Latency within SLA
• Smoke tests passing
• Feature flags working

Post-Deployment

• Verify functionality
• Check error rates
• Monitor performance
• Update documentation
• Clean up old versions

References

For detailed information, load these reference files:

• references/blue-green.md — Blue-green implementation details
• references/canary.md — Canary release patterns
• references/feature-flags.md — Feature flag best practices