Scaling & High Availability

Architecture

Diagram source

graph TB
    LB[Load Balancer] --> API1[API Server 1]
    LB --> API2[API Server 2]
    LB --> API3[API Server N]
    API1 --> DB[(PostgreSQL Primary)]
    API2 --> DB
    API3 --> DB
    DB --> Replica[(Read Replica)]
    API1 --> Redis[(Redis Cluster)]
    API2 --> Redis
    API3 --> Redis
    API1 --> S3[(Object Storage)]

Horizontal Scaling

API Servers

Run multiple API server instances behind a load balancer:

# docker-compose.yml
services:
  api:
    image: ghcr.io/ever-co/gauzy-api:latest
    deploy:
      replicas: 3

Requirements for Multi-Instance

Database Scaling

Read Replicas

Configure read replicas for read-heavy workloads:

DB_REPLICATION=true
DB_REPLICA_HOST=replica.db.host

Connection Pooling

Use PgBouncer for efficient connection management:

[databases]
gauzy = host=localhost dbname=gauzy

[pgbouncer]
pool_mode = transaction
max_client_conn = 1000
default_pool_size = 40

Redis Clustering

For high availability, use Redis Sentinel or Redis Cluster:

REDIS_HOST=sentinel-host
REDIS_PORT=26379
REDIS_SENTINEL_MASTER=mymaster

Related Pages

• Production Deployment — initial setup
• Monitoring — production monitoring