Multi-Region Deployment

Deploy YugabyteDB across multiple geographic regions to build globally distributed applications with low latency and high availability.

Overview

YugabyteDB supports two primary multi-region deployment configurations:

3+ Data Center (3DC) Deployment: Single cluster stretched across 3+ regions with synchronous replication
xCluster Replication: Independent clusters with asynchronous replication

3DC Deployment (Synchronous Replication)

A single YugabyteDB universe deployed across 3 or more data centers with synchronous replication using Raft consensus.

Architecture

Data consistency: Global strong consistency
Replication: Synchronous via Raft
Failover: Automatic
Write latency: Affected by WAN latency
Use cases: Applications requiring strong consistency across regions

Prerequisites

3+ regions (or data centers)
Low-latency network between regions (less than 100ms RTT recommended)
RF ≥ 3 (typically RF=3 or RF=5)
Stable WAN connectivity

Deploy 3DC Cluster with yugabyted

Start first node in Region 1

# Node in us-east-1
./bin/yugabyted start \
  --advertise_address=10.1.1.11 \
  --cloud_location=aws.us-east-1.us-east-1a \
  --fault_tolerance=region

Note: Use --fault_tolerance=region for multi-region deployments.

Start first node in Region 2

# Node in us-west-2
./bin/yugabyted start \
  --advertise_address=10.2.1.12 \
  --join=10.1.1.11 \
  --cloud_location=aws.us-west-2.us-west-2a \
  --fault_tolerance=region

Start first node in Region 3

# Node in eu-west-1
./bin/yugabyted start \
  --advertise_address=10.3.1.13 \
  --join=10.1.1.11 \
  --cloud_location=aws.eu-west-1.eu-west-1a \
  --fault_tolerance=region

Add additional nodes in each region

For a 9-node cluster (3 nodes per region):

# Region 1 - additional nodes
./bin/yugabyted start --advertise_address=10.1.2.14 \
  --join=10.1.1.11 --cloud_location=aws.us-east-1.us-east-1b \
  --fault_tolerance=region

./bin/yugabyted start --advertise_address=10.1.3.15 \
  --join=10.1.1.11 --cloud_location=aws.us-east-1.us-east-1c \
  --fault_tolerance=region

# Repeat for other regions...

Configure placement policy

Set replica placement across regions:

./bin/yb-admin -master_addresses 10.1.1.11:7100,10.2.1.12:7100,10.3.1.13:7100 \
  modify_placement_info \
  aws.us-east-1.us-east-1a,aws.us-west-2.us-west-2a,aws.eu-west-1.eu-west-1a 3

Preferred Region Configuration

Optimize read latency by setting a preferred region:

-- Set preferred region for a tablespace
CREATE TABLESPACE us_east_ts WITH (
  replica_placement='{"num_replicas": 3, "placement_blocks": [
    {"cloud":"aws","region":"us-east-1","zone":"us-east-1a","min_num_replicas":1,"leader_preference":1},
    {"cloud":"aws","region":"us-west-2","zone":"us-west-2a","min_num_replicas":1},
    {"cloud":"aws","region":"eu-west-1","zone":"eu-west-1a","min_num_replicas":1}
  ]}'
);

-- Create table in the tablespace
CREATE TABLE users (id INT PRIMARY KEY, name TEXT) TABLESPACE us_east_ts;

Geo-Partitioning

Partition data by geography for lower latency:

-- Create partitioned table
CREATE TABLE orders (
  order_id INT,
  user_id INT,
  region TEXT,
  order_date DATE,
  PRIMARY KEY (region, order_id)
) PARTITION BY LIST (region);

-- Create partition for US in us-east tablespace
CREATE TABLE orders_us PARTITION OF orders
  FOR VALUES IN ('US') TABLESPACE us_east_ts;

-- Create partition for EU in eu-west tablespace
CREATE TABLE orders_eu PARTITION OF orders  
  FOR VALUES IN ('EU') TABLESPACE eu_west_ts;

-- Create partition for APAC in ap-south tablespace
CREATE TABLE orders_apac PARTITION OF orders
  FOR VALUES IN ('APAC') TABLESPACE ap_south_ts;

xCluster Replication (Asynchronous)

Two or more independent YugabyteDB universes connected via change data capture (CDC).

Architecture

Data consistency: Eventually consistent
Replication: Asynchronous
Failover: Manual (application-managed)
Write latency: No WAN impact
Use cases: Disaster recovery, active-active applications

Deployment Types

Unidirectional Replication

One-way replication from primary to standby:

Primary (us-east) ──────> Standby (us-west)

Bidirectional Replication

Two-way replication for active-active:

Cluster A (us-east) <────> Cluster B (eu-west)

Set Up xCluster Replication

Deploy two independent clusters

Cluster 1 (us-east):

./bin/yugabyted start --advertise_address=10.1.1.11
./bin/yugabyted start --advertise_address=10.1.1.12 --join=10.1.1.11
./bin/yugabyted start --advertise_address=10.1.1.13 --join=10.1.1.11

Cluster 2 (eu-west):

./bin/yugabyted start --advertise_address=10.2.1.11
./bin/yugabyted start --advertise_address=10.2.1.12 --join=10.2.1.11  
./bin/yugabyted start --advertise_address=10.2.1.13 --join=10.2.1.11

Create identical schemas on both clusters

On both clusters:

CREATE DATABASE myapp;

\c myapp

CREATE TABLE users (
  id SERIAL PRIMARY KEY,
  name TEXT,
  email TEXT
);

Set up replication from Cluster 1 to Cluster 2

./bin/yb-admin -master_addresses 10.1.1.11:7100 \
  setup_universe_replication \
  cluster2_replication \
  10.2.1.11:7100,10.2.1.12:7100,10.2.1.13:7100 \
  myapp.users

(Optional) Set up bidirectional replication

For active-active, replicate from Cluster 2 to Cluster 1:

./bin/yb-admin -master_addresses 10.2.1.11:7100 \
  setup_universe_replication \
  cluster1_replication \
  10.1.1.11:7100,10.1.1.12:7100,10.1.1.13:7100 \
  myapp.users

Verify replication status

./bin/yb-admin -master_addresses 10.1.1.11:7100 \
  get_replication_status

Monitor Replication Lag

# Check replication lag
./bin/yb-admin -master_addresses 10.1.1.11:7100 \
  get_replication_lag

Acceptable lag depends on your use case:

< 1 second: Real-time applications
< 5 seconds: Most business applications
< 60 seconds: Analytics/reporting

Read Replicas

Deploy read-only replicas in remote regions for low-latency reads.

Deploy Read Replica Cluster

./bin/yb-admin -master_addresses 10.1.1.11:7100 \
  add_read_replica_placement_info \
  aws.ap-south-1.ap-south-1a,aws.ap-south-1.ap-south-1b,aws.ap-south-1.ap-south-1c 3

Configure Timeline-Consistent Reads

-- Set session to read from replicas
SET yb_read_from_followers = true;
SET yb_follower_read_staleness_ms = 30000;  -- 30 seconds

SELECT * FROM users WHERE region = 'APAC';

Multi-Cloud Deployment

Deploy across multiple cloud providers for maximum resilience.

# AWS node
./bin/yugabyted start \
  --advertise_address=10.1.1.11 \
  --cloud_location=aws.us-east-1.us-east-1a \
  --fault_tolerance=cloud

# GCP node
./bin/yugabyted start \
  --advertise_address=10.2.1.12 \
  --join=10.1.1.11 \
  --cloud_location=gcp.us-central1.us-central1-a \
  --fault_tolerance=cloud

# Azure node  
./bin/yugabyted start \
  --advertise_address=10.3.1.13 \
  --join=10.1.1.11 \
  --cloud_location=azure.eastus.eastus-1 \
  --fault_tolerance=cloud

Network Configuration

VPN/VPC Peering

Set up secure network connectivity: AWS VPC Peering:

aws ec2 create-vpc-peering-connection \
  --vpc-id vpc-11111111 \
  --peer-vpc-id vpc-22222222 \
  --peer-region us-west-2

GCP VPC Peering:

gcloud compute networks peerings create peer-gcp-to-aws \
  --network=yugabyte-network \
  --peer-network=aws-network

Firewall Rules

Open required ports between regions:

# Allow YB-Master RPC
7100/tcp

# Allow YB-TServer RPC  
9100/tcp

# Allow client connections
5433/tcp (YSQL)
9042/tcp (YCQL)

Performance Optimization

Minimize Cross-Region Traffic

Use geo-partitioning to keep data local
Set preferred regions for leader placement
Use follower reads for read-heavy workloads

Optimize Network Latency

Choose regions with low inter-region latency
Use cloud provider backbone networks
Enable compression for WAN traffic:

./bin/yugabyted start \
  --tserver_flags="stream_compression_algo=3"  # LZ4 compression

Monitor WAN Metrics

Key metrics:

Round-trip time (RTT) between regions
Packet loss rate
Bandwidth utilization
Replication lag

Disaster Recovery

Backup to Multiple Regions

# Backup to S3 in multiple regions
./bin/yb-admin create_snapshot ysql.mydb

# Copy backup to remote region
aws s3 sync s3://backups-us-east/ s3://backups-eu-west/ \
  --source-region us-east-1 \
  --region eu-west-1

Failover Procedures

For 3DC deployment:

Automatic failover if region becomes unavailable
No manual intervention required
Cluster remains available with RF-2 nodes

For xCluster:

Update application connection strings
Point to standby cluster
Verify data consistency
Switch xCluster direction when primary recovers

Best Practices

Choose the Right Deployment Model

3DC: Use for strong consistency requirements
xCluster: Use for disaster recovery or when WAN latency is high
Read Replicas: Use for read-heavy workloads in distant regions

Network Planning

Test inter-region latency before deployment
Use dedicated interconnects when possible
Plan for network redundancy
Monitor bandwidth usage

Data Placement Strategy

Use geo-partitioning for multi-tenant applications
Set leader preferences to minimize latency
Consider data sovereignty requirements
Balance load across regions

Testing

Test failover scenarios regularly
Measure application latency from all regions
Validate replication lag under load
Simulate network partitions

Next Steps

Monitor Your Cluster

Set up monitoring for multi-region deployments

Backup & Restore

Configure cross-region backups

Get Started

Core Concepts

Deployment

Develop

Operations

Security

Advanced Features

​Overview

​3DC Deployment (Synchronous Replication)

​Architecture

​Prerequisites

​Deploy 3DC Cluster with yugabyted

​Preferred Region Configuration

​Geo-Partitioning

​xCluster Replication (Asynchronous)

​Architecture

​Deployment Types

​Unidirectional Replication

​Bidirectional Replication

​Set Up xCluster Replication

​Monitor Replication Lag

​Read Replicas

​Deploy Read Replica Cluster

​Configure Timeline-Consistent Reads

​Multi-Cloud Deployment

​Network Configuration

​VPN/VPC Peering

​Firewall Rules

​Performance Optimization

​Minimize Cross-Region Traffic

​Optimize Network Latency

​Monitor WAN Metrics

​Disaster Recovery

​Backup to Multiple Regions

​Failover Procedures

​Best Practices

​Next Steps

Monitor Your Cluster

Backup & Restore

Build docs developers (and LLMs) love

Overview

3DC Deployment (Synchronous Replication)

Architecture

Prerequisites

Deploy 3DC Cluster with yugabyted

Preferred Region Configuration

Geo-Partitioning

xCluster Replication (Asynchronous)

Architecture

Deployment Types

Unidirectional Replication

Bidirectional Replication

Set Up xCluster Replication

Monitor Replication Lag

Read Replicas

Deploy Read Replica Cluster

Configure Timeline-Consistent Reads

Multi-Cloud Deployment

Network Configuration

VPN/VPC Peering

Firewall Rules

Performance Optimization

Minimize Cross-Region Traffic

Optimize Network Latency

Monitor WAN Metrics

Disaster Recovery

Backup to Multiple Regions

Failover Procedures

Best Practices

Next Steps