VPS Clustering and High Availability

• Virtual IP (VIP): Single IP address users connect to
• Cluster manager: Keeps VIP on healthy node (Pacemaker, Keepalived)
• Heartbeat: Constant communication between nodes
• Shared state: Database replication or shared storage

Active-Passive Setup

One active node, others on standby:

# Keepalived config for active-passive
vrrp_instance VI_1 {
    state MASTER
    interface eth0
    virtual_router_id 51
    priority 100          # Higher = more priority to be master
    advert_int 1

    virtual_ipaddress {
        192.168.1.50      # VIP (virtual IP)
    }

    track_script {
        check_web         # Custom health check
    }
}

Advantages:

• Simple setup
• One node always ready
• Less resource usage

Disadvantages:

• Standby nodes wasted (not serving traffic)
• Failover takes seconds

Active-Active Setup

All nodes serve traffic (better efficiency):

# Load balancer distributes to all nodes
Node 1: 192.168.1.101 (active)
Node 2: 192.168.1.102 (active)
Node 3: 192.168.1.103 (active)

# If Node 1 fails, load balancer removes it
# Node 2 and 3 handle all traffic
# When Node 1 recovers, it rejoins cluster

Advantages:

• All nodes serve traffic (100% resource utilization)
• Can handle more load
• Better cost efficiency

Disadvantages:

• More complex configuration
• All nodes must have identical state
• Database synchronization critical

Database Replication

MySQL replication (primary-replica):

# Primary server (Node 1) config
# /etc/mysql/mysql.conf.d/mysqld.cnf
server-id = 1
log_bin = /var/log/mysql/mysql-bin.log
binlog_do_db = myapp_db

# Create replication user
CREATE USER 'replicator'@'10.0.0.%' IDENTIFIED BY 'password';
GRANT REPLICATION SLAVE ON *.* TO 'replicator'@'10.0.0.%';

# Replica server (Node 2) config
server-id = 2
relay-log = /var/log/mysql/mysql-relay-bin
relay-log-index = /var/log/mysql/mysql-relay-bin.index
relay-log-recovery = ON

# Start replication from primary
CHANGE MASTER TO
    MASTER_HOST='10.0.0.1',
    MASTER_USER='replicator',
    MASTER_PASSWORD='password',
    MASTER_LOG_FILE='mysql-bin.000001',
    MASTER_LOG_POS=154;
START SLAVE;

Heartbeat and Detection

# Corosync provides cluster communication
# /etc/corosync/corosync.conf
nodename: node1
cluster_name: ha-cluster

interface {
    ringnumber: 0
    bindnetaddr: 192.168.1.0
    mcastaddr: 239.255.1.1
    mcastport: 5405
    ttl: 1
}

# Start cluster
systemctl start corosync
systemctl start pacemaker

# Check cluster status
crm status
corosync-cfgtool -s

Automatic Failover

# Pacemaker health check (primitive)
crm configure primitive web-app \
    ocf:heartbeat:apache \
    op monitor interval="30s" timeout="60s" \
    meta failure-timeout="5m"

# Failover policy: restart 3 times, then switch to next node
crm configure property -threshold="3"
crm configure property failure-timeout="5m"

# Set resource stickiness (prefer current node if healthy)
crm configure rsc_defaults resource-stickiness="100"

Test Your Cluster

# Simulate node failure
systemctl stop httpd            # Stop web server

# Monitor failover
watch crm status                # See VIP move to another node

# Kill entire node (harsh test)
systemctl poweroff              # Reboot node

# Traffic should route to remaining nodes within seconds

# Test failback (recovery)
systemctl start httpd           # Bring node back online
crm node online node1           # Mark node as online
# VIP should move back to preferred node

Related: Load balancing | VPS scaling | High availability | Database replication