Redis offers three main clustering solutions: Redis Cluster, master/slave replication, and Sentinel mode for automatic failover and fault recovery.

Sentinel Mode Overview

Sentinel mode builds upon the master/slave replication model by introducing monitoring agents that automatically handle failures. Traditional master/slave setups require manual intervention when the master server fails, which is time-consuming and results in service downtime. Sentinel addresses this limitation by providing automated failover capabilities.

Sentinel State Persistence

Sentinel states are persistently stored in the sentinel configuration file. Each time a new configuration is received or created, it's written to disk with a version timestamp. This ensures safe stopping and restarting of santinel processes.

Sentinel Functions

• Master server health monitoring
• Automatic master-slave failover when master fails
• Configuration updates across master, slave, and sentinel configurations

How Sentinel Operates

Each Sentienl instance performs several periodic tasks:

• Sends PING commands to master, slaves, and other Sentinels every second
• Marks instances as subjectively down if they don't respond within the configured timeout
• Confirms master failure with other Sentinels when enough agree on the failure
• Sends INFO commands to all known masters and slaves every 10 seconds
• Increases INFO command frequency to slaves when master is down

Three Core Monitoring Tasks

1. Topology Discovery: Every 10 seconds, each Sentinel sends INFO commands to master and slaves to discover the current cluster topology. This allows Sentinels to detect new slaves as they join.
2. Sentinel Communication: Every 2 seconds, Sentinels publish their status and master assessments to a common channel while subscribing to the same channel to learn about other Sentinels.
3. Heartbeat Monitoring: Every second, Sentinels send PING commands to master, slaves, and other Sentinels to verify node health.

Sentinel Deployment Process

First, ensure Redis master-slave replication is properly configured.

Create Sentinel Directories

mkdir -p /opt/redis/sentinel

Create Sentinel Configuration Files

cat > /opt/redis/sentinel/sentinel.conf << EOF
port 26380
dir "/opt/redis/sentinel"
logfile /opt/redis/sentinel/sentinel.log
protected-mode no
daemonize yes
sentinel monitor redis-primary 192.168.1.10 6379 2
sentinel down-after-milliseconds redis-primary 5000
sentinel auth-pass redis-primary secure123
EOF

Configuration Parameters:

• sentinel monitor redis-primary 192.168.1.10 6379 2: Defines the initial master server and requires 2 Sentinel votes for failover decisions
• sentinel down-after-milliseconds redis-primary 5000: Marks master as down after 5 seconds without response
• sentinel auth-pass redis-primary secure123: Authentication password for Redis cluster access

Start Sentinel

redis-sentinel /opt/redis/sentinel/sentinel.conf
netstat -lntup | grep 26380
tail -f /opt/redis/sentinel/sentinel.log

Simulate Master Failure

redis-cli -a secure123 -p 6379 shutdown
redis-cli -a secure123 -p 6379 info replication

Sentinel will automatically detect the master failure and initiate failover, promoting a slave to master. The new master will be 192.168.1.11 in this example.

Recover Failed Master

systemctl start redis
netstat -lntup | grep 6379

The recovered node will automatically rejoin as a slave.

Common Sentinel Commands

PING

Check if Redis or Sentinel instances are responding:

redis-cli -a secure123 -p 26380 PING
# Returns: PONG

SENTINEL masters

List monitored master servers and their status:

redis-cli -a secure123 -p 26380 sentinel masters

SENTINEL slaves

List slaves for a specific master:

redis-cli -a secure123 -p 26380 sentinel slaves redis-primary

SENTINEL get-master-addr-by-name

Retrieve IP and port of the current master:

redis-cli -a secure123 -p 26380 sentinel get-master-addr-by-name redis-primary

SENTINEL reset

Reset monitoring state for a master:

redis-cli -a secure123 -p 26380 sentinel reset redis-primary

SENTINEL failover

Force manual failover:

redis-cli -a secure123 -p 26380 sentinel failover redis-primary