Step-by-Step Guide to Building and Benchmarking a Redis Master‑Slave Architecture

Hello, I am Wukong. After last week’s deep dive into Redis master‑slave principles, many readers appreciated the plain‑language explanation.

Prerequisite Articles

Previous cache‑practice series (parts 1‑7) are linked for reference.

Master‑Slave Architecture Setup

The setup consists of one master node handling writes and one slave node handling reads; successful data copy from master to slave confirms the architecture.

Redis Environment Installation

Install Redis on both machines following the official documentation.

Configure Master‑Slave and Read‑Write Separation

Required Settings

Slave read‑only flag (slave‑read‑only) – default, no change needed.

Authentication passwords: masterauth on the slave, requirepass on the master.

Slave‑of configuration to bind the slave to its master.

Bind Redis to the machine’s IP address.

Configuration Steps

(1) Shut down Redis on both servers: redis-cli shutdown (2) On the second machine (slave), edit /etc/redis/6379.conf:

vi /etc/redis/6379.conf

slave-read-only yes
masterauth abc123
slaveof eshop-cache01 6379
bind 192.168.10.115

(3) On the first machine (master), edit /etc/redis/6379.conf:

vi /etc/redis/6379.conf

slave-read-only yes
requirepass abc123
bind 192.168.10.86

(4) Restart Redis instances, starting the master first then the slave:

cd /etc/init.d
./redis_6379 start

(5) Verify that Redis processes are running: ps -ef | grep redis (6) Check slave information:

# Connect to the slave
redis-cli -h 192.168.10.115 -p 6379 -a abc123
# Show replication info
info replication

(7) Check master information:

# Connect to the master
redis-cli -h 192.168.10.86 -p 6379 -a abc123
# Show replication info
info replication

Open port 6379 on both machines if needed:

iptables -A INPUT -p tcp --dport 6379 -j ACCEPT

Testing Master‑Slave Synchronization

Verify Existing Data Sync

Check that keys are identical on both nodes: get keys Result shows matching key count and names.

Check a specific key: get key11 Values match, confirming synchronization.

Real‑Time Sync Test

On the master, set a new key: set key12 120 On the slave, retrieve it: get key12 The slave returns 120, proving real‑time replication.

Read‑Only Slave Test

Attempting to write on the slave should fail, confirming the read‑only setting.

Benchmarking the Master‑Slave Setup

Redis includes the redis-benchmark tool.

Typical benchmark command:

redis-benchmark [-h <host>] [-p <port>] [-c <clients>] [-n <requests>] [-k <boolean>]

We use the default parameters: 50 concurrent clients, 100 000 total requests, 2‑byte payload.

Concurrency: 50

Total requests: 100 000

Data size per GET/SET: 2 bytes

Run the benchmark against the slave:

cd /usr/local/bin
./redis-benchmark -h 192.168.10.115 -p 6379 -c 50 -n 100000 -d 2

Sample GET command result:

====== GET ======
  100000 requests completed in 1.51 seconds
  50 parallel clients
  2 bytes payload
  keep alive: 1

99.57% <= 1 milliseconds
99.77% <= 4 milliseconds
99.86% <= 5 milliseconds
99.92% <= 6 milliseconds
99.94% <= 7 milliseconds
99.99% <= 9 milliseconds
100.00% <= 9 milliseconds
66050.20 requests per second

Interpretation:

100 000 requests finished in 1.51 s.

99.57% completed within 1 ms.

All requests finished within 9 ms.

Throughput reached ~66 k requests/second.

Other command results are similar.

With the setup and benchmark complete, refer to the detailed article on master‑slave principles for deeper understanding.

Community Invitation

The author also promotes a knowledge‑sharing community (“星球”) with daily study challenges and interview questions, offering a low‑cost subscription.