This repository contains two files for benchmarking MySQL replication:
-
measure.py: This is a Python script that measures how long it takes for a MySQL follower to catch up with its leader in a leader-follower replication setup. It continuously monitors two values:Exec_Master_Log_PosandSeconds_Behind_Master. It should run on the follower. -
test.cpp: This is a C++ program that stress-tests a MySQL database by performing a large number of updates within a specified period of time, and measures the time taken to perform each update operation. It uses multiple-threading to perform these updates concurrently. It should run on the leader.
- A MySQL server (tested with MySQL 5.6 and above)
- Leader-follower replication setup
- Python 3.6 or above
- A C++ compiler that supports C++17
- MySQL C++ connector installed (
sudo apt-get install libmysqlcppconn-dev)
This repository includes a Makefile for building the C++ program:
make
This will produce an executable named testapp.
To clean up the build artifacts:
make clean
Please make sure to set the MySQL credentials in the C++ and Python script before running any.
To run the measure.py script:
python3 measure.py
The testapp program supports two commands:
- Prepare: This command prepares a specified number of rows in the MySQL table. To use it:
./testapp prepare <num_rows>
- Update: This command starts a specified number of threads, each of which performs updates on the MySQL table for a specified duration. To use it:
./testapp update <end_id> <num_threads> <num_seconds>