mysql 5.7 sysbench_MySQL 5.7和8.0 MHA架构下sysbench压测

sysbench压测环境：

sysbench server：172.16.7.105

压测数据库服务器：MHA架构，172.16.7.100(主节点)、172.16.7.101、172.16.7.102

MySQL在172.16.7.105(安装MySQL5.7.26)部署目录：/opt/mysql/

主要压测172.16.7.100主节点

172.16.7.100配置：

MySQL5.7.26

CPU：4核

内存：8G

innodb_buffer_pool_size = 28G

innodb_log_file_size = 512M

innodb_log_files_in_group = 2

innodb_log_buffer_size=16M

MySQL8.0.17

CPU：4核

内存：8G

innodb_buffer_pool_size = 28G

innodb_log_file_size = 512M

innodb_log_files_in_group = 2

innodb_log_buffer_size=16M

前置工作

1.完成MySQL的安装

2.完成动态库文件的安装

yum install -y openssl libtool

#可能遇到的报错

1.在make之前，需要修改lib库，否则会得到一个报错:/usr/bin/ld: cannot find -lmysqlclient_r

因为安装中指向的lib是/opt/mysql/lib

cd /opt/mysql/lib

ln -s libmysqlclient.so.20.3.9 libmysqlclient_r.so

2.make的时候没有明显错误，执行./sysbench时报错

./sysbench: error while loading shared libraries: libmysqlclient.so.18: cannot open shared object file: No such file or directory

ln -s /opt/mysql/lib/libmysqlclient.so.18 /usr/lib64

如果报错为

/usr/local/sysbench/bin/sysbench: error while loading shared libraries: libmysqlclient.so.20: cannot open shared object file: No such file or directory

那么就执行

ln -s /opt/mysql/lib/libmysqlclient.so.20 /usr/lib64

下载sysbench：

wget http://imysql.com/wp-content/uploads/2014/09/sysbench-0.4.12-1.1.tgz

部署：

tar -xf sysbench-0.4.12-1.1.tgz

mv sysbench-0.4.12-1.1 sysbench

cp -r sysbench /usr/local/sysbench

cd /usr/local/sysbench

./autogen.sh

#生成configure文件

./configure --prefix=/usr/local/sysbench/  --with-mysql=/opt/mysql/  --with-mysql-includes=/opt/mysql/include/ --with-mysql-libs=/opt/mysql/lib/

注意此处的includes必须是mysql对应的include目录，libs也是mysql对应的lib目录

完成安装

make&&make install

(确认没有报错)

验证安装

/usr/local/sysbench/bin/sysbench --help

如果出现下列文字，说明部署成功

[root@GTID01 bin]# /usr/local/sysbench/bin/sysbench --help

Missing required command argument.Usage: sysbench [general-options]... --test= [test-options]... command

General options:

--num-threads=N            number of threads to use [1]

--max-requests=N            limit for total number of requests [10000]

...

增加环境变量

echo -e "export PATH=\$PATH:/usr/local/sysbench/bin" >> /etc/profile

source /etc/profile

使用sysbench进行测试

先到被测试的库里建库，建用户，授权(这里通过105去测试100库)

mysql -h 172.16.7.100 -uroot -p

...

create database sbtest; ##因为lua脚本里面这里设置的就是sbtest库，库名需要与脚本里的内容保持一致

create user tpcc@'%' identified by 'tpcc';

grant all privileges on sbtest.* to tpcc@'%';

flush privileges;

MySQL8.0需要加：

alter user tpcc@'%' identified with 'mysql_native_password' by 'tpcc';

会产生报错的操作(将oltp.lua中的mysiam修改成innodb后，会找不到表)

cd /usr/local/sysbench/sysbench/tests/db/

cp oltp.lua oltp_innodb.lua

sed -i 's/myisam/innodb/g' /usr/local/sysbench/sysbench/tests/db/oltp_innodb.lua

测试多表的压测

执行prepare

sysbench --mysql-host=172.16.7.100 --mysql-port=3306 --mysql-user=tpcc --mysql-password=tpcc --test=/usr/local/sysbench/sysbench/tests/db/oltp.lua --oltp_tables_count=10 --oltp-table-size=100000 --rand-init=on prepare

执行run

sysbench --mysql-host=172.16.7.100 --mysql-port=3306 --mysql-user=tpcc --mysql-password=tpcc --test=/usr/local/sysbench/sysbench/tests/db/oltp.lua --oltp_tables_count=10 --oltp-table-size=100000 --num-threads=4 --oltp-read---report-interval=5 --rand-type=uniform --max-time=30 --max-requests=0 --percentile=99 run

--num-threads=4　　　　//线程数为4

--max-time=30　　　　　　//测试时间为30s

--report-interval=5　　　　//报告打印周期为5s

--oltp-read->

--max-requests=0                //最大执行次数这里不做限制，只由max-time进行限制

MySQL8.0我是用的30001端口

MySQL5.7用32线程压测结果：

sysbench 0.5:  multi-threaded system evaluation benchmark

Running the test with following options:

Number of threads: 32

Report intermediate results every 5 second(s)

Random number generator seed is 0 and will be ignored

Threads started!

[   5s] threads: 32, tps: 842.19, reads/s: 11867.59, writes/s: 3376.34, response time: 172.23ms (99%)

[  10s] threads: 32, tps: 1013.21, reads/s: 14187.90, writes/s: 4053.03, response time: 96.36ms (99%)

[  15s] threads: 32, tps: 1195.60, reads/s: 16709.15, writes/s: 4778.79, response time: 63.71ms (99%)

[  20s] threads: 32, tps: 1095.19, reads/s: 15341.21, writes/s: 4383.35, response time: 87.04ms (99%)

[  25s] threads: 32, tps: 1111.01, reads/s: 15567.40, writes/s: 4446.26, response time: 78.19ms (99%)

[  30s] threads: 32, tps: 1045.20, reads/s: 14628.00, writes/s: 4174.40, response time: 85.90ms (99%)

OLTP test statistics:

queries performed:

read:                            441616

write:                           126176

other:                           63088

total:                           630880

transactions:                        31544  (1050.71 per sec.)

deadlocks:                           0      (0.00 per sec.)

read/write requests:                 567792 (18912.78 per sec.)

other operations:                    63088  (2101.42 per sec.)

General statistics:

total time:                          30.0216s

total number of events:              31544

total time taken by event execution: 954.2759s

response time:

min:                                  9.17ms

avg:                                 30.25ms

max:                                389.96ms

approx.  99 percentile:              94.36ms

Threads fairness:

events (avg/stddev):           985.7500/40.24

execution time (avg/stddev):   29.8211/0.03

read: 441616　　　　 //读总数，select语句

write: 126176　　　　//写总数，insert、delete、update语句

other: 63088　　　 //其它语句，如commit等

total: 630880           //总的执行语句数

transactions: 31544 (1050.71 per sec.)　　//总的事物数(★每秒处理事物数：TPS★)

read/write requests: 567792 (18912.78 per sec.)　　//读写请求次数(★每秒的读写次数：QPS★)

other operations: 63088 (2101.42 per sec.)　　　　//其它操作的每秒执行数

General statistics:

total time: 30.0216s　　　　　　　　//总时间

total number of events:

31544　　　　//★事物总数★

total time taken by event execution: 954.2759s　　 //所有事务耗时相加(不考虑并行因素)

response time:　　　　//应答时间

min: 9.17ms　　　　//最小

avg: 30.25ms　　　　//平均

max: 389.96ms　　　　//最大

approx. 99 percentile: 94.36ms　　//99%语句执行时间

Threads fairness:　　//线程公平性

events (avg/stddev): 985.7500/40.24

execution time (avg/stddev): 29.8211/0.03

需要重点关注的几个测试结果

★总的事物数，每秒事务数TPS，QPS，时间统计信息(最大、最小、平均、99%以上语句响应时间)★

通过sysbench结果来判定数据库的能力主要还是通过TPS

执行清除

sysbench --mysql-host=172.16.7.100 --mysql-port=3306 --mysql-user=tpcc --mysql-password=tpcc --mysql-db=sbtest --oltp-tables-count=10 --oltp-table-size=100000 --num-threads=16 --max-requests=0 --max-time=30 --report-interval=1 --test=/usr/local/sysbench/sysbench/tests/db/oltp.lua cleanup

sysbench 0.5: multi-threaded system evaluation benchmark

Dropping table 'sbtest1'...

Dropping table 'sbtest2'...

Dropping table 'sbtest3'...