PostgreSQL 主机性能测试方法 - 单机多实例

5 minute read

背景

业界有一些通用的数据库性能测试模型,可以用来测试硬件在不同测试模型下的性能表现。
参考
http://www.tpc.org/
https://github.com/oltpbenchmark/oltpbench
http://oltpbenchmark.com/

本文主要以PostgreSQL为例,向大家介绍一下,如何使用PostgreSQL来测试硬件的性能。

PostgreSQL 的功能非常的强大,所以可以适用于几乎所有的测试模型,同时用户还可以根据自己的应用场景设计测试模型。

一、机器部署

LVM部署

可选,如果你的主机有多个块设备,可以使用LVM的条带化,提高整体的吞吐和IOPS能力,当然你也可以使用多个表空间来利用不同的块设备。

举例

常用的配置,例如4个组,每个组为3块盘组成的RAID 5  
4 * (3 DISK RAID 5)   

lvcreate 参数举例

-i 4: 4 个组,所以条带宽度设置为4
-I 16: (3-1) * 8K   每个组3块盘,实际数据盘为2块,所以将条带大小设置为 2*8KB (8K指数据块的块大小,或者WAL的块大小)  

分区与对齐举例(假设为SSD,需要对齐)

parted -s /dev/sdb mklable gpt
parted -s /dev/sdc mklable gpt
parted -s /dev/sdd mklable gpt
parted -s /dev/sde mklable gpt
parted -s /dev/sdb mkpart primary 1MiB xxxxGB    # 对齐, 起步为条带大小的倍数 1MiB/16K 
parted -s /dev/sdc mkpart primary 1MiB xxxxGB
parted -s /dev/sdd mkpart primary 1MiB xxxxGB
parted -s /dev/sde mkpart primary 1MiB xxxxGB

逻辑卷

pvcreate /dev/sd[bcde]1
vgcreate -s 128MB vgdata01 /dev/sd[bcde]1
lvcreate -i 4 -I 16 -n lv01 -L 5T vgdata01
lvcreate -i 4 -I 16 -n lv02 -l 100%FREE vgdata01

创建文件系统(mkfs.ext4可以感知lvm条带配置,因此不需要设置mkfs.ext4的条带配置)

mkfs.ext4 /dev/mapper/vgdata01-lv01 -m 0 -O extent,uninit_bg -E lazy_itable_init=1 -T largefile -L lv01
mkfs.ext4 /dev/mapper/vgdata01-lv02 -m 0 -O extent,uninit_bg -E lazy_itable_init=1 -T largefile -L lv02
mkdir /u01
mkdir /u02
mount -o defaults,noatime,nodiratime,nodelalloc,barrier=0,data=writeback LABEL=lv01 /u01
mount -o defaults,noatime,nodiratime,nodelalloc,barrier=0,data=writeback LABEL=lv02 /u02

目录

mkdir -p /data01/digoal
mkdir -p /data02/digoal
chown digoal /data01/digoal
chown digoal /data02/digoal

io 调度 : SSD建议使用deadline 或 noop

echo deadline > /sys/block/sdb/queue/scheduler
echo deadline > /sys/block/sdc/queue/scheduler
echo deadline > /sys/block/sdd/queue/scheduler
echo deadline > /sys/block/sde/queue/scheduler

OS内核配置

配置举例,请根据环境调整
1. /etc/sysctl.conf

fs.aio-max-nr = 1048576
fs.file-max = 76724600
kernel.core_pattern= /data01/corefiles/core_%e_%u_%t_%s.%p
kernel.sem = 4096 2147483647 2147483646 512000
kernel.shmall = 107374182
kernel.shmmax = 274877906944
kernel.shmmni = 819200
net.core.netdev_max_backlog = 10000
net.core.rmem_default = 262144  
net.core.rmem_max = 4194304     
net.core.wmem_default = 262144  
net.core.wmem_max = 4194304     
net.core.somaxconn = 4096
net.ipv4.tcp_max_syn_backlog = 4096
net.ipv4.tcp_keepalive_intvl = 20
net.ipv4.tcp_keepalive_probes = 3
net.ipv4.tcp_keepalive_time = 60
net.ipv4.tcp_mem = 8388608 12582912 16777216
net.ipv4.tcp_fin_timeout = 5
net.ipv4.tcp_synack_retries = 2
net.ipv4.tcp_syncookies = 1   
net.ipv4.tcp_timestamps = 1   
net.ipv4.tcp_tw_recycle = 0   
net.ipv4.tcp_tw_reuse = 1     
net.ipv4.tcp_max_tw_buckets = 262144
net.ipv4.tcp_rmem = 8192 87380 16777216
net.ipv4.tcp_wmem = 8192 65536 16777216
net.nf_conntrack_max = 1200000
net.netfilter.nf_conntrack_max = 1200000
vm.dirty_background_bytes = 409600000      
vm.dirty_expire_centisecs = 3000            
vm.dirty_ratio = 95                         
vm.dirty_writeback_centisecs = 100           
vm.extra_free_kbytes = 4096000
vm.min_free_kbytes = 2097152
vm.mmap_min_addr = 65536
vm.overcommit_memory = 0    
vm.overcommit_ratio = 90    
vm.swappiness = 0           
vm.zone_reclaim_mode = 0    
net.ipv4.ip_local_port_range = 40000 65535    

2. /etc/security/limits.conf

* soft    nofile  655360
* hard    nofile  655360
* soft    nproc   655360
* hard    nproc   655360
* soft    memlock unlimited
* hard    memlock unlimited
* soft    core    unlimited
* hard    core    unlimited

3. disable thp

cat /sys/kernel/mm/transparent_hugepage/enabled
cat /sys/kernel/mm/transparent_hugepage/defrag
cat /sys/kernel/mm/redhat_transparent_hugepage/enabled
cat /sys/kernel/mm/redhat_transparent_hugepage/defrag

[never]

4. io schedular

/boot/grub/grub.conf

title xxxxxxxxx
        root(0,0)  
        kernel /vmlinuz-............. numa=off elevator=deadline  

5. disable selinux

cat /etc/selinux/config

SELINUX=disabled
SELINUXTYPE=targeted

6. cgroup

yum install -y libcgroup

mkdir -p /cgroup/cpu
mkdir -p /cgroup/cpuacct
mkdir -p /cgroup/memory
mkdir -p /cgroup/blkio

mount -t cgroup -o cpu cpu /cgroup/cpu
mount -t cgroup -o cpuacct cpuacct /cgroup/cpuacct
mount -t cgroup -o memory memory /cgroup/memory
mount -t cgroup -o blkio blkio /cgroup/blkio

二、自定义测试模型

100个实例, 每个实例1亿记录 。

压测 insert on conflict update 。 (前期插入,后期全更新,对IO考验较大)

每个实例 2GB shared buffer 。

(使用CGROUP)

被测试机器部署

1. install postgresql 9.6

yum -y install rsync coreutils glib2 lrzsz sysstat e4fsprogs xfsprogs ntp readline-devel zlib zlib-devel openssl openssl-devel pam-devel libxml2-devel libxslt-devel python-devel tcl-devel gcc make smartmontools flex bison perl perl-devel perl-ExtUtils* OpenIPMI-tools openldap openldap-devel

tar -jxvf postgresql-9.6rc1.tar.bz2
cd postgresql-9.6rc1
./configure --prefix=/home/digoal/pgsql9.6rc1
make world -j 32
make install-world

2. 配置环境变量

$ vi ~/env.sh
export PS1="$USER@`/bin/hostname -s`-> "
export PGPORT=$1
export PGDATA=/data02/digoal/pg_root$PGPORT
export LANG=en_US.utf8
export PGHOME=/home/digoal/pgsql9.6rc1
export LD_LIBRARY_PATH=$PGHOME/lib:/lib64:/usr/lib64:/usr/local/lib64:/lib:/usr/lib:/usr/local/lib:$LD_LIBRARY_PATH
export DATE=`date +"%Y%m%d%H%M"`
export PATH=$PGHOME/bin:$PATH:.
export MANPATH=$PGHOME/share/man:$MANPATH
export PGHOST=$PGDATA
export PGUSER=postgres
export PGDATABASE=postgres
alias rm='rm -i'
alias ll='ls -lh'
unalias vi

3. 初始化数据库集群脚本

$ vi ~/init.sh

for ((i=1921;i<1921+$1;i++))
do
  . ~/env.sh $i
  initdb -D $PGDATA -E UTF8 --locale=C -U postgres -X /data01/digoal/pg_xlog$i
  echo "local   all             all                                     trust" > $PGDATA/pg_hba.conf
  echo "host    all             all             127.0.0.1/32            trust" >> $PGDATA/pg_hba.conf
  echo "host    all             all             ::1/128                 trust" >> $PGDATA/pg_hba.conf
  echo "host    all             all             0.0.0.0/0               trust" >> $PGDATA/pg_hba.conf
done

4. 初始化数据库集群

$ rm -rf /data01/digoal/*
$ rm -rf /data02/digoal/*
$ cd ~ 
$ . ~/init.sh 100

5. 获取DM设备 major minor

# dmsetup ls
vgdata01-lv02   (253, 1)
vgdata01-lv01   (253, 0)

6. cgroup 限制

xlog盘限制iops 4000
data盘限制iops 800
cpu 0.7核, 周期1秒, quota 0.7秒
cpu real-time 调度统一 周期1秒, runtime 0.001秒
memory, 4G, 打开oom

7. 启动数据库集群脚本

$ vi ~/start.sh 

for ((i=1921;i<1921+$1;i++))
do
  . /home/digoal/env.sh $i
  cgcreate -g cpu:RULE$i
  cgcreate -g cpuacct:RULE$i
  cgcreate -g memory:RULE$i
  cgcreate -g blkio:RULE$i
  
  echo "253:0 4000" > /cgroup/blkio/RULE$i/blkio.throttle.write_iops_device
  echo "253:0 4000" > /cgroup/blkio/RULE$i/blkio.throttle.read_iops_device
  echo "253:1 800" > /cgroup/blkio/RULE$i/blkio.throttle.write_iops_device
  echo "253:1 800" > /cgroup/blkio/RULE$i/blkio.throttle.read_iops_device
  echo "70" > /cgroup/cpu/RULE$i/cpu.shares
  echo "1000000" > /cgroup/cpu/RULE$i/cpu.cfs_period_us
  echo "700000" > /cgroup/cpu/RULE$i/cpu.cfs_quota_us
  echo "1000000" > /cgroup/cpu/RULE$i/cpu.rt_period_us
  echo "1000" > /cgroup/cpu/RULE$i/cpu.rt_runtime_us
  echo "4294967296" > /cgroup/memory/RULE$i/memory.limit_in_bytes
  
  cgexec -g cpu:RULE$i -g cpuacct:RULE$i -g memory:RULE$i -g blkio:RULE$i su - digoal -c ". ~/env.sh $i ; nohup postgres -B 1GB -c port=$i -c listen_addresses='0.0.0.0' -c synchronous_commit=on -c full_page_writes=on -c wal_buffers=128MB -c wal_writer_flush_after=0 -c bgwriter_delay=10ms -c max_connections=100 -c bgwriter_lru_maxpages=1000 -c bgwriter_lru_multiplier=10.0 -c unix_socket_directories='.' -c max_wal_size=4GB -c checkpoint_timeout=30min -c checkpoint_completion_target=0.5 -c log_checkpoints=on -c log_connections=on -c log_disconnections=on -c log_error_verbosity=verbose -c autovacuum_vacuum_scale_factor=0.002 -c autovacuum_max_workers=4 -c autovacuum_naptime=5s -c random_page_cost=1.0 -c constraint_exclusion=on -c log_destination='csvlog' -c logging_collector=on -c maintenance_work_mem=256MB -c autovacuum_work_mem=256MB -D $PGDATA -k $PGDATA >/dev/null 2>&1 &"
done

8. 启动数据库集群
因为需要设置CGROUP,需要超级用户执行

$ sudo bash -c "bash"
# . /home/digoal/start.sh 100

9. 停集群脚本
为了尽快停库(checkpoint可能耗费大量IO),可以先将资源放大,然后停库。

$ vi ~/stop.sh 

for ((i=1921;i<1921+$1;i++))
do
  . /home/digoal/env.sh $i
  cgdelete cpu:RULE$i
  cgdelete cpuacct:RULE$i
  cgdelete blkio:RULE$i
  su - digoal -c ". ~/env.sh $i ; pg_ctl stop -m fast -w -t 600 -D $PGDATA"
  echo "0" > /cgroup/memory/RULE$i/memory.force_empty
  cgdelete memory:RULE$i
done

需要超级用户执行

$ sudo bash -c "bash"
# . /home/digoal/stop.sh 100

测试客户端机器部署

假设已安装postgresql 9.6

1. 目录

mkdir /data01/digoal
chown digoal /data01/digoal

2. 环境脚本

$ vi ~/env.sh

export PS1="$USER@`/bin/hostname -s`-> "
export PGPORT=$1
export PGDATA=/data02/digoal/pg_root$PGPORT
export LANG=en_US.utf8
export PGHOME=/home/digoal/pgsql9.6rc1
export LD_LIBRARY_PATH=$PGHOME/lib:/lib64:/usr/lib64:/usr/local/lib64:/lib:/usr/lib:/usr/local/lib:$LD_LIBRARY_PATH
export DATE=`date +"%Y%m%d%H%M"`
export PATH=$PGHOME/bin:$PATH:.
export MANPATH=$PGHOME/share/man:$MANPATH
export PGHOST=$PGDATA
export PGUSER=postgres
export PGDATABASE=postgres
alias rm='rm -i'
alias ll='ls -lh'
unalias vi

3. 生成压测数据结构以及pgbench调用的测试脚本
定制内容都在这里。

$ vi ~/pgbench_init.sh 

for ((i=1921;i<1921+$1;i++))
do
  . ~/env.sh $i
  export PGHOST=$2
  psql -c "drop table if exists test; create table test(id int primary key, info text, crt_time timestamp);"
  echo "\set id random(1,100000000)" > ~/test$i.sql
  echo "insert into test (id,info,crt_time) values (:id, md5(random()::text), now()) on conflict on constraint test_pkey do update set info=excluded.info, crt_time=excluded.crt_time;" >> ~/test$i.sql
done

不需要执行。

4. 压测脚本

$ vi ~/pgbench.sh

mkdir -p /data01/digoal/log
for ((i=1921;i<1921+$1;i++))
do
  . ~/env.sh $i
  export PGHOST=$2
  nohup pgbench -M prepared -f ~/test$i.sql -n -r -P 1 -c 4 -j 4 -T 2592000 >>/data01/digoal/log/$2_$i.log 2>&1 &   # 每个实例测试4个连接
done

不需要执行,执行样例如下

cd ~
. ~/pgbench.sh 100 目标主机IP

5. 压测脚本

$ vi test.sh

#!/bin/bash

cd ~
. ~/pgbench_init.sh 100 $1

for ((i=1;i>0;i=1))
do
sleep 1
CNT=`ps -ewf|grep pgbench|grep -c -v grep`
if [ $CNT -eq 0 ]; then
  . ~/pgbench.sh 100 $1
  exit
fi
sleep 1
done


$ chmod 500 test.sh

6. 调用压测脚本

nohup ./test.sh 目标IP >/dev/null 2>&1 &

7. 压测结果

$ cat /data01/digoal/log/$2_$i.log

$ head -n 30000 /data01/digoal/log/$2_$i.log |tail -n 7200 > /tmp/1
$ cat /tmp/1|awk '{print $4 "," $7 "," $10}' >/tmp/2

输出TPS,RT,标准差。

TPS表示数据库视角的事务处理能力(也就是单个测试脚本的每秒调用次数)。

RT表示响应时间。

标准差可以用来表示抖动,通常应该在1以内(越大,说明抖动越厉害)。

8. 主机性能结果

$ sar -f ....

三、内置的测试模型tpc-b

只有有两处不一样,其他与定制测试一样。

1. 生成压测数据结构

$ vi ~/pgbench_init.sh 

for ((i=1921;i<1921+$1;i++))
do
  . ~/env.sh $i
  export PGHOST=$2
  nohup pgbench -i -s 2800 >/dev/null 2>&1 &   # 2.8 亿数据
done

不需要执行。

2. 压测脚本

$ vi ~/pgbench.sh

mkdir -p /data01/digoal/log
for ((i=1921;i<1921+$1;i++))
do
  . ~/env.sh $i
  export PGHOST=$2
  nohup pgbench -M prepared -n -r -P 10 -S -c 4 -j 4 -T 2592000 >>/data01/data/log/$2_$i.log 2>&1 &  # 每个实例测试4个连接
done

Flag Counter

digoal’s 大量PostgreSQL文章入口