PostgreSQL 参数优化设置 32GB内存（推荐） 内存参数 检查点 日志参数 自动初始化参数shell脚本

1.修改参数列表

（1）执行计划

enable_nestloop = off #默认为on

enable_seqscan = off #默认为on

enable_indexscan = on

enable_bitmapscan = on

max_connections = 1000 #默认为100

 

（2）内存相关

shared_buffers = 16GB # 默认为128MB

effective_cache_size = 24GB #默认为4GB

work_mem = 32MB

temp_buffers = 32MB

wal_buffers = 512MB

maintenance_work_mem = 2GB

 

（3）autovacuum

autovacuum = on

autovacuum_max_workers = 3 #默认为3

autovacuum_work_mem = 256MB

autovacuum_vacuum_scale_factor = 0.05 #默认为0.2

 

（4）检查点

max_wal_size = 4GB #默认为1GB

min_wal_size = 1GB #默认为80MB

checkpoint_completion_target = 0.9 #默认为0.5

checkpoint_timeout = 30min #默认为5min

 

（5）日志

log_destination = csvlog

log_directory = pg_log

logging_collector = on

log_min_duration_statement = 800

log_rotation_size = 1024MB

log_truncate_on_rotation = on

log_filename = 'xl_log_%a.log'

 

2.参数设置公式

这个shell脚本可以使用shell的关联数组对

#!/bin/bash
#command to execute script: su - postgres -c "./create.sh"
#default value refer to standard server(cpu:6cores,memory:32G,storage:12T)
#value according to the actual situation 

#List of parameters to be modified
:<<!
	enable_seqscan = off
	enable_indexscan = on
	enable_bitmapscan = on
	max_connections = 1000
	shared_buffers = 16GB
	effective_cache_size = 24GB
	work_mem = 32MB
	temp_buffers = 32MB
	wal_buffers = 512MB
	maintenance_work_mem = 2GB
	autovacuum_max_workers = 3
	autovacuum_work_mem = 256MB
	checkpoint_timeout = 30min
	max_wal_size = 4GB
	min_wal_size = 1GB
	checkpoint_completion_target = 0.9
	log_destination = csvlog
	log_directory = pg_log
	logging_collector = on
	log_min_duration_statement = 800
	log_rotation_size = 1024MB
	log_truncate_on_rotation = on
	log_filename = 'xl_log_%a.log'
!

ERROR="\033[41;37m ERROR \033[0m"
INFO="\033[42;37m INFO \033[0m"
WARN="\033[43;37m WARN \033[0m"
COMMON_ERROR="some error happened, specific information please see console output"

# Array of parameters
declare -A parameter_array
parameter_array=([enable_seqscan]=off [enable_indexscan]=on [enable_bitmapscan]=on [max_connections]=1000 [shared_buffers]=16GB 
[effective_cache_size]=24GB [work_mem]=32MB [temp_buffers]=32MB [wal_buffers]=512MB [maintenance_work_mem]=2GB 
[autovacuum_max_workers]=3 [autovacuum_work_mem]=256MB [checkpoint_timeout]=30min [max_wal_size]=4GB 
[min_wal_size]=1GB [checkpoint_completion_target]=0.9 [log_destination]=csvlog [log_directory]=pg_log 
[logging_collector]=on [log_min_duration_statement]=800 [log_rotation_size]=1024MB [log_truncate_on_rotation]=on [log_filename]=viid_log_%a.log )

#default value :
pgctl_path=
data_directory=
memory=

# check command exit value, 0 is success
function check_fun(){
	status=$?
	error=${COMMON_ERROR}
	if [[ 0 -ne ${status} ]] ; then
		echo -e "${ERROR} ${error}"
		exit 1
	fi
}

# prepare conditions
function prepare_conditions(){
	data_directory=$(psql -qtAX  -c "show data_directory" | sed 's/[ ]//g')
	check_fun
	if [[ ! -d "${data_directory}" ]] ; then
		echo -e "${ERROR} database's data directory does not exist"
		exit 1
	fi
	# physical machine environment
	memory=$(grep MemTotal /proc/meminfo | awk '{print $2 / 1024 / 1024}' | sed 's/\.[0-9]*//' | tail -n 1)
	check_fun
	# docker environment 
	memory_limit=$(($(awk '{print $1}' /sys/fs/cgroup/memory/memory.limit_in_bytes) / 1024 / 1024 /1024))
	check_fun
	# comparing the two, choose the smaller one.
	if [[ "${memory_limit}" -le "${memory}" ]] ; then
		memory=${memory_limit}
	fi
}

# calculate parameters
function calculate_parameters(){
	# 50%*memory
	parameter_array[shared_buffers]=$((memory * 1024 / 2))"MB"
	# 75%*memory
	parameter_array[effective_cache_size]=$((memory * 1024 * 3 / 4 ))"MB"
	# <1%
	parameter_array[work_mem]=${memory}"MB"
	# <1%
	parameter_array[temp_buffers]=${memory}"MB"
	# 32GB => 512MB
	parameter_array[wal_buffers]=$((memory * 16 ))"MB"
	# 32GB => 2048MB
	parameter_array[maintenance_work_mem]=$((memory * 64 ))"MB"
}

# modify parameters
function modify_parameters(){
	# modify parameters by modifying file postgresql.conf:
	for parameter in ${!parameter_array[*]}
	do
		#check whether the parameters have been modified
		# PostgreSQL's default parameter configuration example: enable_seqscan = on
		# viid's example: enable_seqscan='on'
		check_out=$(grep "^${parameter}=" "${data_directory}"/postgresql.conf | grep -v '#' | tail -n 1)
		# process sleep 
		sleep 0.2s
		if [[ -z "${check_out}" ]] ; then
			echo "${parameter}='${parameter_array[${parameter}]}'" >> "${data_directory}"/postgresql.conf
			check_fun
			echo -e "${INFO} modify ${parameter} successfully"
		else 
			if [[ "${check_out}" = "${parameter}='${parameter_array[${parameter}]}'" ]] ; then
				echo -e "${INFO} ${parameter} is already configured, then skip this step"
			else
				sed -i s!^"${check_out}"!"${parameter}='${parameter_array[${parameter}]}'"!g  "${data_directory}"/postgresql.conf
				check_fun
				echo -e "${INFO} modify ${parameter} successfully"
			fi
		fi
	done
}

# create directory(pg_log and pg_arch), and set user postgres permission
function create_dir(){
	directory_array=("pg_arch" "pg_log")
	for directory in ${directory_array[*]};
	do
		# process sleep 
		sleep 0.2s
		if [[ ! -d "${data_directory}/${directory}" ]] ; then
			mkdir -p "${data_directory}"/"${directory}"
			echo -e "${INFO} path ${data_directory}/${directory} create successfully"
		else 
			echo -e "${INFO} ${data_directory}/${directory} is already exists, then skip this step"
		fi
	# set user postgres permission
	chown postgres:postgres "${data_directory}"/"${directory}"
	done
}

# because the environment is different, need to find the path of the database restart command 'pg_ctl'
function find_cmd(){
	result=$(find / -name pg_ctl 2> /dev/null | grep bin/pg_ctl$ | tail -n 1 )
	# check whether the path exists
	if [[ -z "${result}" ]] ; then
		echo -e "${ERROR} database restart command 'pg_ctl' not exists"
		echo -e "${ERROR} please check to see if the database is installed or the command directory does not have permission to access it"
		exit 1
	else 
		echo -e "${INFO} database restart command 'pg_ctl' path: ${result}"
		pgctl_path=${result}
	fi
}

# user choose whether to restart or not
function check_restart(){
	read -r -p "Is it necessary to restart database immediately?[Enter YES or NO]:" result
	if [[ "${result,,}" = "yes" ]] ; then
		echo -e "${INFO} start to restart database"
		${pgctl_path} restart -D "${data_directory}" >& /dev/null
		if [[ 0 -ne ${status} ]] ; then
			echo -e "${ERROR} restart database failed"
			exit 12
		fi
	elif [[ "${result,,}" = "no" ]] ; then
		echo -e "${WARN} please restart database manually"
		echo -e "${WARN} if you don't restart, database may not be available"
		exit 11
	else
		echo -e "${ERROR} invalid input,please enter again"
		check_restart
	fi
}

# ******* start *******
# prepare conditions:memory ,data_directory
prepare_conditions
# calculate parameters
calculate_parameters
# modify parameters
modify_parameters 
# create directory(pg_log and pg_arch)
create_dir
# the path of the database restart command 'pg_ctl'
find_cmd
# remind user that they need to restart database to take effect
# process sleep 
sleep 0.5s
echo -e ""
echo -e "*******************************************************************"
echo -e "*                                                                 *"
echo -e "*                                                                 *"
echo -e "*                restart database to take effect                  *"
echo -e "*                                                                 *"
echo -e "*                                                                 *"
echo -e "*******************************************************************"
# process sleep 
sleep 0.5s
# user choose whether to restart or not 
# check_restart
echo -e "${INFO} start to restart database"
${pgctl_path} restart -D "${data_directory}" >& /dev/null