SRE文档

深浅模式

系统优化 ​

修改 /etc/grub.conf ​

关闭 numa=off ，修改磁盘IO调度方式 elevator=deadline

修改方法：

grubby --update-kernel=ALL --args="transparent_hugepage=never" --args="elevator=deadline"
验证：
grubby --info=ALL
返回
args="ro crashkernel=auto rhgb quiet numa=off elevator=deadline"
grubby --update-kernel=ALL --args="transparent_hugepage=never" --args="elevator=deadline"
验证：
grubby --info=ALL
返回
args="ro crashkernel=auto rhgb quiet numa=off elevator=deadline"

关闭内存大页 hugepage ​

验证是否开启方法1： 
如果以下文件不存在，则是THP已经从内核中移除。
/sys/kernel/mm/transparent_hugepage or /sys/kernel/mm/redhat_transparent_hugepage
验证是否开启方法2： 
cat /sys/kernel/mm/transparent_hugepage/enabled
[always] madvise never

[always] == 》表示开启
验证是否开启方法1： 
如果以下文件不存在，则是THP已经从内核中移除。
/sys/kernel/mm/transparent_hugepage or /sys/kernel/mm/redhat_transparent_hugepage
验证是否开启方法2： 
cat /sys/kernel/mm/transparent_hugepage/enabled
[always] madvise never

[always] == 》表示开启

禁用内存大页 ​

方法： ​

修改 /etc/rc.local

# vi /etc/rc.local #追加，禁用大页

if test -f /sys/kernel/mm/transparent_hugepage/enabled; then
  echo never > /sys/kernel/mm/transparent_hugepage/enabled
fi
if test -f /sys/kernel/mm/transparent_hugepage/defrag; then
  echo never > /sys/kernel/mm/transparent_hugepage/defrag
fi
# vi /etc/rc.local #追加，禁用大页

if test -f /sys/kernel/mm/transparent_hugepage/enabled; then
  echo never > /sys/kernel/mm/transparent_hugepage/enabled
fi
if test -f /sys/kernel/mm/transparent_hugepage/defrag; then
  echo never > /sys/kernel/mm/transparent_hugepage/defrag
fi

内核优化 ​

vi /etc/sysctl.conf
vi /etc/sysctl.conf

添加内容如下：

vm.swappiness = 0
kernel.shmmax=135497418752
net.core.rmem_max = 4194304
net.core.wmem_max = 4194304
net.core.rmem_default = 262144
net.core.wmem_default = 262144
net.ipv4.ip_local_port_range = 9000 65535
kernel.sem = 50100 64128000 50100 1280
vm.dirty_background_bytes = 102400000
vm.dirty_ratio = 80
vm.nr_hugepages = 102352

/etc/security/limits.conf
* soft nofile 655360
* hard nofile 655360
* soft nproc 655360
* hard nproc 655360
* soft stack unlimited
* hard stack unlimited
* soft   memlock    250000000
* hard   memlock    250000000
vm.swappiness = 0
kernel.shmmax=135497418752
net.core.rmem_max = 4194304
net.core.wmem_max = 4194304
net.core.rmem_default = 262144
net.core.wmem_default = 262144
net.ipv4.ip_local_port_range = 9000 65535
kernel.sem = 50100 64128000 50100 1280
vm.dirty_background_bytes = 102400000
vm.dirty_ratio = 80
vm.nr_hugepages = 102352

/etc/security/limits.conf
* soft nofile 655360
* hard nofile 655360
* soft nproc 655360
* hard nproc 655360
* soft stack unlimited
* hard stack unlimited
* soft   memlock    250000000
* hard   memlock    250000000

块设备预读 ​

blockdev --setra 16384 /dev/sda

设置开机启动生效：

echo "blockdev --setra 16384 /dev/sda" >> /etc/rc.local
blockdev --setra 16384 /dev/sda

设置开机启动生效：

echo "blockdev --setra 16384 /dev/sda" >> /etc/rc.local

应用层 ​

show max_connections;
show shared_buffers;
show wal_buffers;
show checkpoint_completion_target;
show checkpoint_timeout;
show fsync;
show commit_delay;
show autovacuum_work_mem;
show commit_siblings;
show effective_cache_size;
show maintenance_work_mem;
show work_mem;



1.用超级用户运行
postgres=# 

SELECT pg_reload_conf();

2.用UNIX的kill手动发起HUP信号

$kill -HUP PID

3.使用pg_ctl命令触发SIGHUP信号

$pg_ctl reload
show max_connections;
show shared_buffers;
show wal_buffers;
show checkpoint_completion_target;
show checkpoint_timeout;
show fsync;
show commit_delay;
show autovacuum_work_mem;
show commit_siblings;
show effective_cache_size;
show maintenance_work_mem;
show work_mem;



1.用超级用户运行
postgres=# 

SELECT pg_reload_conf();

2.用UNIX的kill手动发起HUP信号

$kill -HUP PID

3.使用pg_ctl命令触发SIGHUP信号

$pg_ctl reload

https://www.kancloud.cn/taobaomysql/monthly/140098