Written by arstercz - 2014-12-18

MySQL numa交叉模式启动说明

numa交叉模式说明: http://www.percona.com/doc/percona-server/5.5/performance/innodb_numa_support.html

db01: 开启numa interleave; RAM 64G; db02: 关闭numa interleave; RAM 64G; 两台db均分配48G内存buffer pool; Percona Server 5.5.33版本;

mysqld_safe部分配置增加numa参数:

#NUMA support
numa_interleave = 1
innodb_buffer_pool_populate = 1
flush_caches=1

innodb_buffer_pool_populate: 如果服务器为NUMA架构，且内存的选择策略为selected, 在buffer pool分配的内存大小大于节点的可用内存时,系统则会进行swap交换操作,即便其它节点还有可用的内存。该选项功能引用自Twitter’s的MySQL补丁, 启用的该选项在buffer cache还是clean的时候(即空机) 启动实例后为InnoDB进行预分配页空间以强制决定NUMA分配的策略; http://www.ibm.com/developerworks/cn/linux/l-numa/

swap对numa架构的影响见: http://blog.jcole.us/2010/09/28/mysql-swap-insanity-and-the-numa-architecture/

db01:

available: 2 nodes (0-1)
node 0 cpus: 0 2 4 6 8 10 12 14 16 18 20 22
node 0 size: 32722 MB
node 0 free: 16432 MB
node 1 cpus: 1 3 5 7 9 11 13 15 17 19 21 23
node 1 size: 32768 MB
node 1 free: 19543 MB
node distances:
node   0   1 
  0:  10  20 
  1:  20  10

db02:

available: 2 nodes (0-1)
node 0 cpus: 0 2 4 6 8 10 12 14 16 18 20 22
node 0 size: 32722 MB
node 0 free: 27778 MB
node 1 cpus: 1 3 5 7 9 11 13 15 17 19 21 23
node 1 size: 32768 MB
node 1 free: 31067 MB
node distances:
node   0   1 
  0:  10  20 
  1:  20  10

可以看到 db01启用interleave模式后, node1, node2同时对外服务, 尽管对于内存而言存在本地和远端模式的访问, 但比起swap操作, 开销更少; db02主机的node1节点是空闲的;

← → Top