MySQL IO线程及相关参数调优 update (进行大量数据更新引起的性能低)

日志内容

I/O thread 0 state: waiting for completed aio requests (insert buffer thread)
I/O thread 1 state: waiting for completed aio requests (log thread)
I/O thread 2 state: waiting for completed aio requests (read thread)
I/O thread 3 state: waiting for completed aio requests (read thread)
I/O thread 4 state: waiting for completed aio requests (read thread)
I/O thread 5 state: waiting for completed aio requests (read thread)
I/O thread 6 state: waiting for completed aio requests (read thread)
I/O thread 7 state: waiting for completed aio requests (read thread)
I/O thread 8 state: waiting for completed aio requests (read thread)
I/O thread 9 state: waiting for completed aio requests (read thread)
I/O thread 10 state: waiting for completed aio requests (write thread)
I/O thread 11 state: waiting for completed aio requests (write thread)
I/O thread 12 state: waiting for completed aio requests (write thread)
I/O thread 13 state: waiting for completed aio requests (write thread)
I/O thread 14 state: waiting for completed aio requests (write thread)
I/O thread 15 state: waiting for completed aio requests (write thread)
I/O thread 16 state: waiting for completed aio requests (write thread)
I/O thread 17 state: waiting for completed aio requests (write thread)

先说优化结果,增加了IO线程后,性能提升很大,批量update 很快就完成了

一、关于一个SQL的简单的工作过程
1、工作前提描述
  1、启动MySQL,在内存中分配一个大空间innodb_buffer_pool(还有log_buffer)
  2、多用户线程连接MySQL,从内存分配用户工作空间(其中排序空间)
  3、磁盘上有数据库文件、ib_logfile、tmp目录、undo
2、SQL的简易流程
  1、DQL操作
    1、首先进行内存读
    2、如果buffer pool中没有所需数据,就进行物理读
    3、物理读数据读入buffer pool,再返回给用户工作空间
  2、DML操作(例update)
    1、内存读,然后进行物理读,读取所需修改的数据行
    2、从磁盘调入undo页到buffer pool中
    3、修改前的数据存入undo页里,产生redo
    4、修改数据行(buffer pool中数据页成脏页),产生redo
    5、生成的redo先是存于用户工作空间,择机拷入log_buffer中
    6、log线程不断的将log_buffer中的记录写入redo logfile中
    7、修改完所有数据行,提交事务,刻意再触发一下log线程
    8、待log_buffer中的相关信息都写完,响应事务提交成功
  至此,日志写入磁盘,内存脏块还在buffer pool中(后台周期写入磁盘,释放buffer pool空间)。

二、影响SQL执行性能的因素,及具体看方式
1、大量物理读
mysql> show global status like 'i%read%';
| Innodb_buffer_pool_reads | 647 |
| Innodb_data_read | 48402944 |
| Innodb_data_reads | 2996 |
| Innodb_pages_read | 2949 |
| Innodb_rows_read | 1002172 |
  1、Innodb_buffer_pool_reads:物理读次数
  2、Innodb_data_read:物理读数据字节量
  3、Innodb_data_reads:物理读IO请求次数
  4、Innodb_pages_read:物理读数据页数
  5、Innodb_rows_read:物理读数据行数
2、Log写性能
复制代码
mysql> show engine innodb status \G

LOG

Log sequence number 144064129  //已经生成的日志量(累计值)/单位:字节
Log flushed up to 144064129  //已经写入的日志量(累计值)
Pages flushed up to 144064129  //已经写入的脏页量(累计值)
Last checkpoint at 144064120  //检查点
0 pending log flushes, 0 pending chkp writes
92 log i/o's done, 0.00 log i/o's/second
复制代码
  关于redo log的写入:
  1、Innodb_os_log_written:日志刷盘的字节数,如果在commit不怎么变化的情况下,这个值出现暴增,说明系统出现大事务了(处理:kill线程,必要情况kill掉mysql进程);
  2、Innodb_log_writes:日志写的次数。
3、磁盘排序
复制代码
mysql> show status like 'Sort_merge_passes';
+-------------------+-------+
| Variable_name | Value |
+-------------------+-------+
| Sort_merge_passes | 0 |
+-------------------+-------+
1 row in set (0.00 sec)
复制代码
  用户所需数据,如果没有内存buffer pool中,就发生物理读;  

  如果需要过滤掉很多数据,就会影响物理读和内存读,因为返回很多的数据(物理读),在内存中需要过滤掉很多数据(内存读);  

  如果涉及到group/order by,会在用户工作空间完成排序等,如果结果集过大,用户空间过小,进行磁盘排序,Sort_merge_passes>0 ,这就很影响数据库性能了。

三、MySQL线程及其工作
  MySQL的工作机制是单进程多线程:IO线程=一个log线程+四个read线程+四个write线程
复制代码
mysql> show engine innodb status \G

FILE I/O

I/O thread 0 state: waiting for i/o request (insert buffer thread)
I/O thread 1 state: waiting for i/o request (log thread)
I/O thread 2 state: waiting for i/o request (read thread)
I/O thread 3 state: waiting for i/o request (read thread)
I/O thread 4 state: waiting for i/o request (read thread)
I/O thread 5 state: waiting for i/o request (read thread)
I/O thread 6 state: waiting for i/o request (write thread)
I/O thread 7 state: waiting for i/o request (write thread)
I/O thread 8 state: waiting for i/o request (write thread)
I/O thread 9 state: waiting for i/o request (write thread)
复制代码
1、读操作:innodb_read_io_threads
  1、发起者:用户线程发起读请求
  2、完成者:读线程执行请求队列中的读请求操作
  3、如何调整读线程的数量
复制代码
mysql> show variables like 'innodb_read_io_threads';
+------------------------+-------+
| Variable_name | Value |
+------------------------+-------+
| innodb_read_io_threads | 4 |
+------------------------+-------+
1 row in set (0.01 sec)
复制代码
    默认是开启4个读线程,静态参数,修改至配置文件中
  4、如何确定是否需要增加读线程的数量
    查看线程的状态:I/O thread 2 state: waiting for i/o request (read thread)
2、写操作:innodb_write_io_threads
  1、发起者:page_cleaner线程发起
  2、完成者:写线程执行请求队列中的写请求操作
  3、如何调整写线程的数量
复制代码
mysql> show variables like 'innodb_write_io_threads';
+-------------------------+-------+
| Variable_name | Value |
+-------------------------+-------+
| innodb_write_io_threads | 4 |
+-------------------------+-------+
1 row in set (0.01 sec)
复制代码
    默认是开启4个写线程,静态参数,修改至配置文件中
  4、如何确定是否需要增加写线程的数量
    查看线程的状态:I/O thread 6 state: waiting for i/o request (write thread)
关于innodb_purge_threads:page cleaner 线程
作用:

  1、负责对 undo 数据页的清空

  2、数据页中 delete 标志行的清除

  3、清理 innodb buffer pool,负责把内存中的脏页发起写请求,write 线程负载把脏页刷新到磁盘上。

3、日志线程
3.1、只有一个日志线程
  1、是否繁忙
    I/O thread 1 state: waiting for i/o request (log thread):闲
  2、日志写性能
复制代码
mysql> show global status like 'Innodb_log_waits';
+------------------+-------+
| Variable_name | Value |
+------------------+-------+
| Innodb_log_waits | 0 |
+------------------+-------+
1 row in set (0.00 sec)
复制代码
    如果log buffer太小,就很容易满,导致无法写入,产生日志等待。

  3、日志写压力
    1、每秒吞吐量
复制代码
mysql> show global status like 'Innodb_os_log_written';  #redo log写的字节数
+-----------------------+-------+
| Variable_name | Value |
+-----------------------+-------+
| Innodb_os_log_written | 57856 |  
+-----------------------+-------+
1 row in set (0.01 sec)
复制代码
    2、每秒写入次数
复制代码
mysql> show global status like 'Innodb_log_writes';
+-------------------+-------+
| Variable_name | Value |
+-------------------+-------+
| Innodb_log_writes | 59 |
+-------------------+-------+
1 row in set (0.01 sec)
复制代码
3.2、对于日志监控来说,三个经典参数
  1、Innodb_log_waits  #redo写入的等待次数
  2、Innodb_log_writes  #redo写入的次数
  3、Innodb_os_log_written  #写入redo logfile中的字节量
3.3、日志写入异常判断
复制代码
mysql> show engine innodb status \G
Pending flushes (fsync) log: 0;

mysql> show global status like 'Innodb_os_log_pending_fsyncs';
+------------------------------+-------+
| Variable_name | Value |
+------------------------------+-------+
| Innodb_os_log_pending_fsyncs | 0 |
+------------------------------+-------+
1 row in set (0.00 sec)
复制代码
  1、fsync:绕过文件系统缓存,直接将内存中的数据写入存储中,实现数据真正写入可靠的介质磁盘里。(对于redo log来说,通过fsync方式写入磁盘才是可靠的保证,因为写入文件系统缓存的提交成功响应并不是真正的将redo写入磁盘的logfile中)(sync:同步)
  2、pending:挂起(写不动),redo写入存储cache过程中,某种原因io繁忙,cache被占满,超时响应,就会被挂起;
  3、>0就说明系统IO出现问题,=0说明is OK。

四、log buffer调整依据
  log buffer:日志缓存,一般都很小,调整100M足够使用
mysql> show variables like "%log_buffer%";
+------------------------+----------+
| Variable_name | Value |
+------------------------+----------+
| innodb_log_buffer_size | 16777216 |
+------------------------+----------+
  5.7默认是16M,5.6默认是8M
1、文件中(log file):每次写的时候全局都写,不会挑着捡着写
  1、日志写线程每一秒redo日志缓冲刷新到重做日志文件
  2、每个事务提交时会将重做日志缓冲刷新到重做日志文件
  3、每当重做日志缓冲池剩余空间小于1/2时,重做日志缓冲刷新到重做日志文件。
2、在企业中往往设置50-100M,最多设置为2、300M,调整依据:
  1、内存空间足够大
  2、日志产生量大,系统io阻塞了,系统的io占用的是一个带宽,log_writes线程被阻塞,log buffer满了,数据库会hang住。
  3、Innodb_log_waits(状态值)
    The number of times that the log buffer was too small and a wait was required for it to be flushed before continuing.

posted @ 2022-08-17 23:24  方东信  阅读(1000)  评论(0编辑  收藏  举报