博客

​InnoDB是MySQL使用最多的存储引擎，通常InnoDB状态可以通过show engine innodb status\G查看。

由于MySQL不同版本采用InnoDB引擎版本不同，5.6后对show engine innodb status信息进行了优化，本文环境为5.7。

show engine innodb status统计信息

需要注意一点：show engine innodb status显示的不是当前状态，而是过去某个时间范围内InnoDB存储引擎的状态。

向右拉

Per second averages calculated from the last 59 seconds

在显示前端可看到以上信息，代表查询的信息为过去59秒内每2秒的平均值。

show engine innodb status主要包括以下几个部分：

BACKGROUND THREAD

InnoDB存储引擎的核心操作大部分都集中在Mater Thread后台线程中。

MySQL5.5版本之前：

-----------------BACKGROUND THREAD-----------------srv_master_thread loops: 846676 1_second, 846675 sleeps, 84665 10_second, 17 background, 17 flushsrv_master_thread log flush and writes: 854189

MySQL 5.6之后对Master Thread进行了优化，去除了sleeps的信息，srv_active为之前的每秒的循环，srv_idle为每10秒的的循环，srv_shutdown为停止的循环，通常为0，只在MySQL关闭时才会增加。

BACKGROUND THREAD-----------------srv_master_thread loops: 3911776 srv_active, 0 srv_shutdown, 309625 srv_idlesrv_master_thread log flush and writes: 4221384

上面可以看出主循环每10秒进行了309625次，每秒进行了3911776次，每10秒的操作符合1：10。

负载低的情况下日志缓冲刷盘次数，4221384 ≈ 3911776+309625。

根据循环次数可大概判断当前数据库负载情况。如果每秒循环次数少，每10秒次数多，证明当前负载很低；如果每秒循环次数多，每10秒次数少，远大于10：1，证明当前负载很高。

SEMAPHORES

当前等待线程的列表及事件计数器，可以评估当前负载情况。

----------SEMAPHORES----------OS WAIT ARRAY INFO: reservation count 58961200OS WAIT ARRAY INFO: signal count 125268732Mutex spin waits 770371493, rounds 6482840874, OS waits 20699077RW-shared spins 0, rounds 115276716, OS waits 14655922RW-excl spins 0, rounds 987115172, OS waits 12384598RW-sx spins 40484350, rounds 419545112, OS waits 4476477Spin rounds per wait: 115276716.00 RW-shared, 987115172.00 RW-excl, 10.36 RW-sx

要明白InnoDB如何处理互斥量(Mutexes)，以及什么是两步获得锁(two-step approach)。

1. 首先进程试图获得一个锁，如果此锁被它人占用。它就会执行所谓的spin wait，即所谓循环的查询“锁被释放了吗？”。
2. 如果在循环过程中，一直未得到锁释放的信息，则其转入OS WAIT，即所谓线程进入挂起(suspended)状态。
3. 直到锁被释放后，通过信号(singal)唤醒线程。

Spin wait的消耗远小于OSwaits。Spin wait利用cpu的空闲时间，检查锁的状态，OS Wait会有所谓content switch，从CPU内核中换出当前执行线程以供其它线程使用。所以应尽量减少OS waits，可以通过innodb_sync_spin_loops参数来平衡spin wait和os wait。Mutex信息可通过show engine innodb mutex查看。

LATEST DETECTED DEADLOCK

记录最近一次死锁信息，只有产生过死锁才会有记录。

------------------------LATEST DETECTED DEADLOCK------------------------190425 18:00:13*** (1) TRANSACTION:TRANSACTION 231E7C5DF, ACTIVE 0 sec starting index readmysql tables in use 1, locked 1LOCK WAIT 4 lock struct(s), heap size 1248, 3 row lock(s)MySQL thread id 1346996, OS thread handle 0x7fd968454700, query id 760545285 10.10.x.x app_user updatingDELETE     FROM db_0.table_0    WHERE ORDER_ID IN (  456787464 , 456787465 )*** (1) WAITING FOR THIS LOCK TO BE GRANTED:RECORD LOCKS space id 5 page no 6064 n bits 824 index `orderId_index` of table `db_0`.`table_0` trx id 231E7C5DF lock_mode X waitingRecord lock, heap no 180 PHYSICAL RECORD: n_fields 2; compact format; info bits 32 0: len 8; hex 80000015eb6a1041; asc      j A;; 1: len 8; hex 800000002018fce2; asc         ;*** (2) TRANSACTION:TRANSACTION 231E7C5DD, ACTIVE 0 sec starting index read, thread declared inside InnoDB 1mysql tables in use 1, locked 15 lock struct(s), heap size 1248, 4 row lock(s)MySQL thread id 1348165, OS thread handle 0x7fd96669f700, query id 760545283 10.10.x.x app_user updatingDELETE     FROM db_0.table_0    WHERE ORDER_ID IN (  456787464 , 456787465 )*** (2) HOLDS THE LOCK(S):RECORD LOCKS space id 5 page no 6064 n bits 824 index `orderId_index` of table `db_0`.`table_0` trx id 231E7C5DD lock_mode X locks rec but not gapRecord lock, heap no 180 PHYSICAL RECORD: n_fields 2; compact format; info bits 32 0: len 8; hex 80000015eb6a1041; asc      j A;; 1: len 8; hex 800000002018fce2; asc         ;;*** (2) WAITING FOR THIS LOCK TO BE GRANTED:RECORD LOCKS space id 5 page no 6064 n bits 824 index `orderId_index` of table `db_0`.`table_0` trx id 231E7C5DD lock_mode X waitingRecord lock, heap no 180 PHYSICAL RECORD: n_fields 2; compact format; info bits 32 0: len 8; hex 80000015eb6a1041; asc      j A;; 1: len 8; hex 800000002018fce2; asc         ;;​*** WE ROLL BACK TRANSACTION (1)

死锁是指两个或两个以上的进程在执行过程中，因争夺资源而造成的一种互相等待的现象。正常死锁会自动释放，innodb有一个内在的死锁检测工具，当死锁超过一定时间后，会回滚其中一个事务，innodb_lock_wait_timeout可配置死锁等待超时时间。

死锁在两情况下最容易产生：

• 高并发同时操作同一条数据

• 存在主键和辅助索引，加锁顺序相反

避免死锁方法即降低并发，操作数据时使加锁顺序相同。

TRANSACTIONS

包含了InnoDB事务(transaction)的统计信息。

------------TRANSACTIONS------------Trx id counter 2409176Purge done for trx's n:o < 2409171 undo n:o < 0 state: running but idleHistory list length 31LIST OF TRANSACTIONS FOR EACH SESSION:---TRANSACTION 421224214038352, not started0 lock struct(s), heap size 1136, 0 row lock(s)---TRANSACTION 421224214044736, not started0 lock struct(s), heap size 1136, 0 row lock(s)---TRANSACTION 421224214039264, not started0 lock struct(s), heap size 1136, 0 row lock(s)---TRANSACTION 2409171, ACTIVE 1549 sec fetching rows, thread declared inside InnoDB 3871mysql tables in use 1, locked 00 lock struct(s), heap size 1136, 0 row lock(s)MySQL thread id 653597, OS thread handle 140289889908480, query id 2528936 127.0.0.1 root Sending dataSELECT /*!40001 SQL_NO_CACHE */ * FROM `table`

当前活跃的事物状态为ACTIVE，事物的详细信息，包括线程ID、执行时间、用户、SQL等。正在使用1个表，涉及锁的表0个。

FILE I/O

在InnoDB中大量使用了AIO（Async IO）来处理IO 请求，IO Thread主要是负责这些IO请求的回调处理，通过调用fsync()函数协调内存与磁盘之间的数据。

--------FILE I/O--------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)Pending normal aio reads: [0, 0, 0, 0, 0, 0, 0, 0] , aio writes: [0, 0, 0, 0, 0, 0, 0, 0] ,ibuf aio reads:, log i/o's:, sync i/o's:Pending flushes (fsync) log: 0; buffer pool: 015234061 OS file reads, 304461183 OS file writes, 73899457 OS fsyncs0.00 reads/s, 0 avg bytes/read, 0.24 writes/s, 0.17 fsyncs/s

InnoDB1.0版本之前有4个IO线程，1.1后做了优化，Purge Thread从Master Thread独立出来，Purge Cleaner Thread从InnoDB1.2版本引入，都是为了减轻Master Thread的工作，提高CPU利用率。

显示各个I/O thread的pending operations,pending的log和buffer pool thread的fsync()调用；

aio：代表的是异步IO(asynchronous I/O)；

OS file：显示了reads writes fsync() 调用次数。

INSERT BUFFER AND ADAPTIVE HASH INDEX

INSERT BUFFER即合并插入缓存，从innodb 1.0.x版本开始引入Change Buffer，是INSERT BUFFER升级版，即MySQL 5.1.x以上版本都支持，不仅包括INSERT BUFFER，还包括UPDATE BUFFER、DELETE BUFFER、PURGE BUFFER。

-------------------------------------INSERT BUFFER AND ADAPTIVE HASH INDEX-------------------------------------Ibuf: size 1356, free list len 149402, seg size 149404, 2004231 mergesmerged operations: insert 1373793, delete mark 316276978, delete 5341003discarded operations: insert 0, delete mark 0, delete 0Hash table size 298897, node heap has 1 buffer(s)Hash table size 298897, node heap has 1 buffer(s)Hash table size 298897, node heap has 0 buffer(s)Hash table size 298897, node heap has 1 buffer(s)Hash table size 298897, node heap has 1 buffer(s)Hash table size 298897, node heap has 1 buffer(s)Hash table size 298897, node heap has 2 buffer(s)Hash table size 298897, node heap has 0 buffer(s)Hash table size 298897, node heap has 1 buffer(s)Hash table size 298897, node heap has 0 buffer(s)Hash table size 298897, node heap has 0 buffer(s)Hash table size 298897, node heap has 1 buffer(s)Hash table size 298897, node heap has 1 buffer(s)193.03 hash searches/s, 713.40 non-hash searches/s

可以看到自适应哈希索引大小、使用情况、每秒使用自适应哈希索引搜索情况。自适应HASH索引，由INNODB存储引擎控制，只适合等值查询，不适合范围查询。可通过innodb_adaptive_hash_index来禁用或启动。

LOG

事物日志的信息。

---LOG---Log sequence number 33859450169594Log flushed up to   33859450169564Pages flushed up to 33859450169210Last checkpoint at  338594501692010 pending log flushes, 0 pending chkp writes15044267 log i/o's done, 0.10 log i/o's/second

InnoDB事物采用Write-Ahead log策略，即事物在提交时，先写重做日志，在修改页。

Write-Ahead Log：如果一个页在写入磁盘时，必须先将内存中小于该页LSN的日志先写入到磁盘中。

重做日志有LSN、每个页有LSN、Checkpoint也有LSN。

LSN记录的是重做日志的总量，单位是字节。以下三种情况会将重做日志缓存刷到重做日志文件：

• Master Thread 每秒刷重做日志缓存到重做日志文件

• innodb_flush_log_at_trx_commit=1时，每次事务提交刷重做日志缓存到重做日志文件

• 重做日志缓冲池剩余空间小于1/2时，刷重做日志缓存到重做日志文件

BUFFER POOL AND MEMORY

innodb_buffer_pool包含数据页、索引页、undo页、insert buffer、数据字典、自适应哈希索引、锁信息等。数据库缓冲池是通过LRU列表管理的。

----------------------BUFFER POOL AND MEMORY----------------------Total large memory allocated 19789774848Dictionary memory allocated 3944999Buffer pool size   1179504Free buffers       8192Database pages     1116347Old database pages 411925Modified db pages  3Pending reads      0Pending writes: LRU 0, flush list 0, single page 0Pages made young 74514305, not young 6499732670.21 youngs/s, 0.17 non-youngs/sPages read 15233915, created 7356668, written 2647396840.00 reads/s, 0.00 creates/s, 0.10 writes/sBuffer pool hit rate 1000 / 1000, young-making rate 0 / 1000 not 0 / 1000Pages read ahead 0.00/s, evicted without access 0.00/s, Random read ahead 0.00/sLRU len: 1116347, unzip_LRU len: 0I/O sum[48]:cur[0], unzip sum[0]:cur[0]

可以看到当前Buffer Pool Size共有1179504页，即1179504*16K。新读取到的页默认插入LRU列表的5/8的位置。此值由innodb_old_blocks_pct控制，即前5/8称为new list，后面3/8的称为old list。Pages made young 显示LRU列表中old list移到new list的次数，not young显示仍在old list的次数。这两个值受innodb_old_blocks_time影响，此值为微秒。如果old list中超过30微秒不再读取，则记录not young，反之记录为Pages made young。

(root@localhost) [(none)] >show global variables like '%blocks%';+------------------------+-------+| Variable_name          | Value |+------------------------+-------+| innodb_old_blocks_pct  | 37    || innodb_old_blocks_time | 30    |+------------------------+-------+

youngs/s,non-youngs/s，表示每秒这两类操作的次数。

Pages read,created,written，表示innodb被读取，创建，写入了多少页及每秒的次数。

Buffer pool hit rate，表示缓冲池命中率，如果低于95%需要具体排查。

Pages read ahead，表示页面预读，随机预读的每秒页数。

LRU中包含unzip_LRU，unzip_LRU是管理非16KB的压缩表。

INDIVIDUAL BUFFER POOL INFO

可通过innodb_buffer_pool_instances 来配置多个缓冲池实例，默认为1。可减少数据库内部资源竞争，增加并发处理能力。如果分配多个缓冲池实例，每个缓冲池大小为 innodb_buffer_pool_size / innodb_buffer_pool_instances 。

----------------------INDIVIDUAL BUFFER POOL INFO-------------------------BUFFER POOL 0Buffer pool size   147438Free buffers       1024Database pages     139530Old database pages 51486Modified db pages  0Pending reads      0Pending writes: LRU 0, flush list 0, single page 0Pages made young 8790743, not young 774674600.00 youngs/s, 0.00 non-youngs/sPages read 1856892, created 916430, written 307271670.00 reads/s, 0.00 creates/s, 0.00 writes/sNo buffer pool page gets since the last printoutPages read ahead 0.00/s, evicted without access 0.00/s, Random read ahead 0.00/sLRU len: 139530, unzip_LRU len: 0I/O sum[6]:cur[0], unzip sum[0]:cur[0]---BUFFER POOL 1---BUFFER POOL 2---BUFFER POOL 3

可以通过information_schema.INNODB_BUFFER_POOL_STATS视图查看每个buffer_pool实例的信息，MySQL默认一个page大小为16K，可以得出POOL_SIZE * innodb_buffer_pool_instances * 16K = innodb_buffer_pool_size。

(root@localhost) [information_schema] >select POOL_ID,POOL_SIZE,FREE_BUFFERS,DATABASE_PAGES,OLD_DATABASE_PAGES,MODIFIED_DATABASE_PAGES,PAGES_MADE_YOUNG,PAGES_NOT_MADE_YOUNG from information_schema.INNODB_BUFFER_POOL_STATS;+---------+-----------+--------------+----------------+--------------------+-------------------------+------------------+----------------------+| POOL_ID | POOL_SIZE | FREE_BUFFERS | DATABASE_PAGES | OLD_DATABASE_PAGES | MODIFIED_DATABASE_PAGES | PAGES_MADE_YOUNG | PAGES_NOT_MADE_YOUNG |+---------+-----------+--------------+----------------+--------------------+-------------------------+------------------+----------------------+|       0 |     90112 |            0 |          90109 |              33279 |                       0 |            18064 |            132278807 ||       1 |     90112 |            0 |          90109 |              33282 |                       0 |            18342 |            132086061 ||       2 |     90112 |            0 |          90110 |              33282 |                       0 |            17631 |            132149779 |+---------+-----------+--------------+----------------+--------------------+-------------------------+------------------+----------------------+

详细说明同上。

ROW OPERATIONS

显示了row 操作及其他一些统计信息。

--------------ROW OPERATIONS--------------0 queries inside InnoDB, 0 queries in queue0 read views open inside InnoDBProcess ID=444943, Main thread ID=139899621590784, state: sleepingNumber of rows inserted 172887566, updated 227534242, deleted 56676133, read 7096670778.77 inserts/s, 8.04 updates/s, 0.00 deletes/s, 10.92 reads/s

queries，表示innodb内核中有多少个线程，队列中有多少个线程。

read views open inside InnoDB，表示有多少个read view 被打开，一个read view 包含事物开始点数据库内容的MVCC快照。

Process ID=444943，表示内核的主线程状态。

Number of rows inserted、updated、deleted、read，表示多少行被插入，更新和删除，读取及每秒信息，可用于监控。

可通过以下命令查看：

(root@localhost) [(none)] >show global status like 'Innodb_rows_%';+----------------------+-----------+| Variable_name        | Value     |+----------------------+-----------+| Innodb_rows_deleted  | 56676133  || Innodb_rows_inserted | 172887566 || Innodb_rows_read     | 709667077 || Innodb_rows_updated  | 227534242 |+----------------------+-----------+(root@localhost) [(none)] >show global status like 'Uptime';+---------------+---------+| Variable_name | Value   |+---------------+---------+| Uptime        | 1757270 |+---------------+---------+

END OF INNODB MONITOR OUTPUT

InnoDB信息结束语。

----------------------------END OF INNODB MONITOR OUTPUT============================

如果看不到这行输出，可能是有大量事务或者有一个大的死锁截断了输出信息。