$df -h
Filesystem Size Used Avail Use% Mounted on
/dev/sda3 8.6G 1.5G 6.7G 18% /
/dev/sda10 784G 325G 420G 44% /u01
/dev/sda5 8.7G 7.9G 407M 96% /tmp
/dev/sda2 15G 2.8G 11G 21% /usr
/dev/sda1 122M 12M 104M 10% /boot
tmpfs 7.9G 4.0K 7.9G 1% /dev/shm

$sudo lsof | grep /tmp | sort -k7 -nr
sleep 18833 peien.htg 1w REG 8,5 8321143673 54 /tmp/netstat.log (deleted)
netstat_2 13571 peien.htg 1w REG 8,5 8321143673 54 /tmp/netstat.log (deleted)
tcprstat 18823 root 2w REG 8,5 43632 49 /tmp/myrt.daemon.log
sh 18822 mysql 2w REG 8,5 43632 49 /tmp/myrt.daemon.log
sh 18822 mysql 1w REG 8,5 43632 49 /tmp/myrt.daemon.log
myrt.pl 26045 mysql 2w REG 8,5 43632 49 /tmp/myrt.daemon.log
myrt.pl 26045 mysql 1w REG 8,5 43632 49 /tmp/myrt.daemon.log
check_age 25298 mysql 2w REG 8,5 22049 33 /tmp/check_agent.log
check_age 25298 mysql 1w REG 8,5 22049 33 /tmp/check_agent.log
mysqld 3784 mysql 6u REG 8,5 15156 13 /tmp/ibH3IFN9 (deleted)

上面lsof输出结果的第二列是PID,倒数第三列是占用空间大小
COMMAND PID USER FD TYPE DEVICE SIZE NODE NAME
可以看到文件/tmp/netstat.log (deleted)占用7个多G的空间，虽然被删除了，但是还是有进程打开它。

然后，用PID看看是哪个程序占用这个文件：

$ps -ef | grep 13571
51717 13571 1 0 2011 ? 00:15:00 /bin/bash /tmp/netstat_20110829.sh
51717 21456 13571 0 09:40 ? 00:00:00 sleep 10
zhuxu 21458 17014 0 09:40 pts/0 00:00:00 grep 13571

将这个进程KILL掉后，就OK了：

$sudo kill -9 13571
$sudo lsof | grep /tmp | sort -k7 -nr | head
tcprstat 22084 root 2w REG 8,5 49339 49 /tmp/myrt.daemon.log
sh 22083 mysql 2w REG 8,5 49339 49 /tmp/myrt.daemon.log
sh 22083 mysql 1w REG 8,5 49339 49 /tmp/myrt.daemon.log
myrt.pl 26045 mysql 2w REG 8,5 49339 49 /tmp/myrt.daemon.log
myrt.pl 26045 mysql 1w REG 8,5 49339 49 /tmp/myrt.daemon.log
check_age 25298 mysql 2w REG 8,5 24583 33 /tmp/check_agent.log
check_age 25298 mysql 1w REG 8,5 24583 33 /tmp/check_agent.log
mysqld 3784 mysql 6u REG 8,5 15156 13 /tmp/ibH3IFN9 (deleted)
su 17013 root cwd DIR 8,5 4096 2 /tmp
sort 22090 zhuxu cwd DIR 8,5 4096 2 /tmp

$df -h
Filesystem Size Used Avail Use% Mounted on
/dev/sda3 8.6G 1.5G 6.7G 18% /
/dev/sda10 784G 325G 420G 44% /u01
/dev/sda5 8.7G 56M 8.2G 1% /tmp
/dev/sda2 15G 2.8G 11G 21% /usr
/dev/sda1 122M 12M 104M 10% /boot
tmpfs 7.9G 4.0K 7.9G 1% /dev/shm

--EOF--

$perror 13
OS error code 13: Permission denied

发现是权限问题，数据目录及其文件的用户和属组都是root，修改正确的权限后，可以show tables操作，但是进行count(*)表的时候，mysql实例重启了。
后面测试了一下：
（1）mysql实例启动的时候，修改数据目录权限，然后访问报ERROR，还原权限后，都正常。
（2）mysql实例启动之前，修改数据目录权限，启动后，然后访问报ERROR，还原权限后，对表进行操作，mysql实例异常重启。

具体测试过程如下：
在mysql实例启动前修改异常权限：

$ps -ef | grep mysqld | grep -v mysql | wc -l
0
$ll
total 10906156
-rw-rw---- 1 mysql dba 5368709120 Feb 3 21:03 ibdata1
-rw-rw---- 1 mysql dba 5368709120 Feb 3 21:03 ibdata2
-rw-rw---- 1 mysql dba 104857600 Feb 3 21:03 ib_logfile0
-rw-rw---- 1 mysql dba 104857600 Feb 2 11:05 ib_logfile1
-rw-rw---- 1 mysql dba 104857600 Feb 2 11:05 ib_logfile2
-rw-rw---- 1 mysql dba 104857600 Feb 2 11:05 ib_logfile3
drwx------ 2 mysql dba 4096 Oct 28 15:20 mysql
drwx------ 2 mysql dba 4096 Dec 30 14:07 test
drwx------ 2 root root 4096 Feb 3 12:47 zx_test

$sudo -u mysql /u01/mysql/bin/mysqld_unsafe &

启动mysql实例时候，alert.log有错误信息：

120203 21:07:24 InnoDB: highest supported file format is Barracuda.
120203 21:07:25 InnoDB: Warning: allocated tablespace 12, old maximum was 0
120203 21:07:25 InnoDB: Operating system error number 13 in a file operation.
InnoDB: The error means mysqld does not have the access rights to
InnoDB: the directory.
120203 21:07:25 InnoDB: Error: trying to open a table, but could not
InnoDB: open the tablespace file './zx_test/tmp.ibd'!
InnoDB: Have you moved InnoDB .ibd files around without using the
InnoDB: commands DISCARD TABLESPACE and IMPORT TABLESPACE?
InnoDB: It is also possible that this is a temporary table #sql...,
InnoDB: and MySQL removed the .ibd file for this.
InnoDB: Please refer to
InnoDB: http://dev.mysql.com/doc/refman/5.1/en/innodb-troubleshooting-datadict.html
InnoDB: for how to resolve the issue

访问该库的时候，报错：

$mysql -uroot zx_test
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 9
Server version: 5.1.48-log Source distribution

Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.
This software comes with ABSOLUTELY NO WARRANTY. This is free software,
and you are welcome to modify and redistribute it under the GPL v2 license

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

root@zx_test 09:11:53>show tables;
ERROR 1018 (HY000): Can't read dir of './zx_test/' (errno: 13)
root@zx_test 09:11:57>
root@zx_test 09:11:59>create table tmp_zx (a int);
ERROR 1005 (HY000): Can't create table 'tmp_zx' (errno: 13)

修改回原来权限之后，对库中已经存在的表进行访问时候，mysql实例重启了，重启后可以正常访问了：

$sudo chown -R mysql:dba zx_test
$mysql -uroot zx_test
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 10
Server version: 5.1.48-log Source distribution

Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.
This software comes with ABSOLUTELY NO WARRANTY. This is free software,
and you are welcome to modify and redistribute it under the GPL v2 license

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

root@zx_test 09:14:06>show tables;
+-------------------+
| Tables_in_zx_test |
+-------------------+
| tmp |
+-------------------+
1 row in set (0.00 sec)

root@zx_test 09:14:11>create table tmp_zx (a int);
Query OK, 0 rows affected (0.01 sec)

root@zx_test 09:14:21>show tables;
+-------------------+
| Tables_in_zx_test |
+-------------------+
| tmp |
| tmp_zx |
+-------------------+
2 rows in set (0.00 sec)

root@zx_test 09:14:24>select * from tmp;
ERROR 2013 (HY000): Lost connection to MySQL server during query
root@zx_test 09:14:35>120203 21:14:35 mysqld_safe Number of processes running now: 0
120203 21:14:35 mysqld_safe mysqld restarted

root@zx_test 09:16:42>
root@zx_test 09:16:44>
root@zx_test 09:16:48>select * from tmp;
ERROR 2006 (HY000): MySQL server has gone away
No connection. Trying to reconnect...
Connection id: 1
Current database: zx_test

Empty set (0.02 sec)

root@zx_test 09:16:48>select * from tmp;
Empty set (0.00 sec)

alert.log报错信息如下：

120203 21:14:35  InnoDB: error: space object of table'zx_test/tmp',
InnoDB: space id 990 did not exist in memory. Retrying an open.
120203 21:14:35  InnoDB: Assertion failure in thread 1242745152 in file fil/fil0fil.c line 2982
InnoDB: Failing assertion: flags != DICT_TF_COMPACT
InnoDB: We intentionally generate a memory trap.
InnoDB: Submit a detailed bug report to http://bugs.mysql.com.
InnoDB: If you get repeated assertion failures or crashes, even
InnoDB: immediately after the mysqld startup, there may be
InnoDB: corruption in the InnoDB tablespace. Please refer to
InnoDB: http://dev.mysql.com/doc/refman/5.1/en/forcing-recovery.html
InnoDB: about forcing recovery.
120203 21:14:35 - mysqld got signal 6 ;
This could be because you hit a bug. It is also possible that this binary
or one of the libraries it was linked against is corrupt, improperly built,
or misconfigured. This error can also be caused by malfunctioning hardware.
We will try our best to scrape up some info that will hopefully help diagnose
the problem, but since we have already crashed, something is definitely wrong
and this may fail.

key_buffer_size=2147483648
read_buffer_size=1048576
max_used_connections=1
max_threads=1100
threads_connected=1
It is possible that mysqld could use up to
key_buffer_size + (read_buffer_size + sort_buffer_size)*max_threads = 3344878 K
bytes of memory
Hope that's ok; if not, decrease some variables in the equation.

thd: 0x1ef29df0
Attempting backtrace. You can use the following information to find out
where mysqld died. If you see no messages after this, something went
terribly wrong...
stack_bottom = 0x4a12c0f0 thread_stack 0x80000
/u01/mysql/libexec/mysqld(my_print_stacktrace+0x35)[0x884b95]
/u01/mysql/libexec/mysqld(handle_segfault+0x31d)[0x5bcfcd]
/lib64/libpthread.so.0[0x351660e7c0]
/lib64/libc.so.6(gsignal+0x35)[0x3515a30265]
/lib64/libc.so.6(abort+0x110)[0x3515a31d10]
/u01/mysql/lib/mysql/plugin/ha_innodb_plugin.so[0x2aab2ca1630f]
/u01/mysql/lib/mysql/plugin/ha_innodb_plugin.so[0x2aab2ca0bc24]
/u01/mysql/lib/mysql/plugin/ha_innodb_plugin.so[0x2aab2ca06b5a]
/u01/mysql/lib/mysql/plugin/ha_innodb_plugin.so[0x2aab2ca32b0c]
/u01/mysql/libexec/mysqld(_ZN7handler7ha_openEP8st_tablePKcii+0x3e)[0x6b760e]
/u01/mysql/libexec/mysqld(_Z21open_table_from_shareP3THDP14st_table_sharePKcjjjP8st_tableb+0x592)[0x61b012]
/u01/mysql/libexec/mysqld[0x613747]
/u01/mysql/libexec/mysqld(_Z10open_tableP3THDP10TABLE_LISTP11st_mem_rootPbj+0x716)[0x616036]
/u01/mysql/libexec/mysqld(_Z11open_tablesP3THDPP10TABLE_LISTPjj+0x627)[0x616c27]
/u01/mysql/libexec/mysqld(_Z28open_and_lock_tables_derivedP3THDP10TABLE_LISTb+0x37)[0x616fd7]
/u01/mysql/libexec/mysqld[0x5c7b2e]
/u01/mysql/libexec/mysqld(_Z21mysql_execute_commandP3THD+0x33e1)[0x5ced01]
/u01/mysql/libexec/mysqld(_Z11mysql_parseP3THDPKcjPS2_+0x20b)[0x5d3a2b]
/u01/mysql/libexec/mysqld(_Z16dispatch_command19enum_server_commandP3THDPcj+0x3e0)[0x5d3e10]
/u01/mysql/libexec/mysqld(_Z10do_commandP3THD+0xe6)[0x5d5066]
/u01/mysql/libexec/mysqld(handle_one_connection+0x5c6)[0x5c55c6]
/lib64/libpthread.so.0[0x35166064a7]
/lib64/libc.so.6(clone+0x6d)[0x3515ad3c2d]
Trying to get some variables.
Some pointers may be invalid and cause the dump to abort...
thd->query at 0x1eec18b0 = select * from tmp
thd->thread_id=10
thd->killed=NOT_KILLED
The manual page at http://dev.mysql.com/doc/mysql/en/crashing.html contains
information that should help you find out what is causing the crash.
120203 21:14:35 mysqld_safe Number of processes running now: 0
120203 21:14:35 mysqld_safe mysqld restarted
...
InnoDB: The InnoDB memory heap is disabled
InnoDB: Mutexes and rw_locks use GCC atomic builtins
InnoDB: Compressed tables use zlib 1.2.3
120203 21:14:35  InnoDB: highest supported file format is Barracuda.
InnoDB: The log sequence number in ibdata files does not match
InnoDB: the log sequence number in the ib_logfiles!
120203 21:14:35  InnoDB: Database was not shut down normally!
InnoDB: Starting crash recovery.
InnoDB: Reading tablespace information from the .ibd files...
120203 21:14:35  InnoDB: Warning: allocated tablespace 343, old maximum was 0
InnoDB: Restoring possible half-written data pages from the doublewrite
InnoDB: buffer...
InnoDB: Last MySQL binlog file position 0 302421247, file name /u01/mysql/log/mysql-bin.000340
120203 21:14:36 InnoDB Plugin 1.0.9 started; log sequence number 16439604278
120203 21:14:36 [Note] Recovering after a crash using /u01/mysql/log/mysql-bin
120203 21:14:36 [Note] Starting crash recovery...
120203 21:14:36 [Note] Crash recovery finished.
120203 21:14:36 [Note] Event Scheduler: Loaded 0 events
120203 21:14:36 [Note] /u01/mysql/libexec/mysqld: ready for connections.
Version: '5.1.48-log'  socket: '/u01/mysql/run/mysql.sock'  port: 3306  Source distribution

--EOF--

原参数配置如下：

root@test 01:47:31>show variables like '%query_cache%';
+------------------------------+----------+
| Variable_name | Value |
+------------------------------+----------+
| have_query_cache | YES |
| query_cache_limit | 1024 |
| query_cache_min_res_unit | 1024 |
| query_cache_size | 31457280 |
| query_cache_type | DEMAND |
| query_cache_wlock_invalidate | OFF |
+------------------------------+----------+
6 rows in set (0.00 sec)

由于参数query_cache_min_res_unit值设置和query_cache_limit值设置一样大小，导致query_cache_type=DEMAND时候，就是SQL中添加SQL_CACHE选项，还是不缓存。

下面测试看到Qcache_hits值一直没有增加，也就是不能命中query cache ：

root@test 02:00:20>flush status;
Query OK, 0 rows affected (0.00 sec)

root@test 02:00:27>show global status like '%qcache%';
+-------------------------+----------+
| Variable_name | Value |
+-------------------------+----------+
| Qcache_free_blocks | 1 |
| Qcache_free_memory | 31437704 |
| Qcache_hits | 0 |
| Qcache_inserts | 0 |
| Qcache_lowmem_prunes | 0 |
| Qcache_not_cached | 23 |
| Qcache_queries_in_cache | 1 |
| Qcache_total_blocks | 4 |
+-------------------------+----------+
8 rows in set (0.00 sec)

root@test 02:00:30>select sql_cache * from test where id = 2 ;
+----+------+
| id | col |
+----+------+
| 2 | b |
+----+------+
1 row in set (0.00 sec)

root@test 02:00:52>show global status like '%qcache%';
+-------------------------+----------+
| Variable_name | Value |
+-------------------------+----------+
| Qcache_free_blocks | 1 |
| Qcache_free_memory | 31437704 |
| Qcache_hits | 0 |
| Qcache_inserts | 0 |
| Qcache_lowmem_prunes | 0 |
| Qcache_not_cached | 256 |
| Qcache_queries_in_cache | 1 |
| Qcache_total_blocks | 4 |
+-------------------------+----------+
8 rows in set (0.00 sec)

root@test 02:00:56>select sql_cache * from test where id = 2 ;
+----+------+
| id | col |
+----+------+
| 2 | b |
+----+------+
1 row in set (0.01 sec)

root@test 02:00:59>show global status like '%qcache%';
+-------------------------+----------+
| Variable_name | Value |
+-------------------------+----------+
| Qcache_free_blocks | 1 |
| Qcache_free_memory | 31437704 |
| Qcache_hits | 0 |
| Qcache_inserts | 0 |
| Qcache_lowmem_prunes | 0 |
| Qcache_not_cached | 308 |
| Qcache_queries_in_cache | 1 |
| Qcache_total_blocks | 4 |
+-------------------------+----------+
8 rows in set (0.00 sec)

设置参数query_cache_min_res_unit值小于和query_cache_limit值的时候，查询就可以命中了，可以看到第一次查询之后Qcache_inserts+1，第二次查询之后命中Qcache_hits+1 ：

root@test 02:03:43>set global query_cache_min_res_unit = 1000;
Query OK, 0 rows affected (0.00 sec)

root@test 02:03:51>show variables like '%query_cache%';
+------------------------------+----------+
| Variable_name | Value |
+------------------------------+----------+
| have_query_cache | YES |
| query_cache_limit | 1024 |
| query_cache_min_res_unit | 1000 |
| query_cache_size | 31457280 |
| query_cache_type | DEMAND |
| query_cache_wlock_invalidate | OFF |
+------------------------------+----------+
6 rows in set (0.00 sec)

root@test 02:04:14>show global status like '%qcache%';
+-------------------------+----------+
| Variable_name | Value |
+-------------------------+----------+
| Qcache_free_blocks | 1 |
| Qcache_free_memory | 31437704 |
| Qcache_hits | 0 |
| Qcache_inserts | 0 |
| Qcache_lowmem_prunes | 0 |
| Qcache_not_cached | 1965 |
| Qcache_queries_in_cache | 1 |
| Qcache_total_blocks | 4 |
+-------------------------+----------+
8 rows in set (0.00 sec)

root@test 02:04:14>select sql_cache * from test where id = 2 ;
+----+------+
| id | col |
+----+------+
| 2 | b |
+----+------+
1 row in set (0.00 sec)

root@test 02:04:20>show global status like '%qcache%';
+-------------------------+----------+
| Variable_name | Value |
+-------------------------+----------+
| Qcache_free_blocks | 1 |
| Qcache_free_memory | 31436192 |
| Qcache_hits | 0 |
| Qcache_inserts | 1 |
| Qcache_lowmem_prunes | 0 |
| Qcache_not_cached | 2145 |
| Qcache_queries_in_cache | 2 |
| Qcache_total_blocks | 6 |
+-------------------------+----------+
8 rows in set (0.00 sec)

root@test 02:04:32>select sql_cache * from test where id = 2 ;
+----+------+
| id | col |
+----+------+
| 2 | b |
+----+------+
1 row in set (0.00 sec)

root@test 02:04:36>show global status like '%qcache%';
+-------------------------+----------+
| Variable_name | Value |
+-------------------------+----------+
| Qcache_free_blocks | 1 |
| Qcache_free_memory | 31436192 |
| Qcache_hits | 1 |
| Qcache_inserts | 1 |
| Qcache_lowmem_prunes | 0 |
| Qcache_not_cached | 2175 |
| Qcache_queries_in_cache | 2 |
| Qcache_total_blocks | 6 |
+-------------------------+----------+
8 rows in set (0.00 sec)

另外，对于query_cache_type=DEMAND的时候，即使SQL不加SQL_CACHE选项，也是需要到query cache中去查询一下，具体如下：

root@test 02:04:37>show global variables like '%query_cache%';
+------------------------------+----------+
| Variable_name | Value |
+------------------------------+----------+
| have_query_cache | YES |
| query_cache_limit | 1024 |
| query_cache_min_res_unit | 1000 |
| query_cache_size | 31457280 |
| query_cache_type | DEMAND |
| query_cache_wlock_invalidate | OFF |
+------------------------------+----------+
6 rows in set (0.00 sec)

root@test 02:10:58>set profiling = 1;
Query OK, 0 rows affected (0.00 sec)

root@test 02:11:15>select * from test where id = 2 ;
+----+------+
| id | col |
+----+------+
| 2 | b |
+----+------+
1 row in set (0.00 sec)

root@test 02:11:29>show profile source for query 1;
+--------------------------------+----------+------------------+---------------+-------------+
| Status | Duration | Source_function | Source_file | Source_line |
+--------------------------------+----------+------------------+---------------+-------------+
| starting | 0.000086 | NULL | NULL | NULL |
| checking query cache for query | 0.000058 | unknown function | sql_cache.cc | 1446 |
| Opening tables | 0.000021 | unknown function | sql_base.cc | 4517 |
| System lock | 0.000010 | unknown function | lock.cc | 258 |
| Table lock | 0.000016 | unknown function | lock.cc | 269 |
| init | 0.000032 | unknown function | sql_select.cc | 2484 |
| optimizing | 0.000016 | unknown function | sql_select.cc | 820 |
| statistics | 0.000070 | unknown function | sql_select.cc | 1011 |
| preparing | 0.000021 | unknown function | sql_select.cc | 1033 |
| executing | 0.000010 | unknown function | sql_select.cc | 1755 |
| Sending data | 0.000020 | unknown function | sql_select.cc | 2309 |
| end | 0.000011 | unknown function | sql_select.cc | 2530 |
| query end | 0.000009 | unknown function | sql_parse.cc | 4986 |
| freeing items | 0.000019 | unknown function | sql_parse.cc | 6012 |
| logging slow query | 0.000009 | unknown function | sql_parse.cc | 1687 |
| cleaning up | 0.000009 | unknown function | sql_parse.cc | 1655 |
+--------------------------------+----------+------------------+---------------+-------------+
16 rows in set (0.00 sec)

其中上面status : checking query cache for query 中对应的source_function就是send_result_to_client，在MYSQL源码中有如下信息：

int Query_cache::send_result_to_client(THD *thd, char *sql, uint query_length)
{
...
if (thd->locked_tables || thd->variables.query_cache_type == 0 ||
query_cache_size == 0)
goto err;
...
}

所以,只有query_cache_type=OFF，关掉query cache时候，才不会再去query cache中查询一次，具体如下：

root@test 02:14:20>show variables like '%query_cache%';
+------------------------------+----------+
| Variable_name | Value |
+------------------------------+----------+
| have_query_cache | YES |
| query_cache_limit | 1024 |
| query_cache_min_res_unit | 1000 |
| query_cache_size | 31457280 |
| query_cache_type | OFF |
| query_cache_wlock_invalidate | OFF |
+------------------------------+----------+
6 rows in set (0.00 sec)

root@test 02:14:31>set profiling=1;
Query OK, 0 rows affected (0.00 sec)

root@test 02:14:33>select * from test where id = 2 ;
+----+------+
| id | col |
+----+------+
| 2 | b |
+----+------+
1 row in set (0.00 sec)

root@test 02:14:52>show profile source for query 1;
+--------------------+----------+------------------+---------------+-------------+
| Status | Duration | Source_function | Source_file | Source_line |
+--------------------+----------+------------------+---------------+-------------+
| starting | 0.000125 | NULL | NULL | NULL |
| Opening tables | 0.000035 | unknown function | sql_base.cc | 4517 |
| System lock | 0.000012 | unknown function | lock.cc | 258 |
| Table lock | 0.000016 | unknown function | lock.cc | 269 |
| init | 0.000034 | unknown function | sql_select.cc | 2484 |
| optimizing | 0.000017 | unknown function | sql_select.cc | 820 |
| statistics | 0.000066 | unknown function | sql_select.cc | 1011 |
| preparing | 0.000023 | unknown function | sql_select.cc | 1033 |
| executing | 0.000010 | unknown function | sql_select.cc | 1755 |
| Sending data | 0.000022 | unknown function | sql_select.cc | 2309 |
| end | 0.000011 | unknown function | sql_select.cc | 2530 |
| query end | 0.000009 | unknown function | sql_parse.cc | 4986 |
| freeing items | 0.000035 | unknown function | sql_parse.cc | 6012 |
| logging slow query | 0.000010 | unknown function | sql_parse.cc | 1687 |
| cleaning up | 0.000010 | unknown function | sql_parse.cc | 1655 |
+--------------------+----------+------------------+---------------+-------------+
15 rows in set (0.00 sec)

root@test 02:15:01>show profiles;
+----------+------------+---------------------------------+
| Query_ID | Duration | Query |
+----------+------------+---------------------------------+
| 1 | 0.00043500 | select * from test where id = 2 |
+----------+------------+---------------------------------+
1 row in set (0.00 sec)

【附】
http://www.alidba.net/index.php/archives/423
http://www.mysqlops.com/2011/08/10/mysql-query-cache.html
--EOF--

wget http://prdownloads.sourceforge.net/oprofile/oprofile-0.9.7.tar.gz
tar -zxvf oprofile-0.9.7.tar.gz
cd oprofile-0.9.7
sudo yum install binutils-devel  --说明：configure报错时候需要装一下这个包 (参考这个网址)
./configure
make
sudo make install

2.2 使用

sudo opcontrol --deinit; sudo /sbin/modprobe oprofile timer=1

sudo opcontrol --setup --no-vmlinux && sudo opcontrol --init && sudo opcontrol --reset && sudo opcontrol --start

$mysqlslap --no-defaults --query=" select count(*) as num from vid_film left join vid_class_map on vid_film.id=vid_class_map.v_id and vid_class_map.s_type=1 and vid_class_map.class_id=1;" --number-of-queries=5000 --concurrency=50 -uroot --socket=/u01/mysql/run/mysql.sock

sudo opcontrol --shutdown

opreport -g -l -p /u01/mysql /u01/mysql/libexec/mysqld --merge all -o tmp.log

2.3 遇到的问题
（1）使用opreport报错：

1、error: no sample files found: profile specification too strict
2、opreport error: No sample file found: try running opcontrol --dump or specify a session containing sample files

的时候，需要删除下面配置

sudo rm -f /root/.oprofile/daemonrc

参考：http://blog.yufeng.info/archives/1283

（2）报告的结果无具体符号表相关信息

$more tmp.log
CPU: CPU with timer interrupt, speed 0 MHz (estimated)
Profiling through timer interrupt
samples % linenr info symbol name
16823 100.000 (no location information) /u01/mysql/libexec/mysqld

问题：这个时候只看到mysqld，看不到线程具体跑的函数统计。通过命令( objdump -t /u01/mysql/libexec/mysqld | head )查看是无符号表。
原因：这个mysql是rpm直接安装的，虽然在configure时候也是有加-g，rpmbuild时候问题
解决：源码重新编译一下，重新跑上面的就OK了，结果如下：

$head -n15 tmp.log
CPU: CPU with timer interrupt, speed 0 MHz (estimated)
Profiling through timer interrupt
samples % linenr info symbol name
4589 16.4628 mf_keycache.c:1666 find_key_block
4123 14.7910 mf_keycache.c:2542 key_cache_read
3781 13.5641 my_handler.c:124 ha_key_cmp
3554 12.7498 bmove512.c:40 bmove512
2403 8.6206 mf_keycache.c:1349 unreg_request
2238 8.0287 mi_search.c:184 _mi_bin_search
679 2.4359 mi_search.c:64 _mi_search
618 2.2170 mi_key.c:220 _mi_pack_key
615 2.2063 sql_select.cc:11896 join_read_key(st_join_table*)
469 1.6825 sql_select.cc:11345 sub_select(JOIN*, st_join_table*, bool)
466 1.6717 sql_select.cc:11413 evaluate_join_record(JOIN*, st_join_table*, int)
409 1.4673 mi_key.c:468 _mi_read_key_record

--EOF--

MySQL 5.1中Innodb存储引擎（Built-in-InnoDB）提供了compact和redundant两种格式来记录行数据，redundant是为了兼容之前版本保留的。默认情况下保存的格式为compact格式。可以通过命令show table staus like 'table_name'查看当前表使用的行格式（对应查询结果中的Row_format值）。

对于compact行记录模式：不管char还是varchar类型，NULL值都不占用存储空间。

对于redundant行记录模式：varchar类型的NUL值不占用存储空间，而char类型需要占用存储空间。

二、dynamic和compressed行记录模式

Innodb plugin引入新的文件格式Barracuda，并将以前支持的compact和redundant行记录模式称为Antelope文件格式。

Barracuda文件格式包含两种新的行记录格式：Dynamic和Compressed。

这两种文件格式在处理行溢出时候，Antelope会存放768个前缀字节，Barracuda只会存放20个字节的指针，见下图。

同时，选择文件格式对应的参数innodb_file_format，默认为Antelope。

其他详细的可以参考：http://dev.mysql.com/doc/innodb/1.1/en/innodb-row-format-dynamic.html

再来一张图：

上面的笔记来自：

1、《MySQL技术内幕InnoDB存储引擎》

2、《InnoDB Internals: InnoDB File Formats and Source Code Structure》

三、问题解决

在一次机器升级中涉及到数据库迁移，MYSQL版本都是5.1.x，从一个没有装innodb plugin的版本的数据库A通过mysqldump出来，恢复到装有innodb plugin版本的数据库B中其中一张表报错，B中报错类似如下：

root@test 04:10:58>create table tmp_zx ( a int) engine = innodb row_format = dynamic;
ERROR 1005 (HY000): Can't create table 'test.tmp_zx' (errno: 1478 )

原因是B中innodb_file_format中默认值为默认的Antelope：

root@test 04:17:58>show variables where variable_name in ('innodb_file_per_table','innodb_file_format','innodb_version');
+-----------------------+----------+
| Variable_name         | Value    |
+-----------------------+----------+
| innodb_file_format    | Antelope |
| innodb_file_per_table | ON       |
| innodb_version        | 1.0.9    |
+-----------------------+----------+
3 rows in set (0.00 sec)

动态修改为innodb_file_format 为 Barracuda 就可以。

但是原来A中没有装plugin怎么能有row_format = dynamic的表？

其实原因在于开发创建时候，虽然选择row_format = dynamic但是实际上转换为Compact（可以参考说明：http://dev.mysql.com/doc/innodb-plugin/1.0/en/innodb-compression-syntax-warnings.html），但是在Create_options: row_format=DYNAMIC中保留这个选项，mysqldump导出表结构时候就保留原来创建时候的选项。在A中情况如下：

root@localhost 16:25:40>show variables where variable_name in ('innodb_file_format','innodb_version');
Empty set (0.00 sec)

root@localhost 16:25:42>create table tmp_zx ( a int) engine = innodb row_format = dynamic;
Query OK, 0 rows affected (0.20 sec)

root@localhost 16:25:50>show table status like 'tmp_zx'\G
*************************** 1. row ***************************
Name: tmp_zx
Engine: InnoDB
Version: 10
Row_format: Compact
Rows: 0
Avg_row_length: 0
Data_length: 16384
Max_data_length: 0
Index_length: 0
Data_free: 0
Auto_increment: NULL
Create_time: 2011-07-24 16:25:50
Update_time: NULL
Check_time: NULL
Collation: gbk_chinese_ci
Checksum: NULL
Create_options: row_format=DYNAMIC
Comment:
1 row in set (0.00 sec)

root@localhost 16:25:56>show create table tmp_zx\G
*************************** 1. row ***************************
Table: tmp_zx
Create Table: CREATE TABLE `tmp_zx` (
`a` int(11) DEFAULT NULL
) ENGINE=InnoDB DEFAULT CHARSET=gbk ROW_FORMAT=DYNAMIC
1 row in set (0.00 sec)

--EOF--

记录一下连接MySQL的脚本，供以后参考：

#!/bin/env perl

use strict;

use warnings;

use DBI;

# -------------------------------------------------------------------

# Func   : Main()

# -------------------------------------------------------------------

my $user   = "root";

my $passwd = "";

my $host   = "10.254.5.151";

my $db     = "test";

my $port   = 3306;

# connect database

my $dbh = &connet_mysql($host,$port,$db,$user,$passwd);

# execute sql

my $sql = qq{select * from tmp_zx limit 10};

my $sth = &execute_sql($dbh,$sql);

# get sql result

while ( my $result_ref = $sth->fetchrow_hashref() ) {

my $delim = "";

foreach ( keys %{$result_ref} ) {

print $delim,$_,"=",$result_ref->{$_};

$delim = " , ";

}

print "\n";

}

# end sql

$sth->finish();

# disconnect from database

$dbh->disconnect();

# -------------------------------------------------------------------

# Func   : Connect Database

# Sample :

#          &connet_mysql($host,$port,$db,$user,$passwd);

# -------------------------------------------------------------------

sub connet_mysql {

# setup database connection variables

my ($host,$port,$db,$user,$passwd) = @_;

my $driver = "mysql";

# connect to database

my $dsn = "DBI:$driver:database=$db;host=$host;port=$port";

my $mysql_dbh = DBI->connect($dsn,$user,$passwd) or die "Connect to mysql database error:". DBI->errstr;

$mysql_dbh->{AutoCommit} = 0;

$mysql_dbh->{RaiseError} = 1;

$mysql_dbh->{PrintError} = 1;

$mysql_dbh->do("set names gbk");

return $mysql_dbh;

}

# -------------------------------------------------------------------

# Func   : Execute SQL

# Sample :

#          &execute_sql($dbh,$sql);

# -------------------------------------------------------------------

sub execute_sql {

my ($dbh, $sql) = @_;

my $sth = $dbh->prepare($sql);

$sth->execute();

return $sth;

}

【附】Perl的引用：http://search.cpan.org/~nwclark/perl-5.8.6/pod/perlreftut.pod

--EOF--

#!/usr/local/bin/perl

use Term::ANSIColor;

use strict;

use warnings;

$SIG{CHLD} = 'IGNORE'; #忽略SIGCHLD信号,回收僵尸子进程

print color('green'),"Program started... pid = $$ \n\n";

print "fork process...\n";

# 如果fork失败，则返回undefined

defined ( my $child = fork() ) or die "Fork Error : $!\n" ;

# 如果fork成功:

# 如果是父进程，则把生成的子进程PID赋值给$child。

# 如果是子进程，则把0赋值给$child。

if ($child == 0){

print color('yellow'),"\n-> child process running ...\n";

print "child process pid : $$ !\n";

# 退出子进程，如果前面没有对子进程处理的话，那么在父进程还没退出期间，子进程就变成僵尸子进程。

exit 0;

} else {

print color('blue'),"\n-> parent process running ...\n";

print "child process pid : $child ; parent process pid : $$ !\n",color('reset');

}

#sleep(10);

print "\nHello World!\n";

执行结果如下：

[zhuxu@xentest8-vm1 my_perl_test]$ ./fork.pl

Program started... pid = 30881

fork process...

-> child process running ...

child process pid : 30882 !

-> parent process running ...

child process pid : 30882 ; parent process pid : 30881 !

Hello World!

[zhuxu@xentest8-vm1 my_perl_test]$

【参考】：Perl fork()

--EOF--

（1）在远程客户端（需要添加选项：--local-infile=1）导入远程客户端文本到MySQL，需指定LOCAL（默认就是ignore）,加ignore选项会放弃数据，加replace选项会更新数据，都不会出现唯一性约束问题。

（2）在本地服务器导入本地服务器文本到MySQL，不指定LOACL，出现唯一性约束冲突，会失败回滚，数据导入不进去，这个时候就需要加ignore或者replace来导入数据。

测试如下：

（1）本地服务器导入本地服务器文本

mysql> show create table tmp_loaddata\G;

*************************** 1. row ***************************

Table: tmp_loaddata

Create Table: CREATE TABLE `tmp_loaddata` (

`id` int(11) NOT NULL,

`name` varchar(10) DEFAULT NULL,

PRIMARY KEY (`id`)

) ENGINE=InnoDB DEFAULT CHARSET=latin1

1 row in set (0.00 sec)

ERROR:

No query specified

mysql> select * from tmp_loaddata;

+----+------+

| id | name |

+----+------+

| 1 | test |

+----+------+

1 row in set (0.00 sec)

mysql>

mysql> system cat /home/zhuxu/1.txt

1,new update

2,new update

mysql>

mysql> LOAD DATA INFILE '/home/zhuxu/1.txt' INTO TABLE tmp_loaddata FIELDS TERMINATED BY ',';

ERROR 1062 (23000): Duplicate entry '1' for key 'PRIMARY'

# 出现唯一性约束冲突，会失败回滚

mysql> select * from tmp_loaddata;

+----+------+

| id | name |

+----+------+

| 1 | test |

+----+------+

1 row in set (0.00 sec)

mysql> LOAD DATA INFILE '/home/zhuxu/1.txt' IGNORE INTO TABLE tmp_loaddata FIELDS TERMINATED BY ',';

Query OK, 1 row affected (0.00 sec)

Records: 2 Deleted: 0 Skipped: 1 Warnings: 0

# 使用IGNORE对于冲突的数据丢弃掉。

mysql> select * from tmp_loaddata;

+----+------------+

| id | name |

+----+------------+

| 1 | test |

| 2 | new update |

+----+------------+

2 rows in set (0.00 sec)

mysql> LOAD DATA INFILE '/home/zhuxu/1.txt' REPLACE INTO TABLE tmp_loaddata FIELDS TERMINATED BY ',';

Query OK, 3 rows affected (0.00 sec)

Records: 2 Deleted: 1 Skipped: 0 Warnings: 0

# 使用REPLACE对于冲突的数据进行更新。

mysql> select * from tmp_loaddata;

+----+------------+

| id | name |

+----+------------+

| 1 | new update |

| 2 | new update |

+----+------------+

2 rows in set (0.00 sec)

（2）远程客户端导入远程客户端文本

[zhuxu@xentest9-vm1 tmp]$ mysql -uzhuxu -pzhuxu test -h10.254.5.151

Welcome to the MySQL monitor. Commands end with ; or \g.

Your MySQL connection id is 15

Server version: 5.1.47-log Source distribution

Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.

This software comes with ABSOLUTELY NO WARRANTY. This is free software,

and you are welcome to modify and redistribute it under the GPL v2 license

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql> select * from tmp_loaddata;

+----+------+

| id | name |

+----+------+

| 1 | test |

+----+------+

1 row in set (0.00 sec)

mysql> system cat /tmp/2.txt

1,new update

2,new update

3,new update

mysql>

mysql> LOAD DATA INFILE '/tmp/2.txt' INTO TABLE tmp_loaddata FIELDS TERMINATED BY ',';

ERROR 13 (HY000): Can't get stat of '/tmp/2.txt' (Errcode: 2)

# 由于数据库服务器没有对应的文本文件，所以报错。

mysql>

mysql> LOAD DATA LOCAL INFILE '/tmp/2.txt' INTO TABLE tmp_loaddata FIELDS TERMINATED BY ',';

ERROR 1148 (42000): The used command is not allowed with this MySQL version

# 进去mysql远程客户端，还需要加--local-infile=1参数指定。

mysql> exit

Bye

[zhuxu@xentest9-vm1 tmp]$ mysql -uzhuxu -pzhuxu test -h10.254.5.151 --local-infile=1 --show-warnings -v -v -v \

> -e "LOAD DATA LOCAL INFILE '/tmp/2.txt' INTO TABLE tmp_loaddata FIELDS TERMINATED BY ','";

--------------

LOAD DATA LOCAL INFILE '/tmp/2.txt' INTO TABLE tmp_loaddata FIELDS TERMINATED BY ','

--------------

Query OK, 2 rows affected (0.00 sec)

Records: 3 Deleted: 0 Skipped: 1 Warnings: 0

Bye

mysql> select * from tmp_loaddata;

+----+------------+

| id | name |

+----+------------+

| 1 | test |

| 2 | new update |

| 3 | new update |

+----+------------+

3 rows in set (0.00 sec)

#

[zhuxu@xentest9-vm1 tmp]$ mysql -uzhuxu -pzhuxu test -h10.254.5.151 --local-infile=1 --show-warnings -v -v -v \

> -e "LOAD DATA LOCAL INFILE '/tmp/2.txt' IGNORE INTO TABLE tmp_loaddata FIELDS TERMINATED BY ','";

--------------

LOAD DATA LOCAL INFILE '/tmp/2.txt' IGNORE INTO TABLE tmp_loaddata FIELDS TERMINATED BY ','

--------------

Query OK, 0 rows affected (0.00 sec)

Records: 3 Deleted: 0 Skipped: 3 Warnings: 0

Bye

mysql> select * from tmp_loaddata;

+----+------------+

| id | name |

+----+------------+

| 1 | test |

| 2 | new update |

| 3 | new update |

+----+------------+

3 rows in set (0.00 sec)

#

[zhuxu@xentest9-vm1 tmp]$ mysql -uzhuxu -pzhuxu test -h10.254.5.151 --local-infile=1 --show-warnings -v -v -v \

> -e "LOAD DATA LOCAL INFILE '/tmp/2.txt' REPLACE INTO TABLE tmp_loaddata FIELDS TERMINATED BY ','";

--------------

LOAD DATA LOCAL INFILE '/tmp/2.txt' REPLACE INTO TABLE tmp_loaddata FIELDS TERMINATED BY ','

--------------

Query OK, 4 rows affected (0.00 sec)

Records: 3 Deleted: 1 Skipped: 0 Warnings: 0

Bye

mysql> select * from tmp_loaddata;

+----+------------+

| id | name |

+----+------------+

| 1 | new update |

| 2 | new update |

| 3 | new update |

+----+------------+

3 rows in set (0.00 sec)

--EOF--

1.1

在MySQL的show status中有2个状态值：Open_tables和Opened_tables。这2个值代表的意思如下：

Open_tables  ：代表当前打开表的数量。

Opened_tables：代表自从MySQL启动后，打开表的数量。

关于MySQL怎么打开关闭表的具体细节参考文档：<How MySQL Opens and Closes Tables>。

（1）对于myisam存储引擎，打开1张表需要2个文件描述符（一个.MYD文件，一个.MYI文件）。

（2）对于innodb存储引擎，开启表的独立表空间（innodb_file_per_table）打开1张表只需要1个文件描述符（一个.ibd文件）。

【MySQL Variable】

对于上面的状态值，对应的5.1.3版本后的MySQL变量参数为table_open_cache，而早期版本为：table_cache，该参数值的代表MySQL可以缓存的打开表时候的最大文件描述符。

1.2

在MySQL 5.1.3之后，还添加了2个状态值：Open_table_definitions和Opened_table_definitions。这2个值代表的意思如下：

Open_table_definitions  ：代表当前缓存了多少.frm文件。

Opened_table_definitions：代表自从MySQL启动后，缓存了.frm文件的数量。

需要注意的是.frm文件是MySQL用于存放表结构的文件，对应myisam和innodb存储引擎都必须有的。

【MySQL Variable】

对于上面的状态值，对应的5.1.3版本后的MySQL变量参数为table_definition_cache，该参数值的代表MySQL可以缓存的表定义的数量。和前面的table cache不同的是，表定义的缓存占用空间很小，而且不需要使用文件描述符，也就是只要打开.frm文件，缓存表定义，然后就可以关闭.frm文件。

1.3

另外，还有2个状态值：Open_files和Opened_files。这2个值的意思同上类似：

Open_files  ：代表当前打开的文件。对应存储引擎（如：innodb）使用存储引擎自己内部函数打开的话，这个值是不会增加的。

Opened_files：代表使用MySQL的my_open()函数打开过的文件数。如果不是使用这个函数打开文件的话，这个值是不会增加的。

【MySQL Variable】

对于上面的状态值，对应的MySQL变量参数为open_files_limit。

二、测试

测试的MySQL版本：

mysql> select version();

+------------+

| version()  |

+------------+

| 5.1.47-log |

+------------+

1 row in set (0.00 sec)

2.1 myisam存储引擎

2.1.1

重启MySQL服务器，可以看到启动时候MySQL已经自动打开相关的系统表。

mysql> show global status like 'Open%_table_definitions';show global status like 'Open%_tables';show global status like 'Open%_files';

+--------------------------+-------+

| Variable_name            | Value |

+--------------------------+-------+

| Open_table_definitions   | 15    |

| Opened_table_definitions | 15    |

+--------------------------+-------+

2 rows in set (0.00 sec)

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_tables   | 8     |

| Opened_tables | 15    |

+---------------+-------+

2 rows in set (0.00 sec)

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_files    | 19    |

| Opened_files  | 62    |

+---------------+-------+

2 rows in set (0.00 sec)

在上面的状态输出，可以看到Open_tables的值为8，代表了当前打开了8张表，同样的通过mysqladmin status命令也可以看到：

[mysql@xentest8-vm1 test]$ mysqladmin status

Uptime: 631  Threads: 2  Questions: 14  Slow queries: 0  Opens: 15  Flush tables: 1  Open tables: 8 Queries per second avg: 0.22

可以通过SHOW OPEN TABLES命令查看，当前的table cache到底打开了哪些表：

mysql> show open tables;

+----------+--------------+--------+-------------+

| Database | Table        | In_use | Name_locked |

+----------+--------------+--------+-------------+

| mysql    | servers      |      0 |           0 |

| mysql    | db           |      0 |           0 |

| mysql    | host         |      0 |           0 |

| mysql    | columns_priv |      0 |           0 |

| mysql    | user         |      0 |           0 |

| mysql    | procs_priv   |      0 |           0 |

| mysql    | event        |      0 |           0 |

| mysql    | tables_priv  |      0 |           0 |

+----------+--------------+--------+-------------+

8 rows in set (0.00 sec)

可以看到打开了mysql库里面的8张系统表（都是myisam存储引擎），另外，还可以通过lsof（lsof的用法参见<使用lsof查找打开的文件>）查看MySQL打开的对应文件：

[zhuxu@xentest8-vm1 ~]$ sudo lsof -u mysql | grep /home/mysql/mysql-5.1.47/data/mysql/

mysqld    14683 mysql   16u   REG              253,1      2048 87889757 /home/mysql/mysql-5.1.47/data/mysql/host.MYI

mysqld    14683 mysql   17u   REG              253,1         0 87889758 /home/mysql/mysql-5.1.47/data/mysql/host.MYD

mysqld    14683 mysql   18u   REG              253,1      2048 87889760 /home/mysql/mysql-5.1.47/data/mysql/user.MYI

mysqld    14683 mysql   19u   REG              253,1       380 87889761 /home/mysql/mysql-5.1.47/data/mysql/user.MYD

mysqld    14683 mysql   20u   REG              253,1      5120 87889754 /home/mysql/mysql-5.1.47/data/mysql/db.MYI

mysqld    14683 mysql   21u   REG              253,1       880 87889755 /home/mysql/mysql-5.1.47/data/mysql/db.MYD

mysqld    14683 mysql   22u   REG              253,1      4096 87889772 /home/mysql/mysql-5.1.47/data/mysql/tables_priv.MYI

mysqld    14683 mysql   23u   REG              253,1         0 87889773 /home/mysql/mysql-5.1.47/data/mysql/tables_priv.MYD

mysqld    14683 mysql   24u   REG              253,1      4096 87889775 /home/mysql/mysql-5.1.47/data/mysql/columns_priv.MYI

mysqld    14683 mysql   25u   REG              253,1         0 87889776 /home/mysql/mysql-5.1.47/data/mysql/columns_priv.MYD

mysqld    14683 mysql   26u   REG              253,1      4096 87889808 /home/mysql/mysql-5.1.47/data/mysql/procs_priv.MYI

mysqld    14683 mysql   27u   REG              253,1         0 87889809 /home/mysql/mysql-5.1.47/data/mysql/procs_priv.MYD

mysqld    14683 mysql   28u   REG              253,1      1024 87889769 /home/mysql/mysql-5.1.47/data/mysql/servers.MYI

mysqld    14683 mysql   29u   REG              253,1         0 87889770 /home/mysql/mysql-5.1.47/data/mysql/servers.MYD

mysqld    14683 mysql   30u   REG              253,1      2048 87889817 /home/mysql/mysql-5.1.47/data/mysql/event.MYI

mysqld    14683 mysql   31u   REG              253,1         0 87889818 /home/mysql/mysql-5.1.47/data/mysql/event.MYD

2.1.2

接着打开一张myisam存储引擎的表：

mysql> select count(*) from myisam_1;

+----------+

| count(*) |

+----------+

|        0 |

+----------+

1 row in set (0.00 sec)

mysql> show global status like 'Open%_table_definitions';show global status like 'Open%_tables';show global status like 'Open%_files';

# 可以看到Opened_table_definitions/Opened_table_definitions/Open_tables/Opened_tables的值都加1

+--------------------------+-------+

| Variable_name            | Value |

+--------------------------+-------+

| Open_table_definitions   | 16    |

| Opened_table_definitions | 16    |

+--------------------------+-------+

2 rows in set (0.00 sec)

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_tables   | 9     |

| Opened_tables | 16    |

+---------------+-------+

2 rows in set (0.00 sec)

# 可以看到Open_files的值（19+2）加2，Opened_files的值（62+3）加3

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_files    | 21    |

| Opened_files  | 65    |

+---------------+-------+

2 rows in set (0.00 sec)

mysql> show create table myisam_1\G

*************************** 1. row ***************************

Table: myisam_1

Create Table: CREATE TABLE `myisam_1` (

`a` int(11) DEFAULT NULL,

KEY `idx_myisam_1` (`a`)

) ENGINE=MyISAM DEFAULT CHARSET=latin1

1 row in set (0.00 sec)

mysql> show open tables from test;

+----------+----------+--------+-------------+

| Database | Table    | In_use | Name_locked |

+----------+----------+--------+-------------+

| test     | myisam_1 |      0 |           0 |

+----------+----------+--------+-------------+

1 row in set (0.00 sec)

[zhuxu@xentest8-vm1 ~]$ sudo lsof -u mysql | grep /home/mysql/mysql-5.1.47/data/test/

mysqld    14683 mysql  33u REG              253,1      1024 87851510 /home/mysql/mysql-5.1.47/data/test/myisam_1.MYI

mysqld    14683 mysql   34u REG              253,1         0 87851511 /home/mysql/mysql-5.1.47/data/test/myisam_1.MYD

# 可见，除了需要打开.frm文件（打开，缓存后，就关闭），对于myisam存储引擎，打开1张表需要2个文件描述符（一个.MYD文件，一个.MYI文件）

2.1.3

对应table cache中的缓存的表，可以通过flush tables命令来把它人工移出table cache:

mysql> flush tables myisam_1;

Query OK, 0 rows affected (0.00 sec)

mysql> show open tables from test;

Empty set (0.00 sec)

mysql> show global status like 'Open%_table_definitions';show global status like 'Open%_tables';show global status like 'Open%_files';

# 可以看到Open_table_definitions和Open_tables的值都减1

+--------------------------+-------+

| Variable_name            | Value |

+--------------------------+-------+

| Open_table_definitions   | 15    |

| Opened_table_definitions | 16    |

+--------------------------+-------+

2 rows in set (0.00 sec)

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_tables   | 8     |

| Opened_tables | 16    |

+---------------+-------+

2 rows in set (0.00 sec)

# 可以看到Open_files的值（21-2）减2

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_files    | 19    |

| Opened_files  | 65 |

+---------------+-------+

2 rows in set (0.00 sec)

[zhuxu@xentest8-vm1 ~]$ sudo lsof -u mysql | grep /home/mysql/mysql-5.1.47/data/test/ | wc -l

0

2.2 innodb存储引擎

2.2.1

打开一张innodb存储引擎的表：

mysql> show variables like 'innodb_file_per_table';

+-----------------------+-------+

| Variable_name         | Value |

+-----------------------+-------+

| innodb_file_per_table | ON    |

+-----------------------+-------+

1 row in set (0.00 sec)

mysql> select count(*) from t0001;

+----------+

| count(*) |

+----------+

|        2 |

+----------+

1 row in set (0.01 sec)

mysql> show global status like 'Open%_table_definitions';show global status like 'Open%_tables';show global status like 'Open%_files';

# 可以看到Opened_table_definitions/Opened_table_definitions/Open_tables/Opened_tables的值都加1

+--------------------------+-------+

| Variable_name            | Value |

+--------------------------+-------+

| Open_table_definitions   | 16    |

| Opened_table_definitions | 17    |

+--------------------------+-------+

2 rows in set (0.00 sec)

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_tables   | 9     |

| Opened_tables | 17    |

+---------------+-------+

2 rows in set (0.00 sec)

# 可以看到Open_files的值不变（即innodb使用存储引擎自己内部函数打开的话，这个值是不会增加）

# Opened_files的值（65+1）加1（需要打开.frm文件，缓存后，就关闭）

+---------------+-------+

| Variable_name | Value |

+---------------+-------+

| Open_files    | 19    |

| Opened_files  | 66    |

+---------------+-------+

2 rows in set (0.00 sec)

mysql> show open tables from test;

+----------+-------+--------+-------------+

| Database | Table | In_use | Name_locked |

+----------+-------+--------+-------------+

| test     | t0001 |      0 |           0 |

+----------+-------+--------+-------------+

1 row in set (0.00 sec)

mysql> show create table t0001\G

*************************** 1. row ***************************

Table: t0001

Create Table: CREATE TABLE `t0001` (

`a` int(11) DEFAULT NULL

) ENGINE=InnoDB DEFAULT CHARSET=latin1

1 row in set (0.00 sec)

[zhuxu@xentest8-vm1 ~]$ sudo lsof -u mysql | grep /home/mysql/mysql-5.1.47/data/test/

mysqld    14683 mysql   33uW REG              253,1     98304 87851074 /home/mysql/mysql-5.1.47/data/test/t0001.ibd

2.2.2

对于innodb存储引擎，table cache对其影响其实不是很大，innodb有自己的数据字典，可以缓存相关表信息。

其他相关资料：

（1）在<High.Performance.MySQL>中的The InnoDB Data Dictionary（P280页）中描述：

InnoDB has its own per-table cache, variously called a table definition cache or data dictionary, which you cannot configure. When InnoDB opens a table, it adds a corresponding object to the data dictionary. Each table can take up 4 KB or more of memory (although much less space is required in MySQL 5.1). Tables are not removed from the data dictionary when they are closed.

The main performance issue—besides memory requirements—is opening and computing statistics for the tables, which is expensive because it requires a lot of I/O. In contrast to MyISAM, InnoDB doesn’t store statistics in the tables permanently; it recomputes them each time it starts. This operation is serialized by a global mutex in current versions of MySQL, so it can’t be done in parallel. If you have a lot of tables, your server can take hours to start and fully warm up, during which time it might not be doing much other than waiting for one I/O operation after another. We mention this to make sure you know about it, even though there’s nothing you can do to change it. It’s normally a problem only when you have many (thousands or tens of thousands) large tables, which cause the process to be I/O-bound.

If you use InnoDB’s innodb_file_per_table option (described later in “Configuring the tablespace” on page 291), there’s also a separate limit on the number of .ibd files InnoDB can keep open at any time. This is handled by the InnoDB storage engine, not the MySQL server, and is controlled by innodb_open_files. InnoDB doesn’t open files the same way MyISAM does: whereas MyISAM uses the table cache to hold file descriptors for open tables, in InnoDB there is no direct relationship between open tables and open files. InnoDB uses a single, global file descriptor for each .ibd file. If you can afford it, it’s best to set innodb_open_files large enough that the server can keep all .ibd files open simultaneously.

（2）<How innodb_open_files affects performance>

（3）<SHOW OPEN TABLES – what is in your table cache>里面有一句描述：

Note however if you’re starting MySQL Command line client without “-A” option it opens all tables in the active database to allow tab completion which can screw results.

在实际的测试中，对于5.0.67-log版本会出现上面的情况，对于5.1版本的测试，上面的现象是不存在。

--EOF--

[zhuxu@xentest8-vm1 ~]$ free

total       used       free     shared    buffers     cached

Mem:       4194304    4164812      29492          0     248892    2148968

-/+ buffers/cache:    1766952    2427352

Swap:       524280     343820     180460

[zhuxu@dbadb1 ~]$ watch --help

Usage: watch [-dhntv] [--differences[=cumulative]] [--help] [--interval=<n>] [--no-title] [--version] <command>

-d, --differences[=cumulative]        highlight changes between updates

(cumulative means highlighting is cumulative)

-h, --help                            print a summary of the options

-n, --interval=<seconds>              seconds to wait between updates

-v, --version                         print the version number

-t, --no-title                        turns off showing the header

[zhuxu@dbadb1 ~]$

[zhuxu@dbadb1 ~]$ watch -d -n 0.5 free