关键词：
- 阿里云服务器 MySQL 经常自动停止、挂掉、重启。
- 阿里云CentOS WordPress 运行经常数据库连接不上。
- CentOS + nginx + PHP + MySQL.
打开 MySQL 的 error.log 错误信息，在阿里云 CentOS 的路径为 /alidata/log/mysql/error.log,如下：
2016-03-13 00:16:37 0 [Warning] TIMESTAMP with implicit DEFAULT value is deprecated. Please use --explicit_defaults_for_timestamp server option (see documentation for more details).
2016-03-13 00:16:37 30576 [Note] Plugin 'FEDERATED' is disabled.
2016-03-13 00:16:37 30576 [Note] InnoDB: Using atomics to ref count buffer pool pages
2016-03-13 00:16:37 30576 [Note] InnoDB: The InnoDB memory heap is disabled
2016-03-13 00:16:37 30576 [Note] InnoDB: Mutexes and rw_locks use GCC atomic builtins
2016-03-13 00:16:37 30576 [Note] InnoDB: Memory barrier is not used
2016-03-13 00:16:37 30576 [Note] InnoDB: Compressed tables use zlib 1.2.3
2016-03-13 00:16:37 30576 [Note] InnoDB: Using Linux native AIO
2016-03-13 00:16:37 30576 [Note] InnoDB: Using CPU crc32 instructions
2016-03-13 00:16:37 30576 [Note] InnoDB: Initializing buffer pool, size = 64.0M
2016-03-13 00:16:37 30576 [Note] InnoDB: Completed initialization of buffer pool
2016-03-13 00:16:37 30576 [Note] InnoDB: Highest supported file format is Barracuda.
2016-03-13 00:16:37 30576 [Note] InnoDB: The log sequence numbers 52296883 and 52296883 in ibdata files do not match the log sequence number 93636989 in the ib_logfiles!
2016-03-13 00:16:37 30576 [Note] InnoDB: Database was not shutdown normally!
2016-03-13 00:16:37 30576 [Note] InnoDB: Starting crash recovery.
2016-03-13 00:16:37 30576 [Note] InnoDB: Reading tablespace information from the .ibd files...
2016-03-13 00:16:37 30576 [Note] InnoDB: Restoring possible half-written data pages
2016-03-13 00:16:37 30576 [Note] InnoDB: from the doublewrite buffer...
InnoDB: Last MySQL binlog file position 0 9828, file name mysql-bin.000335
2016-03-13 00:16:37 30576 [Note] InnoDB: 128 rollback segment(s) are active.
2016-03-13 00:16:37 30576 [Note] InnoDB: Waiting for purge to start
2016-03-13 00:16:37 30576 [Note] InnoDB: 5.6.21 started; log sequence number 93636989
2016-03-13 00:16:38 30576 [Note] Recovering after a crash using mysql-bin
2016-03-13 00:16:38 30576 [Note] Starting crash recovery...
2016-03-13 00:16:38 30576 [Note] Crash recovery finished.
2016-03-13 00:16:40 30576 [Note] Server hostname (bind-address): '*'; port: 3306
2016-03-13 00:16:40 30576 [Note] IPv6 is not available.
2016-03-13 00:16:40 30576 [Note] - '0.0.0.0' resolves to '0.0.0.0';
2016-03-13 00:16:40 30576 [Note] Server socket created on IP: '0.0.0.0'.
2016-03-13 00:16:41 30576 [Note] Event Scheduler: Loaded 0 events
2016-03-13 00:16:41 30576 [Note] /alidata/server/mysql/bin/mysqld: ready for connections.
Version: '5.6.21-log' socket: '/tmp/mysql.sock' port: 3306 MySQL Community Server (GPL)
160313 00:16:41 mysqld_safe Number of processes running now: 0
160313 00:16:41 mysqld_safe mysqld restarted
但从 MySQL 的 error.log 中，我们很难找出出是哪里出了问题。去 Google 了一下，给出的答案基本是修改”/alidata/server/mysql/my.cnf”:
innodb_buffer_pool_size = 64M
# 默认为128M，修改成64M、32M或者8M，但实际上重启 MySQL 之后以为能运行，实际上过了一会 MySQL 还是会出现不知原因地自动关闭。
但是，现在注意去看其中的一行：
2016-03-13 00:16:37 30576 [Note] InnoDB: Database was not shutdown normally!
这句话说明了 MySQL 非正常关闭，MySQL 也不知道自己为什么被关闭掉了。。。于是， MySQL 觉得真奇怪，自己又重启了自己的进程开始 Crash Recoverying，如下：
2016-03-13 00:16:37 30576 [Note] InnoDB: Starting crash recovery.
接着，大家又猜想这条日志：
2016-03-13 00:16:37 30576 [Note] InnoDB: The log sequence numbers 52296883 and 52296883 in ibdata files do not match the log sequence number 93636989 in the ib_logfiles!
继续 Google，发现原来就是这里捣鬼，才出现了 MySQL 再重启之后，又继续自动关闭，再重启。于是，尝试按照 StackOverflow 上面的 suggestions 去做：
Drop your ib_log files and Put innodb_force_recovery=6 in config file and restart your mysql it will resolve
重启，又可以了。。。没一会，嚓，MySQL 又宕了！这次不仅宕，而且 error.log 又多了因 ** innodb_force_recovery = 6 ** 而引起的错误。
怎么办？现在我们回到下面这条日志：
2016-03-13 00:16:37 30576 [Note] InnoDB: Database was not shutdown normally!
现在我们要看看，为什么 MySQL 会不正常关闭呢！？现在我们打开top看看。下面使用 ** htop ** 进行查看：
一看 mysql 进程占用的 VIRT 怎么这么多！这时，你要是多刷新几下你服务器上搭建的 WordPress 网站的主页的话，会出现 mysql 内存占用率超过 50% 的情况，这时候 mysql 进程就会被 linux 内核杀死。也就出现了我们在 MySQL 的 error.log 中看到的“[Note] InnoDB: Database was not shutdown normally!” 日志了。
为了确定这个猜测是否真实，我们去看看 kernel 日志。
# 打开系统 log 目录
cd /var/log/
#接着查看 messages-20160313
vim messages-20160313
接着你会看到：
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [ 1598] 0 1598 61593 495 0 0 0 AliYunDun
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [24853] 500 24853 59954 6188 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [24855] 500 24855 59460 5555 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [24856] 500 24856 60347 5472 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [24859] 500 24859 59771 5529 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [24860] 500 24860 59898 5773 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [24861] 500 24861 59924 4836 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [24862] 500 24862 60051 5560 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [25035] 0 25035 11302 30 0 0 0 nginx
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [25037] 500 25037 17793 1220 0 0 0 nginx
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [25038] 500 25038 17881 309 0 0 0 nginx
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [ 4296] 0 4296 199625 547 0 0 0 AliHids
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [11606] 500 11606 60371 6174 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [11608] 500 11608 60058 6237 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [11609] 500 11609 59881 5386 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [11612] 500 11612 59734 5895 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [11613] 500 11613 59410 5886 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [17646] 500 17646 56833 5508 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [17647] 500 17647 59049 5495 0 0 0 php-fpm
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [30706] 0 30706 2868 76 0 0 0 mysqld_safe
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [31748] 0 31748 4346 53 0 0 0 anacron
Mar 13 03:14:50 iZ941u4kh5rZ kernel: [31809] 501 31809 204818 26845 0 0 0 mysqld
Mar 13 03:14:50 iZ941u4kh5rZ kernel: Out of memory: Kill process 31809 (mysqld) score 52 or sacrifice child
Mar 13 03:14:50 iZ941u4kh5rZ kernel: Killed process 31809, UID 501, (mysqld) total-vm:819272kB, anon-rss:103500kB, file-rss:3880kB
看到最后的两条：
Mar 13 03:14:50 iZ941u4kh5rZ kernel: Out of memory: Kill process 31809 (mysqld) score 52 or sacrifice child
Mar 13 03:14:50 iZ941u4kh5rZ kernel: Killed process 31809, UID 501, (mysqld) total-vm:819272kB, anon-rss:103500kB, file-rss:3880kB
好了，终于找出原因了，内存不够，杀死了 mysqld 进程。
这时怎么办呢？
** 下面是我的优化方案： **
1.创建 SWAP 分区：
参见教程：How do you add swap to an EC2 instance?
a.逐条运行下面的命令：
sudo /bin/dd if=/dev/zero of=/var/swap.1 bs=1M count=1024
sudo /sbin/mkswap /var/swap.1
sudo /sbin/swapon /var/swap.1
b.将下面一行添加到 /etc/fstab,服务器重启时自动启动swap：
/var/swap.1 swap swap defaults 0 0
2.降低数据库 InnoDB 引擎的缓冲区大小，以及限制 MySQL 的最大连接数(max_connections)：
在 /alidata/server/mysql/my.cnf 的 mysqld 下添加下面两句：
# 降低 InnoDB 缓冲区大小为 64M 或者 32M
innodb_buffer_pool_size = 64M
#限制最大连接数为100，在服务器配置很低时可以继续降低
max_connections = 100
修改完重启 MySQL：service mysqld restart
注释：max_connections 的默认值是 151，可以动态更改这个值。参见：max_connections
3.nginx 优化：
参见：Optimizing Nginx Configuration
打开 “/alidata/server/nginx/conf/nginx.conf” ：
vim /alidata/server/nginx/conf/nginx.conf
优化配置如下：
worker_processes 1;
worker_connections 256;
修改完重启 nginx：service nginx restart
如有必要，可以重启服务器: reboot
好了，问题解决了。
— EOF —
但行好事，莫问前程。
声明：转载需要注明来源地址。
参考资料：
Optimizing Nginx Configuration
How do you add swap to an EC2 instance?
MySQL crashes randomly with “Database was not shut down normally” message in log