MySQL一直在崩溃

Question

我在运行ubuntu 12.10的云中有一台服务器,只有512Mb的RAM才能运行wordpress网站(每天访问aprox 1000次).

MySQL崩溃然后我启用4Gb交换,看看是否可以工作......但仍然崩溃...我需要每次重启服务...

从mysql检查错误日志我注意到InnoDB似乎处于循环中试图恢复某些东西,但我认为它不能......任何人都可以帮助我吗？

131009 17:56:57 InnoDB: 5.5.32 started; log sequence number 242183133
131009 17:56:57 [Note] Server hostname (bind-address): '127.0.0.1'; port: 3306
131009 17:56:57 [Note]   - '127.0.0.1' resolves to '127.0.0.1';
131009 17:56:57 [Note] Server socket created on IP: '127.0.0.1'.
131009 17:56:57 [Note] Event Scheduler: Loaded 0 events
131009 17:56:57 [Note] /usr/sbin/mysqld: ready for connections.
Version: '5.5.32-0ubuntu0.12.10.1'  socket: '/var/run/mysqld/mysqld.sock'  port: 3306  (Ubuntu)
131009 17:57:25 [Note] Plugin 'FEDERATED' is disabled.
131009 17:57:25 InnoDB: The InnoDB memory heap is disabled
131009 17:57:25 InnoDB: Mutexes and rw_locks use GCC atomic builtins
131009 17:57:25 InnoDB: Compressed tables use zlib 1.2.7
131009 17:57:25 InnoDB: Using Linux native AIO
131009 17:57:25 InnoDB: Initializing buffer pool, size = 128.0M
131009 17:57:25 InnoDB: Completed initialization of buffer pool
131009 17:57:25 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!
131009 17:57:25  InnoDB: Database was not shut down normally!
InnoDB: Starting crash recovery.
InnoDB: Reading tablespace information from the .ibd files...
InnoDB: Restoring possible half-written data pages from the doublewrite
InnoDB: buffer...
131009 17:57:26  InnoDB: Waiting for the background threads to start
131009 17:57:27 InnoDB: 5.5.32 started; log sequence number 242183133
131009 17:57:27 [Note] Server hostname (bind-address): '127.0.0.1'; port: 3306
131009 17:57:27 [Note]   - '127.0.0.1' resolves to '127.0.0.1';
131009 17:57:27 [Note] Server socket created on IP: '127.0.0.1'.
131009 17:57:27 [Note] Event Scheduler: Loaded 0 events
131009 17:57:27 [Note] /usr/sbin/mysqld: ready for connections.
Version: '5.5.32-0ubuntu0.12.10.1'  socket: '/var/run/mysqld/mysqld.sock'  port: 3306  (Ubuntu)
131009 17:58:05 [Note] Plugin 'FEDERATED' is disabled.
131009 17:58:05 InnoDB: The InnoDB memory heap is disabled
131009 17:58:05 InnoDB: Mutexes and rw_locks use GCC atomic builtins
131009 17:58:05 InnoDB: Compressed tables use zlib 1.2.7
131009 17:58:05 InnoDB: Using Linux native AIO
131009 17:58:05 InnoDB: Initializing buffer pool, size = 128.0M
131009 17:58:05 InnoDB: Completed initialization of buffer pool
131009 17:58:06 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!
131009 17:58:06  InnoDB: Database was not shut down normally!
InnoDB: Starting crash recovery.
InnoDB: Reading tablespace information from the .ibd files...
InnoDB: Restoring possible half-written data pages from the doublewrite
InnoDB: buffer...
131009 17:58:06  InnoDB: Waiting for the background threads to start
131009 17:58:07 InnoDB: 5.5.32 started; log sequence number 242183143
131009 17:58:07 [Note] Server hostname (bind-address): '127.0.0.1'; port: 3306
131009 17:58:07 [Note]   - '127.0.0.1' resolves to '127.0.0.1';
131009 17:58:07 [Note] Server socket created on IP: '127.0.0.1'.
131009 17:58:07 [Note] Event Scheduler: Loaded 0 events
131009 17:58:07 [Note] /usr/sbin/mysqld: ready for connections.
Version: '5.5.32-0ubuntu0.12.10.1'  socket: '/var/run/mysqld/mysqld.sock'  port: 3306  (Ubuntu)
131009 17:58:33 [Note] Plugin 'FEDERATED' is disabled.
131009 17:58:33 InnoDB: The InnoDB memory heap is disabled
131009 17:58:33 InnoDB: Mutexes and rw_locks use GCC atomic builtins
131009 17:58:33 InnoDB: Compressed tables use zlib 1.2.7
131009 17:58:33 InnoDB: Using Linux native AIO
131009 17:58:33 InnoDB: Initializing buffer pool, size = 128.0M
131009 17:58:34 InnoDB: Completed initialization of buffer pool
131009 17:58:34 InnoDB: highest supported file format is Barracuda.
InnoDB: Log scan progressed past the checkpoint lsn 242183143
131009 17:58:34  InnoDB: Database was not shut down normally!
InnoDB: Starting crash recovery.
InnoDB: Reading tablespace information from the .ibd files...
InnoDB: Restoring possible half-written data pages from the doublewrite
InnoDB: buffer...
InnoDB: Doing recovery: scanned up to log sequence number 242183153
131009 17:58:34  InnoDB: Waiting for the background threads to start
131009 17:58:35 InnoDB: 5.5.32 started; log sequence number 242183153
131009 17:58:35 [Note] Server hostname (bind-address): '127.0.0.1'; port: 3306
131009 17:58:35 [Note]   - '127.0.0.1' resolves to '127.0.0.1';
131009 17:58:35 [Note] Server socket created on IP: '127.0.0.1'.
131009 17:58:35 [Note] Event Scheduler: Loaded 0 events
131009 17:58:35 [Note] /usr/sbin/mysqld: ready for connections.
Version: '5.5.32-0ubuntu0.12.10.1'  socket: '/var/run/mysqld/mysqld.sock'  port: 3306  (Ubuntu)
131009 17:58:44 [Note] Plugin 'FEDERATED' is disabled.
131009 17:58:44 InnoDB: The InnoDB memory heap is disabled
131009 17:58:44 InnoDB: Mutexes and rw_locks use GCC atomic builtins
131009 17:58:44 InnoDB: Compressed tables use zlib 1.2.7
131009 17:58:44 InnoDB: Using Linux native AIO
131009 17:58:45 InnoDB: Initializing buffer pool, size = 128.0M
131009 17:58:45 InnoDB: Completed initialization of buffer pool
131009 17:58:45 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!
131009 17:58:45  InnoDB: Database was not shut down normally!
InnoDB: Starting crash recovery.
InnoDB: Reading tablespace information from the .ibd files...
InnoDB: Restoring possible half-written data pages from the doublewrite
InnoDB: buffer...
131009 17:58:45  InnoDB: Waiting for the background threads to start
131009 17:58:47 [Note] Plugin 'FEDERATED' is disabled.
131009 17:58:47 InnoDB: The InnoDB memory heap is disabled
131009 17:58:47 InnoDB: Mutexes and rw_locks use GCC atomic builtins
131009 17:58:47 InnoDB: Compressed tables use zlib 1.2.7
131009 17:58:47 InnoDB: Using Linux native AIO
131009 17:58:47 InnoDB: Initializing buffer pool, size = 128.0M
131009 17:58:47 InnoDB: Completed initialization of buffer pool
131009 17:58:47 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!
131009 17:58:47  InnoDB: Database was not shut down normally!
InnoDB: Starting crash recovery.
InnoDB: Reading tablespace information from the .ibd files...
InnoDB: Restoring possible half-written data pages from the doublewrite
InnoDB: buffer...
131009 17:58:47  InnoDB: Waiting for the background threads to start
131009 17:58:48 InnoDB: 5.5.32 started; log sequence number 242183153
131009 17:58:48 [Note] Server hostname (bind-address): '127.0.0.1'; port: 3306
131009 17:58:48 [Note]   - '127.0.0.1' resolves to '127.0.0.1';
131009 17:58:48 [Note] Server socket created on IP: '127.0.0.1'.
131009 17:58:48 [Note] Event Scheduler: Loaded 0 events
131009 17:58:48 [Note] /usr/sbin/mysqld: ready for connections.
Version: '5.5.32-0ubuntu0.12.10.1'  socket: '/var/run/mysqld/mysqld.sock'  port: 3306  (Ubuntu)

Answer 1

我发现当MySQL耗尽太多内存时就会发生这种情况.这个问题和这篇文章都有助于测试和解决方案.

对我来说,将此添加到我的/etc/mysql/my.cnf文件修复了该问题.

innodb_buffer_pool_size = 12M

这是我决定分配多少内存的方式.启动除MySQL之外的所有内容并检查可用内存量.将InnoDB缓冲区设置为该数量的10%.对我来说这是12M.

Answer 2

有几件事可以帮助您的安装运行而不会发生任何崩溃。我过去在使用 WordPress 时也遇到过类似的问题，很可能是你没有正确使用 InnoDB。从上面的日志的外观以及帖子中的详细信息来看，您添加了一个重要的交换，这可能没有做任何事情，或者使问题变得更糟......

首先，您的 4GB 交换内存对于这么小的服务器来说太高了。大多数进程可能是通过交换内存使用的。将交换内存减少到 512mb 或 1gb 并从那里排除故障。第二件事是检查交换配置。如果它太高，那么您的系统将积极使用交换内存 - 这会浪费周期，并减慢数据恢复速度，如果它太低，那么您可能根本没有它。默认值为 60，我建议从 10 开始，然后逐渐增加，直到找到适合您的服务器的“最佳位置”。请参阅这篇 Ubuntu 文章，了解如何在系统上更改它： https://help.ubuntu.com/community/SwapFaq#What_is_swappiness_and_how_do_I_change_it.3F

您的 InnoDB 缓冲区非常小，为 128.0MB。MySQL 可能永远不会触及 4GB 交换内存（因此 Apache/PHP 可能正在使用您的交换内存）。这可能就是它崩溃的原因。将 InnoDB 增加至少一倍，实际上我建议尝试在该缓冲区中保留完整的数据库，但我知道由于 RAM 限制，这实际上是不可能的。尝试将 RAM 大小设为 1/2，看看它与以前相比运行情况如何，然后根据结果降低或增加它。MySQL Workbench 有一个“服务器状态”工具，可以告诉您 InnoDB 缓冲区使用情况，理想情况下您希望它以略低于 100% 的速度运行（理想情况下为 80-90%），并有足够的冗余空间，以防出现任何峰值。

快速说明：如果您将整个数据库放入 RAM，您将需要进行更强大和更频繁的数据库备份（它是不稳定的，如果所有内容都在其中并且系统断电，那么您就完蛋了）。

如果这些仍然没有产生任何大的影响，那么请考虑为您的 WP 站点提供更强大的缓存。根据每个请求加载每个页面将是相当费力的，并且不需要太多努力就可以解决。查看 MySQL 上的服务器统计页面，并查看您的站点每秒运行的查询数，与每秒 InnoDB 读/写数进行比较。