Redis自发地失败了：“失败打开.rdb保存：权限被拒绝”

我在Rackspace Cloud Server的Ubuntu 14.04上运行redis-server（版本2.8.4）。 redis服务器进程正在由redis用户运行。

今天早上，Redis服务器自发地失败了“打开.rdb进行保存”。 这个过程发生了20多天没有发生。

正如你可以从下面的日志中看到的那样，Redis正常运行，然后在UTC时间上午8:08开始失败。 稍后closures服务器电源解决了问题。

[26083] 31 Aug 07:54:47.089 * RDB: 2 MB of memory used by copy-on-write [17183] 31 Aug 07:54:47.186 * Background saving terminated with success [17183] 31 Aug 07:59:48.028 * 10 changes in 300 seconds. Saving... [17183] 31 Aug 07:59:48.029 * Background saving started by pid 26653 [26653] 31 Aug 07:59:48.032 * DB saved on disk [26653] 31 Aug 07:59:48.032 * RDB: 2 MB of memory used by copy-on-write [17183] 31 Aug 07:59:48.129 * Background saving terminated with success [17183] 31 Aug 08:04:49.061 * 10 changes in 300 seconds. Saving... [17183] 31 Aug 08:04:49.061 * Background saving started by pid 27197 [27197] 31 Aug 08:04:49.064 * DB saved on disk [27197] 31 Aug 08:04:49.065 * RDB: 2 MB of memory used by copy-on-write [17183] 31 Aug 08:04:49.162 * Background saving terminated with success [17183] 31 Aug 08:07:56.881 * DB saved on disk [17183] 31 Aug 08:08:05.880 # Failed opening .rdb for saving: Permission denied [17183] 31 Aug 08:12:57.047 * 10 changes in 300 seconds. Saving... [17183] 31 Aug 08:12:57.047 * Background saving started by pid 28087 [28087] 31 Aug 08:12:57.048 # Failed opening .rdb for saving: Permission denied [17183] 31 Aug 08:12:57.148 # Background saving error [17183] 31 Aug 08:13:03.068 * 10 changes in 300 seconds. Saving... [17183] 31 Aug 08:13:03.069 * Background saving started by pid 28088 [28088] 31 Aug 08:13:03.070 # Failed opening .rdb for saving: Permission denied [17183] 31 Aug 08:13:03.169 # Background saving error [17183] 31 Aug 08:13:09.087 * 10 changes in 300 seconds. Saving... [17183] 31 Aug 08:13:09.087 * Background saving started by pid 28089 [28089] 31 Aug 08:13:09.088 # Failed opening .rdb for saving: Permission denied (... continues like this until server restart) 

所以，我的问题：这怎么可能发生？

• 了解每个请求32个字节的networking性能影响
• 如果没有运行，则创build一个cron作业，每30分钟运行一次redis
• 在EC2 Micro上的Resque（Redis）复制到3个可用区域
• 如何检查redis master是否可以？
• Redis on EC2（Heroku）：RedisToGo v / s MyRedis v / s OpenRedis v / s RedisCloud v / s Others

我很难过，但这里是我模糊的预感：

• 内存使用率太高（我只运行一个512MB的实例，Redis警告我在/etc/sysctl.conf中设置'vm.overcommit_memory = 1'）
• Redis在连续运行20天后遇到了某种内部限制（所以这会在另外的20天内再次发生）。
• Rackspace在上午8点（美国东部时间凌晨4点）做了一些事情，这会影响我的服务器。