NFS客户端阻塞睡眠问题与配置调研

Linux NFS客户端需要很小心地配置，否则在NFS服务器崩溃时，访问NFS的程序会被挂起，用ps查看，进程状态（STAT）处于D，意为（由于IO阻塞而进入）不可中断睡眠（如果是D+，+号表示程序运行于前台进程组）[1]。

为此，最新的Hadoop HA方案Quorum-Journal的作者在设计文档中对NFS共享存储保存NameNode元数据的作法如下诟病[2]：

Custom hardware - the hardware requirements of a NAS device and remotely controllable PDU can be expensive, and also may be different than the standard deployments used elsewhere within some "filer-free" organizations.
Complicated deployment - even after HDFS is installed, the administrator must take extra steps to congure NFS mounts, custom fencing scripts, etc. This complicates HA deployment and may even cause unavailability if misconfigured.
Poor NFS client implementations - in many versions of Linux, NFS client implementations can be buggy, or easy to misconfigure. For example, it is easy for an administrator to misconfigure mount options in such a way that the NameNodes will freeze unrecoverably in some outage scenarios.

此处第3点就是这次需要调研的内容：NFS客户端正确配置以防程序被长时间阻塞。

1. Hard mount vs Soft mount[3][4]

挂载NFS目录时可以使用hard或soft参数指定client如何处理server的异常（server崩溃或网络连接异常）；它们的区别是：

注：soft挂载可能导致隐性的数据损坏，因而应该在客户端响应性比数据完整性更重要的前提下使用该参数；通过增大retrans参数增加重试次数可以削减soft参数带来的风险；

2. timeo and retrans[3]

soft超时失败由timeo和retrans参数决定：

通过测试（在client挂载时指定参数，然后将server进程关闭，使用ls访问nfs目标），发现：

超时时间测试：

参考：

秒客网