ZFS和BTRFS是不是不适合做比如mongodb的文件系统

一般用 xfs 并且 mount noatime

@phuslu
http://xfs.org/index.php/XFS_FAQ
说没必要mount noatime。读的时候不记录atime是可以避免读的时候的写操作，但是写不一定是马上写到硬盘上的，应该对io的影响方面，关系不大吧。
不过如果只用作mongodb的文件系统，也确实没必要记录atime
引用：
The default atime behaviour is relatime, which has almost no overhead compared to noatime but still maintains sane atime values. All Linux filesystems use this as the default now (since around 2.6.30), but XFS has used relatime-like behaviour since 2006, so no-one should really need to ever use noatime on XFS for performance reasons.

@oldcai 这个之前在 Amazon EC2 的 EBS 测过，需要关掉。

@phuslu 能具体一点么，比如告诉我下为什么要关掉。谢谢。

楼主对写时复制的理解有误吧， 写时复制是用来做快照的

不一定是复制整个文件，ZFS那么先进的文件系统没那么脑残。

@lixm
@pubby 恩，可能是以block为单位的吧。但是这样不是会有很多碎片了。

@oldcai 碎片。。。。太深奥了，我觉得要不实际测试一下看看性能吧。

@pubby
性能影响是有的，但是程度还未知，所以还是试试 @phuslu 推荐的xfs先，看了下，支持超大文件，支持快照，支持碎片整理，还是不错的。
https://groups.google.com/forum/#!topic/mongodb-user/HWLECIxg9ug

That said, there may be ways to disable COW on ZFS since most
databases expect in-place updates. However I am not enough of an
expert in ZFS to know how to do this. You may already have it set up
that way for MySQL.

少复制了一段。。
We've heard reports of slowdowns due to it's copy-on-write behavior
leading to objects being out-of-order on disk and needing more
seeking. To some extent, MongoDB's performance (both read and write)
relies on in-place updating.