Does Network File System has a smart cache

Latest response

Hi all,

It has been told that the NFSv4.2 is already being implemented in the Red Hat Enterprise Linux 7. I have a question concerning the NFS cache. Does NFS have a smart cache which can smart prefetch and/or smart commit?

LP
Log in to join the conversation

Responses

Jamie Bainbridge's picture Red Hat Guru 7257 points
14 October 2017 11:58 PM

Correct:

NFS uses the regular filesystem cache.

The method to configure readahead on an NFS mount is described at:

I am not sure what you mean by smart commit, but NFS uses asynchronous I/O and the regular Linux dirty page flush mechanism. The page flush behaviour can be tuned with the regular sysctl -a | grep dirty tunables.

The nfs(5) man page describes the attribute cache mount options and the sync mount option which can alter caching/flushing behaviour on the NFS Client. The exports(5) man page describes the sync and async export options which can alter this behaviour on the NFS Server.

LP Newbie 10 points
16 October 2017 1:48 AM

By smart commit, I mean that if multiple clients write the same NFS share will the file be corrupted or there is a mechanism protect the share?

Jamie Bainbridge's picture Red Hat Guru 7257 points
16 October 2017 2:01 AM

NFS has file locking, so that only one NFS Client can write to a file or an area of a file at the same time.

As long as locking is working, the file should not get corrupted.

Locking is done with a separate RPC daemon in NFSv3, and in-protocol with StateID in NFSv4.

Close

Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.