[Nottingham] "atime" or "noatime", that is a question...

[Martin]

Chris Burton wrote:
>> Any comments on FS atime?
> 
> Waste of time IMO.
> 
> There was a thread about similar things a while ago
> (http://kerneltrap.org/node/14148), havnt seen anything more recent
> though I havnt been looking.

Thanks for that one, quite a read!

I rather like this solution:

http://kerneltrap.org/node/14148#comment-261650

"So, the atime is updated only when the ctime or mtime is changed, or
[after] 24 hours out date. It means that when the file is opened read
only for less than 24 hrs, there will be no atime update."

A further good suggestion is that being as the atime is then
deliberately inaccurate, to have the timestamp set to be to the nearest
hour or 24 hours so as to be obviously inaccurate.


So... Anyone know if this modified "relatime" idea is included in
present kernels now?

(OK, so I'm still on an old kernel and too lazy to boot up a new one!)

Cheers,
Martin

-- 
----------------
Martin Lomas
martin at ml1.co.uk
----------------