[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#605009: serious performance regression with ext4



On Mon, Nov 29, 2010 at 3:58 PM, Ian Jackson
<ijackson@chiark.greenend.org.uk> wrote:
> Olaf van der Spek writes ("Re: Bug#605009: serious performance regression with ext4"):
>> Probably not an issue for dpkg, but in general:
>> Don't you reset meta-data that way?
>
> Yes.  If you want to keep the metadata you must copy it.

I guess that's hard.

>> Require a second file (name), permission to write to it and assume
>> it's on the same volume?
>
> It will be on the same volume because it's in the same directory.

What about the other issues?

> This is the standard way that ordinary files for which reliability was
> important have been updated on Unix for decades.  fsync is for files
> which need synchronisation with things external to the computer (or at
> least, external to the volume) - eg, email at final dot.

I know. I agree the rename trick should work, but my point is that a
better way for atomic non-durable file updates would be good to have.

Olaf


Reply to: