Hi, I was wondering if it would be such a stupid idea to have ext3 filesystems mounted with the noatime option during the installation. The switch would still be omitted by default in fstab, i.e. from the first boot on it would still default to off.
I am not sure what kind of speedup you would see. I suppose it would help the most at the lower end, on systems where there isn't much RAM: the small disk cache on such machines already forces the system to seek left and right all the time.
This would apply only to fresh installs. I guess that, for updating existing installations, Anaconda already uses the same options that the user has set in fstab.
Are there any packages that would break if installed for the first time using noatime?
On Mon, Oct 11, 2004 at 12:14:47AM +0200, Rudi Chiarito wrote:
I am not sure what kind of speedup you would see. I suppose it would
In case anyone wants to try this, I have a patch that applies cleanly against the latest anaconda in FC3RC3. I only know that the patched RPM builds, other problems prevent me from testing this right now.
http://uberh4x0r.org/~yax/anaconda-noatime.patch http://uberh4x0r.org/~yax/anaconda.spec
The patch should add a noatime switch that can be used either at the boot screen or in a kickstart file. All it should do is mount with the noatime option the partitions to which packages are installed.
Use at your own risk, etc.
On Wed, Oct 27, 2004 at 11:22:10PM +0200, Rudi Chiarito wrote:
In case anyone wants to try this, I have a patch that applies cleanly against the latest anaconda in FC3RC3. I only know that the patched RPM builds, other problems prevent me from testing this right now.
Maybe it's just me, but I think that the doPwMount() function would have a nicer API if the flags it accepts were passed as a bitfield instead of an argument per flag.
Cheers,
Matt
anaconda-devel@lists.stg.fedoraproject.org