Re: Bug#605009: serious performance regression with ext4
Hi,
On Tue, 30 Nov 2010, Michael Biebl wrote:
> Something interesting I noticed:
>
> I created the ext4 file system on a spare partition and installed a chroot.
> After running the test, I exited the chroot, immediately unmounted the partition
> and measured how long it took:
>
> 1.15.8.5: 0.4s
Expected, this version calls sync() so there's nothing to write back when
you umount the filesystem.
> 1.15.8.6: 7.2s
> 1.15.8.6+patch: 18.4s
I can't explain the difference between both but it doesn't matter much IMO.
> If I wait a bit (+60s) after exiting the chroot and then do the unmount, it is
> almost instant (< 0.2s) in all three cases.
60s is more than the default writeback delay so again it's normal.
$ sudo cat /proc/sys/vm/dirty_writeback_centisecs
500
-> writeback after 5 sec
Cheers,
--
Raphaël Hertzog ◈ Debian Developer
Follow my Debian News ▶ http://RaphaelHertzog.com (English)
▶ http://RaphaelHertzog.fr (Français)
Reply to: