On Thu, 2002-08-22 at 14:33, ERG@xxxxxxxxxxxxxxxxx wrote:
> Greetings,
>
> I'm testing out XFS on some LVM2 volumes. These test volumes/FSs were
> created with the simplest invocation of mkfs I could muster: 'mkfs -t xfs
> /dev/vg0/lv0'. Now I'm getting some warnings from device-mapper which
> (apparently) indicate that I need to use the new XFS log format and set up
> my filesystems so that log writes are aligned properly.
>
> I asked on linux-lvm, but it seems that list isn't stocked with XFS gurus
> ;), so I'm having a hard time figuring out to what I need to set my su
> suboption at mkfs time (as in 'mkfs -t xfs -l internal,version=2,su=X
> /dev/vg0/lv0'). I'm not quite understanding the interactions between
> device-mapper, the device (or, in this case, LV) block size, and the
> filesystem block size.
>
> One suggestion I did get on linux-lvm goes like this:
> > To me it looks like you should set su to your block size, so if your
> > block size is 4k set su=8. I don't use xfs myself, so you may want to
> > wait for confirmation before trying this.
>
> Does that sound correct? Does anyone have the time or inclination to
> explain this so I can understand what's going on, rather than just dropping
> a value into my command (though that'd be appreciated, too)? Maybe this is
> easy, and I'm just slow :)
This should be correct, it will make the log writes have the same
alignment as the rest of the metadata. You may also want to increase
your log size from the default, the v2 log code seems to have an
end case in it which will panic the default log size. Using the
most recent commands, something like this should do it:
mkfs -t xfs -f -l sunit=8,size=8192b /dev/vg0/lv0
version 2 logs will get turned on automatically when you specify
the stripe unit.
Steve
--
Steve Lord voice: +1-651-683-3511
Principal Engineer, Filesystem Software email: lord@xxxxxxx
|