On 09/24/2013 10:25 PM, Chris Murphy wrote:
On Sep 24, 2013, at 7:34 PM, William Brown <will...@firstyear.id.au> wrote:
Additionally, with the concerns re device shrink. Yes, XFS won't let you
shrink, but with thin provision LVM that isn't so much an issue: You
just shrink the pv and leave it alone. I would also argue that anyone
who is smart enough to shrink their devices, is smart enough to lookup a
little bit about lvm thinp.
I think LVM thinp commands are a lot more esoteric than ext4, btrfs, or xfs
resize. But really I think the issue is making it easy for users at install
time, so as long as the installer doesn't hit a brick wall being unable to
shrink with the default file system, it probably doesn't matter if it's ext4,
or XFS on thinp, or btrfs. But right now the installer only supports ext4
resize.
I think a really good feature that would go alongside this, would be
discards by default so that lvm thinp can relinquish free blocks also.
But that's really another topic :)
So as it turns out some devices get really busy and slow down when TRIM is
used, so not all devices are well suited for discard and therefore it's
probably not a good idea for it to be set by default. It's also not enabled by
default for ssds with btrfs either. And with dm-crypt there's a security
concern in that it exposes massive zero'd holes on the device instead of
current data being obscured by stale data.
We should not confuse TRIM that gets handled at the device layer (and is a slow,
non-queued S-ATA command for example) and a dm-thin parsing of that same command
in software which just updates the metadata that dm-thin maintains.
Using dm-thin is not for free, but I don't know that we have looked carefully at
the performance profile.
Ric
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
