Request for focused testing on containers (LVM, btrfs) in F19 Beta TCs

[Chris Murphy]

On May 3, 2013, at 5:08 PM, Adam Williamson <awilliam at redhat.com> wrote:
> 
> Well, it's a UI *change* - this is now in the properties of the
> container. If you pick any subvol, you should see a button that lets you
> edit the properties of the volume, and you can set its name, target size
> and RAID level from there. This is part of the major container UI change
> we're trying to test, in fact.

OK. I've filed a bunch of bugs on how this UI deals with sizing. The sizing behaviors are very screwy. I set the white board to newui-container for these.


> 
>> 2. (md) RAID levels 4, 5, 6 refer to "error checking" which is wrong.
>> There is no error checking at all in normal operation of any of these
>> RAID levels. It's a grossly misleading description. I've mentioned it
>> probably 1/2 dozen times on anaconda@ through the F18 lifecycle and
>> it's been ignored. I don't know if I need to use stronger language or
>> what, but the current descriptions effectively lie to users about how
>> these RAID levels function. In a world where words, terms, actually
>> have meaning rather than being meaningless, this needs to be fixed.
> 
> Is there a specific bug report for it?

There is now:
https://bugzilla.redhat.com/show_bug.cgi?id=959701



> 
>>> trying to set up LVM-on-RAID (which now ought to be possible, I
>> believe)
>> 
>> If it is possible, I can't figure out how to do it.
> 
> Similar to the btrfs thing; you set it from the properties of the VG
> now.

Fair enough.

The present implementation first creates conventional md RAID devices, makes them PVs, adds them to the VG. So the raid level is an attribute of the VG. However, LVM directly implements RAID (using md driver code) which makes the raid level an attribute of the LV. I certainly find it preferable because it's less confusing eliminating the visible md RAID devices. It's easier to maintain when adding or replacing devices. And it probably needs to be tested more.

Right now, for Btrfs, raid level is a volume attribute, but it's planned to make it a subvolume attribute. So eventually it'll behave like LVM does.


Chris Murphy