On Aug 9, 2011, at 10:20 AM, Drew Fisher wrote: > > > On 8/9/11 7:55 AM, Darren Kenny wrote: >> On 09/08/2011 14:30, Drew Fisher wrote: >>> Good morning! >>> >>> I wanted to discuss 7032205 >>> (http://monaco.sfbay.sun.com/detail.jsf?cr=7032205) with the group to >>> try to get a consensus with the right approach to take. What I was >>> thinking was setting a value of 1% of the total size of the root pool, >>> with an upper bound of 1gb. >>> >>> For the ultra small disks (32gb SSD), 1% is 320MB. For huge disks (2tb >>> HDD), we'd cap at 1gb. This would provide admins a little bit of >>> breathing room for a full root pool. The problem with doing something >>> like this is some admins would really appreciate the installers doing >>> something like this for them while other admins want to customize each >>> and every single aspect of their install. This leads to a few approaches: >>> >>> 1: Set up an ICT to do the quota setting *after* we successfully >>> install - this is so we can complete the install without zfs getting in >>> the way. >>> >>> 2: Set up controller.py to add an additional zpool option for the quota >>> on the root pool. >>> >>> 3: other? Add something to the default manifest for AI and force the >>> option for GUI and TI? DTD entries specifically for this? >>> >>> Please let me know what you folks think! >> Honestly, and I've said this before, I feel that this is a ZFS bug and >> should be >> fixed in ZFS - even with a 100% full disk it should be possible to make >> space - >> but it isn't, there really should be a 'buffer' in there. > > I totally agree with the sentiment that this is a ZFS issue. Perhaps we > don't need to do anything on our end, aside from escalate the CR already > assigned to ZFS.
The current situation is unclear to me... obviously, no one seems to be on the initial RFE (6803038), but 6976827, which was integrated in snv_167) at least by synopsis, seems to be a fix. Maybe we should check into this first. Back to the point here - I think it's a good workaround to set a quota on the rpool, but if it's not necessary I doubt it's required even optionally. /jb > >> >> But, if we are going to have to go down this route, then I think what you're >> suggesting sounds like a good start and I would prefer option 2, as long as >> there isn't a quota already specified by the user. This gives the best of >> both >> worlds - i.e. automatically settable while leaving it possible to override. > > Thanks, Darren. > > -Drew > _______________________________________________ > caiman-discuss mailing list > [email protected] > http://mail.opensolaris.org/mailman/listinfo/caiman-discuss _______________________________________________ caiman-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
