Joern Michael Krueger
2005-Dec-27 16:28 UTC
[zfs-discuss] ZFS and dynamic LUN reconfiguration
Hello, I have a question about ZFS and the abitlity to online resisize (especially: grow) LUNs. I am aware that with the traditonal ufs/svm I could online resize a LUN exported say by a NetApp-Filer, clear the VTOC and then recreate the VTOC. I then could grow my filesystem. After reading the ZFS Admin Handbook I did not find any hint if and how zfs can handle resized LUNs. Any pointers on what I am missing are greatly appreciated :-) Thanks, J?rn P.S. I did read ZFS On Disk Specification and the comments in vdefs_label.c-File. To my understanding the VTOC block of the ZFS labels is empty, because ZFS uses EFI disklables. Is there any possibility to manipulate EFI labels? This message posted from opensolaris.org
On Tue, Dec 27, 2005 at 08:28:04AM -0800, Joern Michael Krueger wrote:> > I have a question about ZFS and the abitlity to online resisize > (especially: grow) LUNs. I am aware that with the traditonal ufs/svm > I could online resize a LUN exported say by a NetApp-Filer, clear the > VTOC and then recreate the VTOC. I then could grow my filesystem. > > After reading the ZFS Admin Handbook I did not find any hint if and > how zfs can handle resized LUNs. Any pointers on what I am missing > are greatly appreciated :-)ZFS can automatically handle LUN growth. If you have two devices in a mirror, and you grow one and then the other, the mirror will automatically grow to encompass the size of both devices. I believe that this check only happens on reopen (Jeff or Bill can correct me), so you''ll have to do an explicit ''zpool online'' of each device. This will likely be improved in the future.> I did read ZFS On Disk Specification and the comments in > vdefs_label.c-File. To my understanding the VTOC block of the ZFS > labels is empty, because ZFS uses EFI disklables. Is there any > possibility to manipulate EFI labels?Yes, when using EFI labels the first 8k of the label will be empty. ZFS will work with either VTOC or EFI labels, it just chooses (for a variety of reasons) to use EFI labels when given whole disks on the command line. You can manipulate EFI lables using format(1M). - Eric -- Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock
What about if the pool contains only 1 LUN In my case, my mirrorring or other data protection mecanism is done in my storage array so I dont want to setup a mirror on the host side After the LUN is grown, I go in format -e to autodetect my new device size and then I write a new EFI label on it using the new space. Im using emc powerpath as multipathing software on the host ( solaris 10 x64 ) doing a zpool online <pool> <device>, I get a success message but the new space isnot seen neither in df nor in zpool iostat : # zpool iostat -v sasWork capacity operations bandwidth pool used avail read write read write ------------ ----- ----- ----- ----- ----- ----- sasWork 3.28G 36.5G 0 0 0 10.1K emcpower1a 3.28G 36.5G 0 0 0 10.1K ------------ ----- ----- ----- ----- ----- ----- # zpool offline sasWork emcpower1a cannot offline emcpower1a: no valid replicas # zpool online sasWork emcpower1a Bringing device emcpower1a online # zpool iostat -v sasWork capacity operations bandwidth pool used avail read write read write ------------ ----- ----- ----- ----- ----- ----- sasWork 3.28G 36.5G 0 0 0 10.1K emcpower1a 3.28G 36.5G 0 0 0 10.1K ------------ ----- ----- ----- ----- ----- ----- # With UFS and EFI label, I can grow a filesystem without unmounting it, so it remains available for my users i/o at all time, that is what Im trying to acheive with ZFS, as some performance testing shows that i/o writes are a lot faster on a ZFS filesystem This message posted from opensolaris.org
so does anyone know how to <rescan> the LUN(s) part of its pool and detect new size of the LUN ? This message posted from opensolaris.org
from what I know this operation goes via an zpool export, re-label (with format) , then zpool import it''s not online On 6/5/07, Yan <yanmercier at gmail.com> wrote:> so does anyone know how to <rescan> the LUN(s) part of its pool > and detect new size of the LUN ? > > > This message posted from opensolaris.org > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss >