I''ve read on some sun doc that zfs was not supported in live upgrades. I''m using zfs on my desktop and sort of a test server at home and after reading that, looks like I can''t do LU''s. I was also thinking about mv''n another server that holds a TB of stuff to zfs, but again I wont if I can''t LU. I understand there is a diff between "not supported" and "not possible" so I would like to "assume" that it would be possible but if I loose my stuff thats my fault. Which is sort of fine since this 1TB box is a backup for another 1TB server at home. Or, if I want to do LU''s do I just need to re-install my desktop and get away from using zfs? Thanks, This message posted from opensolaris.org
On Tue, Apr 25, 2006 at 01:26:13PM -0700, Justin Conover wrote:> I''ve read on some sun doc that zfs was not supported in live upgrades. > > I''m using zfs on my desktop and sort of a test server at home and after reading that, looks like I can''t do LU''s. I was also thinking about mv''n another server that holds a TB of stuff to zfs, but again I wont if I can''t LU. > > I understand there is a diff between "not supported" and "not possible" so I would like to "assume" that it would be possible but if I loose my stuff thats my fault. Which is sort of fine since this 1TB box is a backup for another 1TB server at home.Storing system software (i.e. package contents) on a ZFS filesystem is definitely "not possible" with live upgrade. Just having ZFS filesystems laying around is possible, although there are some rough edges. For example, live upgrade will create mountpoints for ZFS filesystem even though they''re auto-created for you, which makes ZFS confused when it goes to mount parent datasets and finds a non-empty directory. But nothing that can''t be worked around. Basically, you should be fine as long as you don''t make /usr, /lib, or some other system-provided directory into a ZFS filesystem. You might need to do a little housecleaning afterwards, but you won''t lose data or anything like that. - Eric -- Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock
Folks, NB: this situation won''t last for long - we just need to get the ''whole'' space done. Bev. Eric Schrock wrote:> On Tue, Apr 25, 2006 at 01:26:13PM -0700, Justin Conover wrote: > >> I''ve read on some sun doc that zfs was not supported in live upgrades. >> >> I''m using zfs on my desktop and sort of a test server at home and after reading that, looks like I can''t do LU''s. I was also thinking about mv''n another server that holds a TB of stuff to zfs, but again I wont if I can''t LU. >> >> I understand there is a diff between "not supported" and "not possible" so I would like to "assume" that it would be possible but if I loose my stuff thats my fault. Which is sort of fine since this 1TB box is a backup for another 1TB server at home. >> > > Storing system software (i.e. package contents) on a ZFS filesystem is > definitely "not possible" with live upgrade. Just having ZFS > filesystems laying around is possible, although there are some rough > edges. For example, live upgrade will create mountpoints for ZFS > filesystem even though they''re auto-created for you, which makes ZFS > confused when it goes to mount parent datasets and finds a non-empty > directory. But nothing that can''t be worked around. > > Basically, you should be fine as long as you don''t make /usr, /lib, or > some other system-provided directory into a ZFS filesystem. You might > need to do a little housecleaning afterwards, but you won''t lose data or > anything like that. > > - Eric > > -- > Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20060425/01f3bf01/attachment.html>
See bug 6376420 for a workaround for this problem. Lori Eric Schrock wrote:> On Tue, Apr 25, 2006 at 01:26:13PM -0700, Justin Conover wrote: > >>I''ve read on some sun doc that zfs was not supported in live upgrades. >> >>I''m using zfs on my desktop and sort of a test server at home and after reading that, looks like I can''t do LU''s. I was also thinking about mv''n another server that holds a TB of stuff to zfs, but again I wont if I can''t LU. >> >>I understand there is a diff between "not supported" and "not possible" so I would like to "assume" that it would be possible but if I loose my stuff thats my fault. Which is sort of fine since this 1TB box is a backup for another 1TB server at home. > > > Storing system software (i.e. package contents) on a ZFS filesystem is > definitely "not possible" with live upgrade. Just having ZFS > filesystems laying around is possible, although there are some rough > edges. For example, live upgrade will create mountpoints for ZFS > filesystem even though they''re auto-created for you, which makes ZFS > confused when it goes to mount parent datasets and finds a non-empty > directory. But nothing that can''t be worked around. > > Basically, you should be fine as long as you don''t make /usr, /lib, or > some other system-provided directory into a ZFS filesystem. You might > need to do a little housecleaning afterwards, but you won''t lose data or > anything like that. > > - Eric > > -- > Eric Schrock, Solaris Kernel Development http://blogs.sun.com/eschrock > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
same question Hello ,,,, can liveupgrade be done on the current zfs i am running snv_35 on sparc if so how tks art This message posted from opensolaris.org
Lori Alt <Lori.Alt at Sun.COM> writes:> See bug 6376420 for a workaround for this problem.... which, unfortunately, is not (yet?) visible on bugs.opensolaris.org. Rainer
On Wed, 2006-04-26 at 11:09 +0200, Rainer Orth wrote:> Lori Alt <Lori.Alt at Sun.COM> writes: > > > See bug 6376420 for a workaround for this problem. > > ... which, unfortunately, is not (yet?) visible on bugs.opensolaris.org.Erk, sorry about that. Here''s the workaround from the bug report: "The lucreate/luupgrade will succeed if the zfs file systems are excluded from the lucreate. This can be done with the -x switch to lucreate. *** (#1 of 1): 2006-01-24 17:57:22 GMT lori.alt at sun.com" cheers, tim -- Tim Foster, Sun Microsystems Inc, Operating Platforms Group Engineering Operations http://blogs.sun.com/timf
So, I would basically do this lucreate -m /:/dev/dsk/c0d0s0:ufs -n snv_38 -c snv_36 -x /dev/dsk/c2d0 Because zfs is on the whole drive, or use /dev/dsk/c2d0s0, if i do format on c2d0 it has this: /dev/dsk/c2d0s0 is part of active ZFS pool zfs. So just not sure if adding the s0 matters or not. Thanks, This message posted from opensolaris.org
Justin Conover wrote:> So, I would basically do this > > lucreate -m /:/dev/dsk/c0d0s0:ufs -n snv_38 -c snv_36 -x /dev/dsk/c2d0Actually, as I read the lucreate man page, the argument to -x needs to be a file or directory. So if the zfs file system(s) is mounted at /zfspool, you would add -x /zfspool to the lucreate command. lori> > Because zfs is on the whole drive, or use /dev/dsk/c2d0s0, if i do format on c2d0 it has this: > > /dev/dsk/c2d0s0 is part of active ZFS pool zfs. > > So just not sure if adding the s0 matters or not. > > Thanks, > > > This message posted from opensolaris.org > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Lori Alt wrote:> Justin Conover wrote: > >> So, I would basically do this >> >> lucreate -m /:/dev/dsk/c0d0s0:ufs -n snv_38 -c snv_36 -x /dev/dsk/c2d0 > > > Actually, as I read the lucreate man page, the argument > to -x needs to be a file or directory. So if the zfs file > system(s) is mounted at /zfspool, you would add -x /zfspool > to the lucreate command.It might be better to get all the zfs fs and then exclude them from lucreate. # mount -p | fgrep " zfs " | cut -f3 -d'' '' > /tmp/excludeList # lucreate -m /:/dev/dsk/c0d0s0:ufs -n snv_38 -c snv_36 -f /tmp/excludeList thanks, Raja> > lori > >> >> Because zfs is on the whole drive, or use /dev/dsk/c2d0s0, if i do >> format on c2d0 it has this: >> >> /dev/dsk/c2d0s0 is part of active ZFS pool zfs. >> So just not sure if adding the s0 matters or not. >> >> Thanks, >> >> >> This message posted from opensolaris.org >> _______________________________________________ >> zfs-discuss mailing list >> zfs-discuss at opensolaris.org >> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss > > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss-- Raja Gopal Andra Sun Microsystems Phone: +1 650 786 4273 OP/N1 RPE 17 Network Circle, Fax : +1 650 786 7994 Approachability Group Menlo Park, CA 94025 mailto: andra at sun.com
On Wed, 2006-04-26 at 17:42, Raja Gopal Andra wrote:> It might be better to get all the zfs fs and then exclude them > from lucreate. > > # mount -p | fgrep " zfs " | cut -f3 -d'' '' > /tmp/excludeList > # lucreate -m /:/dev/dsk/c0d0s0:ufs -n snv_38 -c snv_36 -f /tmp/excludeListor you can drop zfs into /etc/default/lu''s LU_UNSUPPORTED_FILE_SYSTEM_TYPES list, which appears to the do the right thing for me.. - Bill
Bill Sommerfeld wrote:> On Wed, 2006-04-26 at 17:42, Raja Gopal Andra wrote: > >>It might be better to get all the zfs fs and then exclude them >>from lucreate. >> >># mount -p | fgrep " zfs " | cut -f3 -d'' '' > /tmp/excludeList >># lucreate -m /:/dev/dsk/c0d0s0:ufs -n snv_38 -c snv_36 -f /tmp/excludeList > > > or you can drop zfs into /etc/default/lu''s > LU_UNSUPPORTED_FILE_SYSTEM_TYPES list, which appears to the do the right > thing for me.. >That does work (and in fact, by default, it''s already there), as long you don''t have nested zfs file systems. But if you do, you run into 6376420 (lucreate creates spurious mount point directories when a boot-environment has nested file systems). The spurious mount point directories created by lucreate don''t give ufs file systems any grief because ufs will allow you to mount over a non-empty directory, but zfs won''t. Lori