Grant Lowe
2010-Mar-10 20:28 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
Please help me out here. I''ve got a V240 with the root drive, c2t0d0 mirrored to c2t1d0. The mirror is having problems, and I''m unsure of the exact procedure to pull the mirrored drive. I see in various googling: zpool replace rpool c2t1d0 c2t1d0 or I''ve seen simply: zpool replace rpool c2t1d0 or I''ve seen the much more complex: zpool offline rpooll c2t1d0 cfgadm -c unconfigure c1::dsk/c2t1d0 (replace the drive) cfgadm -c configure c1::dsk/c2t1d0 zpool replace rpool c2t1d0s0 zpool online rpool c2t1d0s0 So which is it? Also, do I need to include the slice as in the last example? Thanks.
Cindy Swearingen
2010-Mar-10 21:09 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
Hi Grant, I don''t have a v240 to test but I think you might need to unconfigure the disk first on this system. So I would follow the more complex steps. If this is a root pool, then yes, you would need to use the slice identifier, and make sure it has an SMI disk label. After the zpool replace operation and the disk resilvering is complete, apply the boot blocks. The steps would look like this: # zpool offline rpool c2t1d0 #cfgadm -c unconfigure c1::dsk/c2t1d0 (physically replace the drive) (confirm an SMI label and a s0 exists) # cfgadm -c configure c1::dsk/c2t1d0 # zpool replace rpool c2t1d0s0 # zpool online rpool c2t1d0s0 # zpool status rpool /* to confirm the replacement/resilver is complete # installboot -F zfs /usr/platform/`uname -i`/lib/fs/zfs/bootblk /dev/rdsk/c2t1d0s0 Thanks, Cindy On 03/10/10 13:28, Grant Lowe wrote:> Please help me out here. I''ve got a V240 with the root drive, c2t0d0 mirrored to c2t1d0. The mirror is having problems, and I''m unsure of the exact procedure to pull the mirrored drive. I see in various googling: > > zpool replace rpool c2t1d0 c2t1d0 > > or I''ve seen simply: > > zpool replace rpool c2t1d0 > > or I''ve seen the much more complex: > > zpool offline rpooll c2t1d0 > cfgadm -c unconfigure c1::dsk/c2t1d0 > (replace the drive) > cfgadm -c configure c1::dsk/c2t1d0 > zpool replace rpool c2t1d0s0 > zpool online rpool c2t1d0s0 > > So which is it? Also, do I need to include the slice as in the last example? > > Thanks. > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Grant Lowe
2010-Mar-10 23:19 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
Well, this system is Solaris 05/09, with patches form November. No snapshots running and no internal contollers. It''s a file serving and attached to a HDS disk array. Help and please respond ASAP as this is production! Even an IM would be helpful. --- On Wed, 3/10/10, Cindy Swearingen <Cindy.Swearingen at Sun.COM> wrote:> From: Cindy Swearingen <Cindy.Swearingen at Sun.COM> > Subject: Re: [zfs-discuss] Replacing a failed/failed mirrored root disk > To: "Grant Lowe" <glowe at sbcglobal.net> > Cc: zfs-discuss at opensolaris.org > Date: Wednesday, March 10, 2010, 1:09 PM > Hi Grant, > > I don''t have a v240 to test but I think you might need to > unconfigure > the disk first on this system. > > So I would follow the more complex steps. > > If this is a root pool, then yes, you would need to use the > slice > identifier, and make sure it has an SMI disk label. > > After the zpool replace operation and the disk resilvering > is > complete, apply the boot blocks. > > The steps would look like this: > > # zpool offline rpool c2t1d0 > #cfgadm -c unconfigure c1::dsk/c2t1d0 > (physically replace the drive) > (confirm an SMI label and a s0 exists) > # cfgadm -c configure c1::dsk/c2t1d0 > # zpool replace rpool c2t1d0s0 > # zpool online rpool c2t1d0s0 > # zpool status rpool /* to confirm the replacement/resilver > is complete > # installboot -F zfs /usr/platform/`uname > -i`/lib/fs/zfs/bootblk > /dev/rdsk/c2t1d0s0 > > Thanks, > > Cindy > > > On 03/10/10 13:28, Grant Lowe wrote: > > Please help me out here. I''ve got a V240 with the root > drive, c2t0d0 mirrored to c2t1d0. The mirror is having > problems, and I''m unsure of the exact procedure to pull the > mirrored drive. I see in various googling: > > > > zpool replace rpool c2t1d0 c2t1d0 > > > > or I''ve seen simply: > > > > zpool replace rpool c2t1d0 > > > > or I''ve seen the much more complex: > > > > zpool offline rpooll c2t1d0 > > cfgadm -c unconfigure c1::dsk/c2t1d0 > > (replace the drive) > > cfgadm -c configure c1::dsk/c2t1d0 > > zpool replace rpool c2t1d0s0 > > zpool online rpool c2t1d0s0 > > > > So which is it? Also, do I need to include the slice > as in the last example? > > > > Thanks. > > > > _______________________________________________ > > zfs-discuss mailing list > > zfs-discuss at opensolaris.org > > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss >
Cindy Swearingen
2010-Mar-10 23:27 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
Hey list, Grant says his system is hanging after the zpool replace on a v240, running Solaris 10 5/09, 4 GB of memory, and no ongoing snapshots. No errors from zpool replace so it sounds like the disk was physically replaced successfully. If anyone else can comment or help Grant diagnose this issue, please feel free... Thanks, Cindy On 03/10/10 16:19, Grant Lowe wrote:> Well, this system is Solaris 05/09, with patches form November. No snapshots running and no internal contollers. It''s a file serving and attached to a HDS disk array. Help and please respond ASAP as this is production! Even an IM would be helpful. > > --- On Wed, 3/10/10, Cindy Swearingen <Cindy.Swearingen at Sun.COM> wrote: > >> From: Cindy Swearingen <Cindy.Swearingen at Sun.COM> >> Subject: Re: [zfs-discuss] Replacing a failed/failed mirrored root disk >> To: "Grant Lowe" <glowe at sbcglobal.net> >> Cc: zfs-discuss at opensolaris.org >> Date: Wednesday, March 10, 2010, 1:09 PM >> Hi Grant, >> >> I don''t have a v240 to test but I think you might need to >> unconfigure >> the disk first on this system. >> >> So I would follow the more complex steps. >> >> If this is a root pool, then yes, you would need to use the >> slice >> identifier, and make sure it has an SMI disk label. >> >> After the zpool replace operation and the disk resilvering >> is >> complete, apply the boot blocks. >> >> The steps would look like this: >> >> # zpool offline rpool c2t1d0 >> #cfgadm -c unconfigure c1::dsk/c2t1d0 >> (physically replace the drive) >> (confirm an SMI label and a s0 exists) >> # cfgadm -c configure c1::dsk/c2t1d0 >> # zpool replace rpool c2t1d0s0 >> # zpool online rpool c2t1d0s0 >> # zpool status rpool /* to confirm the replacement/resilver >> is complete >> # installboot -F zfs /usr/platform/`uname >> -i`/lib/fs/zfs/bootblk >> /dev/rdsk/c2t1d0s0 >> >> Thanks, >> >> Cindy >> >> >> On 03/10/10 13:28, Grant Lowe wrote: >>> Please help me out here. I''ve got a V240 with the root >> drive, c2t0d0 mirrored to c2t1d0. The mirror is having >> problems, and I''m unsure of the exact procedure to pull the >> mirrored drive. I see in various googling: >>> zpool replace rpool c2t1d0 c2t1d0 >>> >>> or I''ve seen simply: >>> >>> zpool replace rpool c2t1d0 >>> >>> or I''ve seen the much more complex: >>> >>> zpool offline rpooll c2t1d0 >>> cfgadm -c unconfigure c1::dsk/c2t1d0 >>> (replace the drive) >>> cfgadm -c configure c1::dsk/c2t1d0 >>> zpool replace rpool c2t1d0s0 >>> zpool online rpool c2t1d0s0 >>> >>> So which is it? Also, do I need to include the slice >> as in the last example? >>> Thanks. >>> >>> _______________________________________________ >>> zfs-discuss mailing list >>> zfs-discuss at opensolaris.org >>> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
David L Kensiski
2010-Mar-11 22:45 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
At Wed, 10 Mar 2010 15:28:40 -0800 Cindy Swearingen wrote: > Hey list, > > Grant says his system is hanging after the zpool replace on a v240, running Solaris 10 5/09, 4 GB of memory, and no ongoing snapshots. > No errors from zpool replace so it sounds like the disk was physically > replaced successfully. > > If anyone else can comment or help Grant diagnose this issue, please > feel free... > > Thanks, > > Cindy I''m had a similar problem -- swapped out the drive and now when I tried the zfs replace, I got an I/O error: k01_dlk$ sudo zpool replace rpool c1t0d0s0 cannot open ''/dev/dsk/c1t0d0s0'': I/O error I ran format and made sure the partition table matched the good root mirror, then was able to rerun the zpool replace: k01_dlk$ sudo zpool replace -f rpool c1t0d0s0 c1t0d0s0 Please be sure to invoke installgrub(1M) to make ''c1t0d0s0'' bootable. Can I assume zfs_stage1_5 is correct? k01_dlk$ cd /boot/grub/ k01_dlk$ ls bin fat_stage1_5 jfs_stage1_5 nbgrub stage1 ufs_stage1_5 capability ffs_stage1_5 menu.lst pxegrub stage2 vstafs_stage1_5 default install_menu menu.lst.orig reiserfs_stage1_5 stage2_eltorito xfs_stage1_5 e2fs_stage1_5 iso9660_stage1_5 minix_stage1_5 splash.xpm.gz ufs2_stage1_5 zfs_stage1_5 Installgrub succeeded, but I''m a tad nervous about rebooting until I know for sure. k01_dlk$ sudo installgrub zfs_stage1_5 stage2 /dev/rdsk/c1t0d0s0 stage1 written to partition 0 sector 0 (abs 16065) stage2 written to partition 0, 272 sectors starting at 50 (abs 16115) Thanks, --Dave -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100311/388e24b4/attachment.html>
Cindy Swearingen
2010-Mar-11 23:08 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
Hi David, In general, an I/O error means that the slice 0 doesn''t exist or some other problem exists with the disk. The installgrub command is like this: # installgrub /boot/grub/stage1 /boot/grub/stage2 /dev/rdsk/c1t0d0s0 Thanks, Cindy On 03/11/10 15:45, David L Kensiski wrote:> At Wed, 10 Mar 2010 15:28:40 -0800 Cindy Swearingen wrote: > >> Hey list, > >> > >> Grant says his system is hanging after the zpool replace on a v240, > running Solaris 10 5/09, 4 GB of memory, and no ongoing snapshots. > >> No errors from zpool replace so it sounds like the disk was physically >> replaced successfully. > >> > >> If anyone else can comment or help Grant diagnose this issue, please > >> feel free... > >> > >> Thanks, > >> > >> Cindy > > > > I''m had a similar problem -- swapped out the drive and now when I tried the zfs replace, I got an I/O error: > > > k01_dlk$ sudo zpool replace rpool c1t0d0s0 > cannot open ''/dev/dsk/c1t0d0s0'': I/O error > > > I ran format and made sure the partition table matched the good root mirror, then was able to rerun the zpool replace: > > > k01_dlk$ sudo zpool replace -f rpool c1t0d0s0 c1t0d0s0 > Please be sure to invoke installgrub(1M) to make ''c1t0d0s0'' bootable. > > > Can I assume zfs_stage1_5 is correct? > > > k01_dlk$ cd /boot/grub/ > > k01_dlk$ ls > bin fat_stage1_5 jfs_stage1_5 nbgrub stage1 ufs_stage1_5 > capability ffs_stage1_5 menu.lst pxegrub stage2 vstafs_stage1_5 > default install_menu menu.lst.orig reiserfs_stage1_5 stage2_eltorito xfs_stage1_5 > e2fs_stage1_5 iso9660_stage1_5 minix_stage1_5 splash.xpm.gz ufs2_stage1_5 zfs_stage1_5 > > > Installgrub succeeded, but I''m a tad nervous about rebooting until I know for sure. > > > k01_dlk$ sudo installgrub zfs_stage1_5 stage2 /dev/rdsk/c1t0d0s0 > stage1 written to partition 0 sector 0 (abs 16065) > stage2 written to partition 0, 272 sectors starting at 50 (abs 16115) > > > Thanks, > > --Dave > > > > > ------------------------------------------------------------------------ > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
David L Kensiski
2010-Mar-12 15:41 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
On Mar 11, 2010, at 3:08 PM, Cindy Swearingen wrote:> Hi David, > > In general, an I/O error means that the slice 0 doesn''t exist > or some other problem exists with the disk.Which makes complete sense because the partition table on the replacement didn''t have anything specified for slice 0.> > The installgrub command is like this: > > # installgrub /boot/grub/stage1 /boot/grub/stage2 /dev/rdsk/c1t0d0s0I reran it like that. For curiosity sake, what is the zfs_stage1_5 file file?> > Thanks, > > Cindy >Thanks you! --Dave -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20100312/f4578667/attachment.html>
Cindy Swearingen
2010-Mar-12 16:27 UTC
[zfs-discuss] Replacing a failed/failed mirrored root disk
Hi David, Make sure you test booting from the newly attached disk and update the BIOS, if necessary. The boot process occurs over several stages. I''m not sure what the zfs_stage1_5 file provides but is the first two that we need for booting. For more details, go to this site: http://hub.opensolaris.org/bin/view/Community+Group+zfs/boot And review Lori''s slides at the bottom of this page. Thanks, cindy On 03/12/10 08:41, David L Kensiski wrote:> On Mar 11, 2010, at 3:08 PM, Cindy Swearingen wrote: > >> Hi David, >> >> In general, an I/O error means that the slice 0 doesn''t exist >> or some other problem exists with the disk. > > > Which makes complete sense because the partition table on the > replacement didn''t have anything specified for slice 0. > > >> >> The installgrub command is like this: >> >> # installgrub /boot/grub/stage1 /boot/grub/stage2 /dev/rdsk/c1t0d0s0 > > > I reran it like that. For curiosity sake, what is the zfs_stage1_5 file > file? > > >> >> Thanks, >> >> Cindy >> > > > Thanks you! > > --Dave > >