Hi all, I need to know if it is possible to expand the capacity of a zpool without loss of data by growing the LUN (2TB) presented from an HP EVA to a Solaris 10 host. I know that there is a possible way in Solaris Express Community Edition, b117 with the autoexpand property. But I still work with Solaris 10 U7. Besides, when will this feature be integrated in Solaris 10? Is there a workaround? I have checked it out with format tool - without effects. Thanks for any info. Jan -- This message posted from opensolaris.org
On Wed, Jul 29, 2009 at 03:51:22AM -0700, Jan wrote:> Hi all, > I need to know if it is possible to expand the capacity of a zpool > without loss of data by growing the LUN (2TB) presented from an HP EVA > to a Solaris 10 host.Yes.> I know that there is a possible way in Solaris Express Community > Edition, b117 with the autoexpand property. But I still work with > Solaris 10 U7. Besides, when will this feature be integrated in > Solaris 10?Not sure.> Is there a workaround? I have checked it out with format tool - > without effects.What did you try? Since you''re larger than 1T, you certainly have an EFI label. What you have to do is destroy the existing EFI label, then have format create a new one for the larger LUN. Finally, create slice 0 as the size of the entire (now larger) disk. There are four ZFS labels inside the EFI data slice. Two at front, two at end. After enlarging, it probably won''t be able to find the end two, but it should import just fine (and will then write new labels at the end). As always, if you haven''t done this before, you''ll want to test it and make a backup before trying on live data. -- Darren
Hi Darren, thanks for your reply.> What did you try? > Since you''re larger than 1T, you certainly have an EFI label. What you > have to do is destroy the existing EFI label, then have format create a > new one for the larger LUN. Finally, create slice 0 as the size of the > entire (now larger) disk.Yes, I have an EFI label on that device. This is my procedure to try growing the capacity of the device: -> export the zpool -> overwrite the existing EFI label with format tool -> auto-configure it -> import the zpool What do you mean with "then have format create a new one for the larger LUN. Finally, create slice 0 as the size of the entire (now larger) disk."? Could you please give me some more detailed information on your description? Many thanks, jan -- This message posted from opensolaris.org
On Mon, Aug 03, 2009 at 01:15:49PM -0700, Jan wrote:> Yes, I have an EFI label on that device. > This is my procedure to try growing the capacity of the device: > -> export the zpool > -> overwrite the existing EFI label with format tool > -> auto-configure it > -> import the zpool > > What do you mean with "then have format create a > new one for the larger LUN. Finally, create slice 0 as the size of the > entire (now larger) disk."? > > Could you please give me some more detailed information on your description?I typed this up in a thread some months ago. One of the posts has some more detail. http://groups.google.com/group/comp.unix.solaris/browse_thread/thread/b3c996c76c3de860 Does that help? -- Darren
Hi Darren,
i tried exactly the same, but it doesn''t seem to work.
First the size of the disk:
[b] echo |format -e|grep -i 000005A0000[/b] 17.
c6t6001438002A5435A00010000005A0000d0 <HP-HSV210-6200-60.00GB>
/scsi_vhci/ssd at g6001438002a5435a00010000005a0000
Then creating a zpool:
[b]zpool create -m /zones/huhctmp huhctmppool
c6t6001438002A5435A00010000005A0000d0[/b]
[b]zpool list[/b]
NAME SIZE USED AVAIL CAP HEALTH ALTROOT
huhctmppool 59.5G 103K 59.5G 0% ONLINE -
Exporting the zpool:
[b]zpool export huhctmppool[/b]
Then a resize on the EVA, which gives 70GB instead of 60GB:
[b]echo |format -e|grep -i 000005A0000[/b] 17.
c6t6001438002A5435A00010000005A0000d0 <HP-HSV210-6200-70.00GB>
/scsi_vhci/ssd at g6001438002a5435a00010000005a0000
Now labeling and do the autoconfiguration:
format> type
AVAILABLE DRIVE TYPES:
0. Auto configure
1. other
Specify disk type (enter its number)[1]: 0
c6t6001438002A5435A00010000005A0000d0: configured with capacity of 70.00GB
<HP-HSV210-6200-70.00GB>
selecting c6t6001438002A5435A00010000005A0000d0
[disk formatted]
And in the end trying to import:
zpool import huhctmppool
cannot import ''huhctmppool'': no such pool available
Anything else I can try ?
BTW: ZFS is Version 10, Solaris 5/08
Thanks,
Sascha
--
This message posted from opensolaris.org
On Tue, Aug 11, 2009 at 09:35:53AM -0700, Sascha wrote:> Then creating a zpool: > [b]zpool create -m /zones/huhctmp huhctmppool c6t6001438002A5435A00010000005A0000d0[/b] > > [b]zpool list[/b] > NAME SIZE USED AVAIL CAP HEALTH ALTROOT > huhctmppool 59.5G 103K 59.5G 0% ONLINE -I recall your original question was about a larger disk (>1TB). My assumption is that the zpool create has created an EFI label here, but it would be nice to confirm with ''format'' or ''prtvtoc'' output. [skipping ahead]> format> type > > AVAILABLE DRIVE TYPES: > 0. Auto configure > 1. other > Specify disk type (enter its number)[1]: 0 > c6t6001438002A5435A00010000005A0000d0: configured with capacity of 70.00GB > <HP-HSV210-6200-70.00GB> > selecting c6t6001438002A5435A00010000005A0000d0 > [disk formatted]My fear is that with a small disk and not running ''format -e'' it''s placed an SMI label on the disk. Again, confirming that would be nice. The label in this section must match the label in the first section. If they''re both EFI, you should be good. If it put an SMI label on, the data won''t line up. If that''s all that''s wrong, you can repeat this step on your lun and should still be good. (use ''format -e'' and when labeling it should prompt you for the label type).> And in the end trying to import: > zpool import huhctmppoolHmm, you''ve skipped the partitioning step (which should be done after applying the new label). I would make sure that slice 0 encompasses the full 70G. Basically, make the partitioning here look like it used to, but with more data in the slice. So #1. Confirm you have EFI labels before and after running format #2. Partition the disk after the label to look like it used to (but with a larger slice) -- Darren
Hi Darren, thanks for your quick answer.> On Tue, Aug 11, 2009 at 09:35:53AM -0700, Sascha > wrote: > > Then creating a zpool: > > zpool create -m /zones/huhctmp huhctmppool > c6t6001438002A5435A00010000005A0000d0 > > > > zpool list > > NAME SIZE USED AVAIL CAP HEALTH > ALTROOT > huhctmppool 59.5G 103K 59.5G 0% ONLINE - > I recall your original question was about a larger > disk (>1TB).It wasn''t me, I just have the same problem... ;-)> My assumption is that the zpool create has created an > EFI label here, but > it would be nice to confirm with ''format'' or > ''prtvtoc'' output.confirmed, it''s really an EFI Label. (see below) format> label [0] SMI Label [1] EFI Label Specify Label type[1]: 0 Warning: This disk has an EFI label. Changing to SMI label will erase all current partitions. BTW: Is there a smarter way to find out what Label is in place ?> [skipping ahead] > > > format> type > > > My fear is that with a small disk and not running > ''format -e'' it''s > placed an SMI label on the disk. Again, confirming > that would be nice. > The label in this section must match the label in the > first section. If > they''re both EFI, you should be good. If it put an > SMI label on, the > data won''t line up. If that''s all that''s wrong, you > can repeat this > step on your lun and should still be good. > > (use ''format -e'' and when labeling it should prompt > you for the label > type). > > > And in the end trying to import: > > zpool import huhctmppool > > Hmm, you''ve skipped the partitioning step (which > should be done after > applying the new label).Yes I skipped it, because I had no choice to choose the whole size. There was only partition 0 and 8. Even if select 0, delete the slice and newly errect it, it has the same size. Also the number of disk sectors did not change and slice 8 has still it''s starting cylinder at the "old" end of the disk (See below) format> label [0] SMI Label [1] EFI Label Specify Label type[1]: Ready to label disk, continue? yes partition> p Current partition table (original): Total disk sectors available: 146784222 + 16384 (reserved sectors) Part Tag Flag First Sector Size Last Sector 0 usr wm 256 69.99GB 146784222 1 unassigned wm 0 0 0 2 unassigned wm 0 0 0 3 unassigned wm 0 0 0 4 unassigned wm 0 0 0 5 unassigned wm 0 0 0 6 unassigned wm 0 0 0 7 unassigned wm 0 0 0 8 reserved wm 146784223 8.00MB 146800606> I would make sure that > slice 0 encompasses the > full 70G. Basically, make the partitioning here look > like it used to, > but with more data in the slice.I would like to, but how ?> So > #1. Confirm you have EFI labels before and after > running formatOk, I can comfirm that.> #2. Partition the disk after the label to look like > it used to (but > with a larger slice)Well, that dosn''t seem to work :-( Sascha> Darren > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discu > ss-- This message posted from opensolaris.org
Darren, I want to give you a short overview of what I tried: 1. created a zpool on a LUN resized the LUN on the EVA exported the zpool used format -e and label tried to enlarge slice 0 -> impossible (see posting above) 2. Same like 1. but exported the zpool before resizing on the EVA Same result... 3. Created a zpool on a LUN exported the LUN resized the LUN on the EVA relabled the LUN with EFI Label made an auto-reconfigure The partition tool showed the correct size and number of cylinders Unfortunatly, I could not import the zpool (no zpool definded) 4. Tried several variations of 3., but ended always up with no zpool during import. Sascha -- This message posted from opensolaris.org
On Wed, Aug 12, 2009 at 04:53:20AM -0700, Sascha wrote:> confirmed, it''s really an EFI Label. (see below) > > format> label > [0] SMI Label > [1] EFI Label > Specify Label type[1]: 0 > Warning: This disk has an EFI label. Changing to SMI label will erase all > current partitions. > > BTW: Is there a smarter way to find out what Label is in place ?Take a look at partitions. If you have 0-7 or 0-15, you have an SMI label. If you have 0-6 and 8 (7 is skipped), then you have an EFI label.> > Hmm, you''ve skipped the partitioning step (which > > should be done after > > applying the new label). > > Yes I skipped it, because I had no choice to choose the whole size. > There was only partition 0 and 8. > Even if select 0, delete the slice and newly errect it, it has the same size. > Also the number of disk sectors did not change and slice 8 has still it''s starting cylinder at the "old" end of the disk (See below) > > format> label > [0] SMI Label > [1] EFI Label > Specify Label type[1]: > Ready to label disk, continue? yes > > partition> p > Current partition table (original): > Total disk sectors available: 146784222 + 16384 (reserved sectors) > > Part Tag Flag First Sector Size Last Sector > 0 usr wm 256 69.99GB 146784222Okay, you had done that, you just hadn''t confirmed that in the first email. Assuming the first sector is the same before and after, this should all be correct. Can you do zdb -l <diskname> after the resize? Do you get any data? -- Darren
Hi Darren,
took me a while what device is meant by zdb -l...
Original size was 20GB
After resizing in EVA format -e showed the new correct size:
17. c6t6001438002A5435A00010000006D0000d0 <HP-HSV210-6200-30.00GB>
/scsi_vhci/ssd at g6001438002a5435a00010000006d0000
Here is the output of zdb -l:
zdb -l /dev/dsk/c6t6001438002A5435A00010000006D0000d0s0
--------------------------------------------
LABEL 0
--------------------------------------------
version=10
name=''huhctmppool''
state=0
txg=14
pool_guid=8571912358322557497
hostid=2230083084
hostname=''huh014''
top_guid=3763080418932148585
guid=3763080418932148585
vdev_tree
type=''disk''
id=0
guid=3763080418932148585
path=''/dev/dsk/c6t6001438002A5435A00010000006D0000d0s0''
devid=''id1,ssd at n6001438002a5435a00010000006d0000/a''
phys_path=''/scsi_vhci/ssd at
g6001438002a5435a00010000006d0000:a''
whole_disk=1
metaslab_array=14
metaslab_shift=27
ashift=9
asize=21461467136
is_log=0
--------------------------------------------
LABEL 1
--------------------------------------------
version=10
name=''huhctmppool''
state=0
txg=14
pool_guid=8571912358322557497
hostid=2230083084
hostname=''huh014''
top_guid=3763080418932148585
guid=3763080418932148585
vdev_tree
type=''disk''
id=0
guid=3763080418932148585
path=''/dev/dsk/c6t6001438002A5435A00010000006D0000d0s0''
devid=''id1,ssd at n6001438002a5435a00010000006d0000/a''
phys_path=''/scsi_vhci/ssd at
g6001438002a5435a00010000006d0000:a''
whole_disk=1
metaslab_array=14
metaslab_shift=27
ashift=9
asize=21461467136
is_log=0
--------------------------------------------
LABEL 2
--------------------------------------------
version=10
name=''huhctmppool''
state=0
txg=14
pool_guid=8571912358322557497
hostid=2230083084
hostname=''huh014''
top_guid=3763080418932148585
guid=3763080418932148585
vdev_tree
type=''disk''
id=0
guid=3763080418932148585
path=''/dev/dsk/c6t6001438002A5435A00010000006D0000d0s0''
devid=''id1,ssd at n6001438002a5435a00010000006d0000/a''
phys_path=''/scsi_vhci/ssd at
g6001438002a5435a00010000006d0000:a''
whole_disk=1
metaslab_array=14
metaslab_shift=27
ashift=9
asize=21461467136
is_log=0
--------------------------------------------
LABEL 3
--------------------------------------------
version=10
name=''huhctmppool''
state=0
txg=14
pool_guid=8571912358322557497
hostid=2230083084
hostname=''huh014''
top_guid=3763080418932148585
guid=3763080418932148585
vdev_tree
type=''disk''
id=0
guid=3763080418932148585
path=''/dev/dsk/c6t6001438002A5435A00010000006D0000d0s0''
devid=''id1,ssd at n6001438002a5435a00010000006d0000/a''
phys_path=''/scsi_vhci/ssd at
g6001438002a5435a00010000006d0000:a''
whole_disk=1
metaslab_array=14
metaslab_shift=27
ashift=9
asize=21461467136
is_log=0
Is that the output you meant ?
Sascha
--
This message posted from opensolaris.org
Hi Darren, sorry that it took so long before I could answer. The good thing: I found out what went wrong. What I did: After resizing a Disk on the Storage, solaris recognizes it immediately. Everytime you resize a disk, the EVA storage updates the discription which contains the size. So typing "echo |format" shows promptly the new discription. So that seemed to work fine. But It turned out that "format" doesn''t automagically update the amount of cylinders/sectors. So i tried to "auto configure" the disk. When I did that, the first sector of partition 0 changed from sector # 34 to # 256. (which makes label 0 and 1 of zpool unaccessible). The last sector changed to the new end of the disk (which makes label 3 and 4 of the zpool unaccessible). Partition 8 changed accordingly and correctly. If I then relabel the disk the zpool is destroyed. Label 0 and 1 of the zpool are gone and, obviously label 2 and 3 too. What I did to solve it: 1. Changed the size of a disk/LUN on the storage 2. Verified the position of the starting sector of partition 0 3. Did an auto configure 4. Changed starting sector of partition 0 to the formerly starting sector from step #2 5. labeled the disk I then was able to impoort the zpool Do you know why the "auto configure" changes the starting cylinder ? What I also found out is that it did not happen every time. Sometimes the first sector of partition 0 changes, sometimes not. Up to now I can''t find any correlation between when and why. Sascha -- This message posted from opensolaris.org
On Sep 21, 2009, at 8:59 AM, Sascha wrote:> Hi Darren, > > sorry that it took so long before I could answer. > > The good thing: > I found out what went wrong. > > What I did: > After resizing a Disk on the Storage, solaris recognizes it > immediately. > Everytime you resize a disk, the EVA storage updates the discription > which contains the size. So typing "echo |format" shows promptly the > new discription. > So that seemed to work fine. > > But It turned out that "format" doesn''t automagically update the > amount of cylinders/sectors. > So i tried to "auto configure" the disk. > When I did that, the first sector of partition 0 changed from sector > # 34 to # 256. (which makes label 0 and 1 of zpool unaccessible).This was a change made long ago, but it finally caught up with you. You must have created the original EFI label with an older version of [Open]Solaris. If you then relabel with automatic settings, the starting sector will change to the 256 value, which is a much better starting point. -- richard> The last sector changed to the new end of the disk (which makes > label 3 and 4 of the zpool unaccessible). > Partition 8 changed accordingly and correctly. > If I then relabel the disk the zpool is destroyed. > Label 0 and 1 of the zpool are gone and, obviously label 2 and 3 too. > > What I did to solve it: > 1. Changed the size of a disk/LUN on the storage > 2. Verified the position of the starting sector of partition 0 > 3. Did an auto configure > 4. Changed starting sector of partition 0 to the formerly starting > sector from step #2 > 5. labeled the disk > > I then was able to impoort the zpool > > Do you know why the "auto configure" changes the starting cylinder ? > What I also found out is that it did not happen every time. > Sometimes the first sector of partition 0 changes, sometimes not. > Up to now I can''t find any correlation between when and why. > > Sascha > -- > This message posted from opensolaris.org > _______________________________________________ > zfs-discuss mailing list > zfs-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Hej Richard. think I''ll update all our servers to the same version of zfs... That will hopefully make sure that this doesn''t happen again :-) Darren and Richard: Thank you very much for your help ! Sascha -- This message posted from opensolaris.org