[linux-lvm] pvresize complains about too many metadata

[linux-lvm] pvresize complains about too many metadata

[Fabio Coatti]

Hi all, I'm playing a bit with lvm/raid on linux, (on a pretty
experimental box) and now I'm facing a problem that leaves me a bit
stuck

Basically, I've a volume group based on several mdraid partition
(vg1  with md0/md1/md4 as pvs)

recently I've increased the size of md1/md4 arrays by adding more
disks and /proc/mdstats shows that the operation was successful.

the  I've issued a "pvresize" command on /dev/md1 and md4 to make lvm
aware of new size, but I get the following error:

pvresize -v /dev/md1
    Using physical volume(s) on command line
    Archiving volume group "vg1" metadata (seqno 12).
  /dev/md1: too many metadata areas for pvresize
  0 physical volume(s) resized / 1 physical volume(s) not resized


pvck -v /dev/md1
    Scanning /dev/md1
  Found label on /dev/md1, sector 1, type=LVM2 001
  Found text metadata area: offset=4096, size=192512
    Found LVM2 metadata record at offset=10240, size=2048, offset2=0 size2=0
    Found LVM2 metadata record at offset=9728, size=512, offset2=0 size2=0
    Found LVM2 metadata record at offset=8192, size=1536, offset2=0 size2=0
  Found text metadata area: offset=386587820032, size=131072
    Found LVM2 metadata record at offset=386587826176, size=2048,
offset2=0 size2=0
    Found LVM2 metadata record at offset=386587824128, size=2048,
offset2=0 size2=0
    Found LVM2 metadata record at offset=386587822592, size=1536,
offset2=0 size2=0


Can someone give me some hint, to shed some light on what's going on?
obviously I've made something wrong, but I can't figure what (maybe
unusual or not the best way to increase space, ok, but I can't figure
out what's up).

Thanks for any answer.

Re: [linux-lvm] pvresize complains about too many metadata

[Stuart D. Gathman]

[-- Attachment #1: Type: text/plain, Size: 443 bytes --]

Fabio Coatti wrote:
> Hi all, I'm playing a bit with lvm/raid on linux, (on a pretty
> experimental box) and now I'm facing a problem that leaves me a bit
> stuck
>
> Basically, I've a volume group based on several mdraid partition
> (vg1  with md0/md1/md4 as pvs)
>
> recently I've increased the size of md1/md4 arrays by adding more
> disks and /proc/mdstats shows that the operation was successful.
>   
Why not just add another md device?

[-- Attachment #2: stuart.vcf --]
[-- Type: text/x-vcard, Size: 315 bytes --]

begin:vcard
fn:Stuart Gathman
n:Gathman;Stuart
org:Business Management Systems, Inc
adr;dom:;;4001 Williamsburg Court;Fairfax;VA;22032
email;internet:stuart@bmsi.com
tel;work:703 591-0911
tel;fax:703 591-6154
tel;home:703 378-9641
note:AIM: SDGathman
x-mozilla-html:FALSE
url:http://bmsi.com
version:2.1
end:vcard

Re: [linux-lvm] pvresize complains about too many metadata

[Fabio Coatti]

2008/11/30 Stuart D. Gathman <stuart@bmsi.com>:
> Fabio Coatti wrote:
>> Hi all, I'm playing a bit with lvm/raid on linux, (on a pretty
>> experimental box) and now I'm facing a problem that leaves me a bit
>> stuck
>>
>> Basically, I've a volume group based on several mdraid partition
>> (vg1  with md0/md1/md4 as pvs)
>>
>> recently I've increased the size of md1/md4 arrays by adding more
>> disks and /proc/mdstats shows that the operation was successful.
>>
> Why not just add another md device?

Sorry? I've added one disk to a raid5 array, I can't figure out how I
can add more md devices.
basically, the question is: given a correct and successful operation
on the underlying block device, why lvm is complaining about too much
metadata? is there any way to fix this behaviour?

Re: [linux-lvm] pvresize complains about too many metadata

[Alasdair G Kergon]

On Sat, Nov 29, 2008 at 04:13:31PM +0100, Fabio Coatti wrote:
>   /dev/md1: too many metadata areas for pvresize

That's meant to be saying you chose to place a second copy of the metadata at
the end of the device and if you want to increase the size of the data area you
have to shift that metadata first and there is no code that does that.

Workaround is to reduce to 1 metadata area first (backup VG, redo pvcreate with
uuid & restorefile, restore VG), then extend, then put the 2nd metadata area
back (same process) if you still want it.

Alasdair
-- 
agk@redhat.com

Re: [linux-lvm] pvresize complains about too many metadata

[Fabio Coatti]

2008/11/30 Alasdair G Kergon <agk@redhat.com>:
> On Sat, Nov 29, 2008 at 04:13:31PM +0100, Fabio Coatti wrote:
>>   /dev/md1: too many metadata areas for pvresize
>
> That's meant to be saying you chose to place a second copy of the metadata at
> the end of the device and if you want to increase the size of the data area you
> have to shift that metadata first and there is no code that does that.
>
> Workaround is to reduce to 1 metadata area first (backup VG, redo pvcreate with
> uuid & restorefile, restore VG), then extend, then put the 2nd metadata area
> back (same process) if you still want it.
>


Thanks for your explanation, really helpful. Now I can understand
what's going up.

Thanks.

Re: [linux-lvm] pvresize complains about too many metadata

[Fabio Coatti]

2008/11/30 Alasdair G Kergon <agk@redhat.com>:
> On Sat, Nov 29, 2008 at 04:13:31PM +0100, Fabio Coatti wrote:
>>   /dev/md1: too many metadata areas for pvresize
>
> That's meant to be saying you chose to place a second copy of the metadata at
> the end of the device and if you want to increase the size of the data area you
> have to shift that metadata first and there is no code that does that.
>
> Workaround is to reduce to 1 metadata area first (backup VG, redo pvcreate with
> uuid & restorefile, restore VG), then extend, then put the 2nd metadata area
> back (same process) if you still want it.

Done that (basically, having several PV and some space, some pvmove
and pvcreate solved it). But basically this means that the option to
use 2 metadata areas prevents pv to grow, right?
Maybe this situation should be reported on documentation: "never use
two copies of metadata if it can happen to resize pv" (or can it be
considered a bug and solved on code side? basically pvresize should
handle the move of the second metadata at the end of the pv after the
resize operation...)