Cross-device link error

[dictum9]

I am trying to active this volume group and am getting the following error:


# vgchange -a y /dev/vgora
vgchange: Warning: Couldn't attach to the volume group physical volume "/dev/dsk/c10t1d1":
Cross-device link
vgchange: Cross-device link
vgchange: Warning: couldn't query physical volume "/dev/dsk/c10t1d1":
The specified path does not correspond to physical volume attached to
this volume group
vgchange: Warning: couldn't query physical volume "/dev/dsk/c10t1d2":
The specified path does not correspond to physical volume attached to
this volume group
vgchange: Warning: couldn't query all of the physical volumes.
vgchange: Couldn't activate volume group "/dev/vgora":
Quorum not present, or some physical volume(s) are missing.

[Steven E. Protter · Accepted solution]

Shalom,

Your volume group lost a disk, depriving it of quorum, which is 50% of the disks in the volume group being present.

This issue needs to be addresed prior to activation.

SEP

[Jakes Louw]

You can activate the VG without a quorum. However, this will have little value unless your Logical Volumes do not span disk boundaries, ie: you lvols have been created to reside on only one of the disks. Then it would be possible to mount some of the file-systems, or have access to the logical volumes in any case.
SEP is right: the issue of the missing disk must be solved! What does IOSCAN say? Have you had a H/W failure? Are there other disks involved?

[dictum9]

I think LVM is just a symptom of the problem. Here is the entry from syslog.log, what really caused it.


May 8 15:01:03 sba1a vmunix: msgcnt 1 vxfs: mesg 037: vx_metaioerr - vx_write_default - /dev/vgora/lvol1 file system meta data write error in block 530776
May 8 15:01:03 sba1a vmunix: msgcnt 2 vxfs: mesg 031: vx_disable - /dev/vgora/lvol1 file
system disabled
May 8 15:01:04 sba1a vmunix: msgcnt 3 vxfs: mesg 017: vx_write_default - /home/oracle file system inode 24442 marked bad incore
May 8 15:01:04 sba1a vmunix: ?vxfs msgcnt 3 offset 0x00000000 81a4 1 c8 c 9

[Deoncia Grayson_1]

WARNING: msgcnt x: vxfs: mesg 031:
vx_disable - mount_point file system disabled

Message 031 Explanation:

File system disabled, preceded by a message that specifies the reason. This
usually indicates a serious disk problem.

.Action

Unmount the file system and use fsck to run a full structural check.
If the problem is a disk failure, replace the disk before the file system is
mounted for write access.

this document will help you with your error codes:
http://www1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&admit=-682735245+1147260157175+28353475&docId=200000076281156

[Jakes Louw]

What is really concerning, is that this is the 3rd Volume Group related question posted by etc in 3 days: methinks this person is struggling somewhat.
Dear etc: maybe you should give us some more detail on what's been going on with your servers.

[RAC_1]

You seem to have two problems at hand.
cross-device link - Looks like some of your disks are used in more than one volume group. Check with vgdisplay, strings /etc/lvmtab

quorum not present - some of the disks seems to have died/have problem. Try activating volume group without quorum.

vgchange -a y -q n /dev/vgxx

[Jakes Louw]

Here's something I dredged up from a memory: this machine isn't perhaps part of an MC/SG cluster, is it? I've had cases where disks have been erroneously shared between two nodes, and made a member of a VG on both sides. As soon as the VGCREATE is executed on the second node, it obviously overwrites the VG Disk Area on the disk, and the first node can no longer understand why the VG info on the disk is wrong...

[dictum9]

Yes, they are in Service Guard...

[Jakes Louw]

OK, what are you using as an array? EMC, EVA, or something else?
Have you made a list of LUNs on each machine and tried to match them to any of the disks that are causing trouble?

[Albert_31]

Hello,

Do you still need help on this issue, if yes do collect some info and forward it to me.

Logs required
# xd -j8200 -N16 /dev/rdsk/c*t*d*
for c10t1d1 and c10t1d2
# strings /etc/lvmtab
# uname -i (from this server and all servers which can access the luns)
# vgcfgrestore -l -f /etc/lvmconf/vgora.conf
# vgcfgrestore -l -f /etc/lvmconf/vgora.conf.old

The error means that someone has corrupted the LVM headers by use of pvcreate (Most likely) or vgcfgrestore command.

Please add pereira.albert@indiatimes.com in the CC list as well.

regards

Albert