I/O error while reading the VGDA

[Andrey Strougatsky]

Hi,

I have cluster from two K-class machine.
It has common RAID (NIKE) and Lock disk on SCSI bus.

By one node all VG are activated without problems.

On the second node I can not activate any VG from RAID with next error message:
vgchange: Couldn't activate volume group "/dev/vgspace":
I/O error while reading the VGDA.

But all LUN on second node can be viewed ioscan. And normally work diskinfo.
And work without errors with any VG next command:
dd if=/dev/vgspace/... of=/dev/null bs=1024k count=100

But when I try execute
vgchange -a y /dev/vgspace
I received message:
vgchange: Couldn't activate volume group "/dev/vgspace":
I/O error while reading the VGDA.

In the same time in syslog.log I received message:
Apr 8 08:23:11 hp2 vmunix: LVM: vg[2]: pvnum=0 (dev_t=0x1f023700) is POWERFAILED

I did vgexport-vgimport from one to second node some times.
It working, but VG is not activated on second node ï

And such situation and with groups included in cluster and with separate.

And still in syslog.log there are messages of such maintenance (see attachment).

By the way all this situation has taken place after (but it is not necessary because of ) upgrading microware in Nike

(Upgrade was made by engineer from HP because after UPS crash the cache in RAID was dirty and we can not start any way. HP engineer say that he can repair it, but all special utility for it was made for another release of microware.
He upgrade microware to this release. Restore integrity of RAID. After this step we restart cluster and all work good.
But after he was upgrade microware to last version for Nike RAID and when it restarted it work only with one node well, but not working with second, as I describe above)
What can I do ?

[RAC_1]

Do you have problems with disk - dev_t=0x1f023700. Disk - /dev/dsk/c2t3d7.

Does it have alternate path? Is diskinfo ok for this disk?

How did you do vgexport and vgimport on another node? Did you use -s option?
Probaly the disk has gone bad, had problems.

Anil

[Fabio Ettore]

Hi,

try to restore VG information:
# vgcfgrestore -n /dev/vg00 -l
# vgcfgrestore -n /dev/vg00 /dev/rdsk/cXtXdX
on cXtXdX put the first disk of previous command (probably disk is /dev/dsk/c2t37d0 = 0x1f023700).

If after that you obtain error yet then I think that you should substitute that disk.

HTH.

Best regards,
Ettore

[RAC_1]

Check this link to understand disk powerfiled.

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=484727

Anil

[Fabio Ettore]

Sorry, disk dev_t=0x1f023700 is as RAC said: c2t3d7.

Best regards,
Ettore

[Andrey Strougatsky]

To all !

All disks working good, because I can activate the same VG on another node and mount all. All working.
Diskinfo and dd for all disk on node 1 or 2 working without error.
But I can not activate any VG on second node :(

[Ashwani Kashyap]

If you are trying to activate the VG on the second node , while it is already activated on the first node , then you can not do this as the Vg's are activated in exclusive mode .
Deactivate the VG's on the first node and then try to activate them on the second node .

[Hoefnix]

Hi,

When exporting the VG check if the device files for all LUNS in the VG are the same on all the nodes. When it's different because it's connected for example on another BUS-path you have to change the dev-file before running the vgimport.

HTH,
Peter