VG deactivation for temporary disc maintenance?

[David Nixon]

Just wanted to check out the following maintenance scenario: DataPac 7000 RAID
system with failing cache battery pack.
A single VG owns all PVs from the RAID. HP-UX host connection is via an A6829A, dual channel HBA.
If I succeed in running vgchange -a N ; to
de-activate this VG, can I then shutdown my RAID, for battery replacement, without
disturbance (SCSI timeouts etc) to the running host.

[Matthew_50]

remember to umount filesystem first:-)

[Rick Garland]

Cannot deactivate VGs with filesystems mounted.

Do the umount 1st

[A. Clay Stephenson]

Remember, SCSI timeouts are treated as normal events and should be handled gracefully; otherwise, your system couldn't handle cable/controller/interface failures. I'm not familiar with your array but I assume it has dual controllers with batteries on each controller. All of the arrays I use, allow for battery replacement with the removal of a single controller. The activity shifts over to the other SCSI path. You can then use pvchange to shift back. You also have the option of using pvchange to switch paths before the battery replacement.

Having to shutdown for routine tasks is something I avoid but it may be the safer option; however, any array that requires shutdown for this task can hardly be considered a highly-available device.

[David Nixon]

You can certainly have dual redundant
(Infortrend) controllers in DPs 7000
series enclosures - we don't have
this set-up, yet.
No applications are currently using this
RAID; but services would be affected by host shutdown.
So my question about LVM is
does it, in any way, generate I/O to
the PVs of a de-activated VG..

[Devender Khatana]

Hi David,

Once VG is deactivated there will not be any IO to this disks. But if you try to access disks by other means like via dd , diskinfo or vgscan etc. then surely system will try to access disks.

HTH,
Devender

[David Nixon]

Devender, you have confirmed what I
hoped would be the LVM behaviour.