pvchange/lvchange Oracle question

[Chris Cruz_1]

I would like to know if when using the pvchange and lvchange commands, do we need to unmount the the filesystems on those logical volumes and physical volumes? If not, would we need to shut down our Oracle databases before issuing those commands?

[A. Clay Stephenson · Accepted solution]

Essentially the only restriction is if the vg is in shared mode. You can run the lvchange/pvchange commands on mounted filesystems and running Oracle instances. Typically the commands are used to increase the default i/o timeout values for array devices.

[Nguyen Anh Tien]

Do you use Cluster in your enviroment???
Cluster usually use vg in share mode. It is impossible to change vg or lv online in this mode.
If not you can change it online.
tienna

[Devender Khatana]

Hi,

Allthough it do not casue any problems except if it is in cluster or shared mode. But still it will cause huge I/O on the devices which actually depends on size of volume and so should be planned at a time when oracle database utilization is minimal.

HTH,
Devender

[Chris Cruz_1]

Thanks for all the useful feedback. It is very much appreciated. Yes, we do need to run pvchange and lvchange to increase the timeout values for our EMC array.

Is there an order I should do them in (pvchange & lvchange)?

[D Block 2]

Chris,

regarding the EMC and your HPUX host... might want to also take a look at the "queue depth".

see: queue depth in this link:
http://web.tampabay.rr.com/batcave/unixnotes.htm

the point is, pvchange -t option can also be closely associated to the SCSI controller's queue depth for pending I/O's.


--
Change the timeout value to 60:

# pvchange -t 60 /dev/dsk/c?t?d?

Check the queue depth, should be 2:
# scsictl -a /dev/rdsk/c?t?d?

Change it to the value of 2:

# scsictl -m queue_depth=2 /dev/rdsk/c?t?d?

[Sandman!]

Chris,

pvchange the timeout value on all PVs that belong to the oracle VGs. Don't need to run lvchange on the lvols unless you want to change another parameter like (strict allocation policy).

cheers!