Well said by John.
I'd add one minor disclaimer.
Backups and restores ARE different when dealing with user volumes.
Assuming that you include the ;DIRECTORY option on your backup, you would
need to also include:
;onvs=mpexl_system_volume_set,uservol1,uservol2,...
On any future system install, you would need to re-create the user volume
set environment with VOLUTIL and restore the properly-stored directories
from backup before initiating the restore.
I still think that this and other inconveniences would discourage me from
using user volumes in an environment already protected by hardware mirroring.
At 01:24 PM 2004-03-10, John Clogg wrote:
>One of the main benefits of private volumes is fault isolation. If a disk
>drive in a private volume set were to fail, you would only need to restore
>that volume set, rather than doing a complete install of the whole
>system. Since you have mirrored arrays, that benefit diminishes in
>importance. I don't find system administration to be enough more
>difficult with private volumes to even consider as a factor. About the
>only consideration is creating and deleting groups. For each group you
>create you have to do two NEWGROUP commands - one to add the group to the
>system directory, and one for the volume set directory. There are command
>files available on Jazz that automate that extra step for you, if you find
>that problematic. Backups and restores are no different with private
>volume sets.
>
>Yes, you can manipulate class names without a reload. Others have already
>provided instructions.
>
>Regarding maintenance of your Nikes, you have three choices: (1) HP will
>probably still support them on a time and materials basis, (2) You can get
>a maintenance agreement with a third-party support organization, or (3)
>Stock you own set of spares. In my opinion, the best answer is either
>number 2 or a combination of 1 and 3. In other words, have spare drives
>available to replace any that fail, and pay T&M charges if a more involved
>repair is needed.
>
>One more thing I want to comment on: Craig characterized XM as a write
>cache. I disagree. XM is essentially a log of physical transactions in
>progress, whose purpose is to allow the reversal of incomplete
>transactions in the event of a system failure. Disc caching happens
>regardless of whether you have XM enabled or not. If having multiple
>instances of XM does anything for performance, it's because of increased
>parallelism and less frequent checkpoints.
>
>* To join/leave the list, search archives, change list settings, *
>* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
-------------------------------------------------------------------------------------------------
Gilles Schipper
GSA Inc.
HP System Administration Specialists
300 John Street, Box 87651 Thornhill, ON Canada L3T 7R4
Voice: 905.889.3000 Fax: 905.889.3001
email: [log in to unmask] web: http://www.gsainc.com
-------------------------------------------------------------------------------------------------
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|