NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/comp/unix/aix/13682

Internet Message Format | 1993-01-24 | 3.9 KB

Path: sparky!uunet!paladin.american.edu!howland.reston.ans.net!zaphod.mps.ohio-state.edu!menudo.uh.edu!navier.math.uh.edu!mcbryde
From: [e-mail] (Jack Mcbryde)
Newsgroups: comp.unix.aix
Subject: Re: Another Newbie Asks A Question ...
Date: 24 Jan 1993 07:13:43 GMT
Organization: UH Dept of Math
Lines: 57
Distribution: na
Message-ID: <[e-mail]>
References: <[e-mail]>
NNTP-Posting-Host: navier.math.uh.edu

In article <[e-mail]> Ron Wigmore <[e-mail]> writes:
>
>We just had three RS6K boxes installed on our computer floor.  I have been
>reading over a few of the manuals, checking things, etc., and I came upon
>the part which says that "the default installation option is to spread
>rootvg over all available disks".  I checked and, yes, that is the way
>it was installed.
>
>Most of my experience is on mainframes and we would never put the system
>components on each and every disk, since losing any one disk means your
>system crashes.  We tend to have 'system' disks, 'application' disks, and
>'user' disks.  This way, if you lose a disk, the chances are the majority
>of the users can continue 'business as usual'.
[reasoning deleted]

Basically you've got it pegged. The only reason to have rootvg encompass all
six disks (other than some kind of performance scheme for a mongo database
or somesuch) is to maximize the use of your disk space. If you have people
in charge who are willing to spend money on a reasonable amount of disk you
are much better off having more volume groups. (for the sort of reasons you
suggest) Assuming all of the disks are of reasonable size, I'd put rootvg
on hdisk0. Move \home out of rootvg since user accounts are there and put
static software on rootvg. If you need another disk then add it. Do a mksysb
on rootvg when it's installed and whenever you update software. No need to do
it on a weekly or daily basis. Then partition the other disks into volume
groups as you see fit. My inclination would be to avoid having more than two
disks in a volume group unless it were necessary. If they are large disks
I'd probably try to have one disk per volume group.

If you have, say, 6 1Gb disks then it's hard to imagine that the software
would take up more than two disks. (probably fit on one) Then you only
have a 1/6 or 2/6 chance of being completely out of commission when a disk
bites the dust. Plus you don't have to make regular backups of stuff you
already have backed up. You only need to back up rootvg when you modify it.
(This is why I suggest moving /home out of rootvg.) If you want to do
mirroring the setup might be different. I'm not gonna comment on whether
or not I think you should set up mirroring. Depends on what kind of media
you have and how long it would take to get a replacement disk and how
crucial it is that you are functional in the mean time. (ie, how long is your
total downtime, etc) The remaining volume groups you can back up using
whatever scheme happens to suit your needs as often as suits your needs.

Of course this is all subject to your situation. If you have 500 users accessing
a 5.5Gb database in a situation where four hours of downtime costs the company
a million dollars your situation is different than if you have 3 users
developing software that uses large data sets for test cases. (In the first
situation I would set up mirroring. In the second I'd let the 3 guys take the
rest of the day off while I restored rootvg from the mksysb tape, unless I had
more disk space than I could possibly use. I've never seen more disk space than
I could possibly use on any machine I've ever had.)

jack
--
jack      [e-mail]    * All I ask of Thee, Lord
* Christ died for our sins.  Dare we    * Is to be a drinker and fornicator
* make his martyrdom meaningless by     * An unbeliever and a sodomite
* not committing them? - Jules Feiffer  * And then to die. - Claude de Chauvigny