rebuilding disk systems on 2100

From: Jon Earle <jearle_at_dmcs.dnd.ca>
Date: Thu, 16 Apr 1998 12:17:44 -0400

<EXT:PLAINTEXT_OK>

Hello all,

I've been tasked to rebuild the disk systems in an Alpha server to
alleviate some of the stresses that Oracle is placing on the system. The
DBA has selected a disk/partition setup/allocation schema that should suit
his needs. He's going to take care of the database side of things, I just
need to get the system operating.

(Which is why I was asking the question yesterday about the RZ1DB disk)

I've used the SWXCRMGR utility on Alpha system many times to build disk
arrays and the like, so that is no problem. I'm curious now as to any
other "gotchas" that I might be overlooking when rebuilding this system.
Here is my plan for the operation:

1. Previous night, full vdump of all file systems. (Oracle export will be
taken care of)
2. Brew pot of coffee. (Critical) :)
3. Replace disks / rebuild RAID and standalone disks to new requirements.
4. Install OSF/1 4.0a from CDROM via Advanced option (so I can create
AdvFS, Partition disks ...) Other things?
5. Restore previous system (/, /usr/, /var, /users, etc) from tape via
vrestore.
6. Reboot system, use management option from CD to massage files like
/etc/fstab. Others?

This is the first time I've done anything like this in OSF/1 (It'll be the
first time I've installed it), so I'm interested in anything that I may
have overlooked or that might bite me once this has begun. I've asked for
all of the product CDs, license paks, etc, and I'm assuming that the
installation process will request the proper license key, CD, etc when
required. Anything specific that I should have at my fingertips?

For the Oracle gurus, here is the situation leading up to this:

The system is currently residing on a big RAID5 drive - all data, indexes,
etc are located on this one drive. The DBA has a requirement from time to
time, to import data into the database during working hours. This of
course, causes the usage of the RAID drive to skyrocket. Users can still
connect, but it's really slow. What he wants to do, is eliminate the
RAID5, and build a series of smaller RAID0 drives - each will separately
hold the data, indexes, etc which according to his readings and whatnot,
will provide quicker response times when importing data during working hours.

As I know nothing about Oracle, I nod wisely and say Ok. Can anyone
provide verification that this is a good way to proceed, or is there a
better way to provide quick access to the database while a large import is
taking place? (The import is occasionally required as the data is critical,
and generally needed as soon as the tapes are received.)

Thanks!
Jon



---
Jonathan Earle
Technical Architect
HUB Computer Consulting Inc.
Pager: (613)751-4948
Email: jearle_at_*DELETETHIS*hubcc.ca
       "You can never burn out as long as you're doing what you enjoy. 
When it stops being fun, then you burn out.  And I'm still having fun."
                       -- J. Michael Straczynski, creator Babylon 5
Received on Thu Apr 16 1998 - 18:23:05 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT