upgrade netbackup and solaris

[dennis_sexton]

Have a new machine I'm intending to install in place of current media/master server.
I only have the one master/media server. The good news is
that the database is not local but on a SAN disk that I can just mount from
the new host.

I am currently on Netbackup 4.5 fp6 on solaris 9. I want to get to at least Netbackup 5.1
and solaris 10. As best as I can tell, I cannot install 4.5 on solaris 10 even as master.
Thus the steps must be something like

install solaris 9 on new host.
install netbackup 4.5 on new host.
shut down old host.
turn on new host.
confirm all is working.
upgrade netbackup to 5.1.
confirm all is working.
upgrade solaris to 10, using live upgrade.
bring up solaris 10
confirm all is working.

To work on the new host before substituting it for the old host, i need to
give it a different name from the master server. when I install veritas,
can i just put the original master name in bp.conf or is the current hostname
squirreled away in some dark and deep places?

Any comments on the workability/gotchas in this plan appreciated.

[Raghava_Chary]

Dennis,

Modification of the hostname of a NetBackup server to which backup or archive images have been written will result in the inability to easily restore from those images. This is because the server's hostname is associated with the image in the NetBackup image database.

Another problem will be encountered when the images expire, NetBackup not be unable to delete these expired images.

I have the below documented with me but i never tried this before .. it may help.

*******************************************************************************************
Overriding the Media Host
-------------------------

NetBackup allows the administrator to define a mapping in the NetBackup
configuration (bp.conf on UNIX, registry on Windows NT) to request that all
restores that need to use media server A be changed to use media server B
instead. This method requires administrator intervention. Some examples of
when to use this method:

* Two (or more) servers are sharing a robot, each have connected
drives. A restore is requested while one of the servers is either
temporarily unavailable or is busy doing backups.

* A media server has been removed from the NetBackup configuration,
and is no longer accessible.

Note: The media server override applies to all storage units on
the overridden media server. This means restores for any storage
unit on fromhost will go to tohost.

To override the NetBackup media server for a restore:

1. If necessary, physically move the media and update the Media
Manager volume database, using the Media Manager administrative
utilities.

2. Modify the NetBackup configuration on the master server:

UNIX NetBackup server:

As root user, add the following entry to the
/usr/openv/netbackup/bp.conf file:

FORCE_RESTORE_MEDIA_SERVER = fromhost tohost

where fromhost is the media server that performed the original
backup and tohost is the desired media host to use for the
restore.

Windows NT NetBackup server:
Use the RegEdit application to modify the key

HKEY_LOCAL_MACHINE\SOFTWARE\VERITAS\NetBackup\CurrentVersion\Config

to add the value FORCE_RESTORE_MEDIA_SERVER and then set the
newly added value to fromhost tohost, where fromhost is the media
server that performed the original backup and tohost is the
desired media host to use for the restore.


To make expired media available again
-------------------------------------

Here are the options to be run with the bpexdate command to
make media available:

The bpexpdate can fail with this error for a number of reasons:

The command was issued on a master server, the images are
expired but the mediaDB has info on this tape.
use the -justmedia option appended to the command.

The command was issued, images exist but the mediaDB is not
available.
use the -justimages option appended to the command.

If you have to use the -justmedia or -justimages option, you will have to
deassign the tape manually.

The pool number and status number can be obtained by:

vmquery -m
then vmquery -deassignbyid

Example:

run /usr/openv/volmgr/bin/vmquery -m to get info needed on master
server. example: vmquery -m A00001 yeilds:

======================================================================
media ID: A00001
media type: 8MM cartridge tape (4)
barcode: 00A00001
description: Added by Volume Manager
volume pool: djmtest (3)
robot type: TL8 - Tape Library 8MM (6)
robot number: 22
robot slot: 42 (F02)
robot host: wave
volume group: 00_022_TL8
created: Fri Jan 10 16:31:35 1997
assigned: ---
last mounted: Thu Apr 3 12:34:08 1997
first mount: Thu Feb 6 10:24:28 1997
expiration date: ---
number of mounts: 32
max mounts allowed: ---
======================================================================
next, write down the pool number -in parens- our example pool number is 3
ex. (volume pool: djmtest (3)
then get the status number... usually looks like 0x0 (netbackup) or 0x1
(netbackup database) you'll want the last number 0 or 1
Note: you may not see the status listed (like in my example) use 0 or 1 as
appropriate anyways.


Then on master run vmquery -deassignbyid

It will look like vmquery -deassignbyid A00001 3 0

Then run:
/usr/openv/netbackup/bin/admincmd/bpexpdate -justmedia -ev A00001 -d 0

[dennis_sexton]

Raghava,

Thanks for this info. Knowing the difficulty about the database, I do not intend for the new host to
get a new name. It will take over the
name of the old master server. But in prepping it, I can't use its eventual name, I have to
use some other name. What I'm concerned about is whether there are some config files
somplace that might have the unwanted name lurking that will cause me difficulties once
I make the switchover.

[Raghava_Chary]

Dennis,

I would say NO, other than volume database, device and catalog.
I would like to suggest you to contact Symantec support before you change the hostname.

Raghava.

I just requested that info, when i get that doc i will let you know.Message was edited by:
Raghava Chary

[Raghava_Chary]

Dennis,

Unfortunately, the netbackup catalogs are directly tied to the hostname of the mster server. As such no way exists within the software to change the hostname of a master server since each and every record is hard coded with the name of the master server.

If you must change the name, this can only be accomplished by engaging Netbackup Professional Services to come on site and go through the catalog part by part to make sure each and every entry now reflects the new hostname.

Because of the complexity and risk of data loss associated with this task, this is something that requires Professional Services on site.

Raghava.

[Chia_Tan_Beng]

Dennis,

My suggestion is that you install & prepare your new Solaris 9 and NBU45 with same hostname but without connecting to the network. Perform some prelim tests such as mounting SAN disk (catalog) to new server, startup NBU, test catalog backup, backup/restore a part of master server's partition, Vault, etc, Lastly, shutdown the current server when you connect new server to network and try network backup, etc.

Have a 2 phase cutover, 1st cutover to new server with Solaris 9 & NBU45 (perform a backup of catalog to tape and/or snapshot another volume to save keep for contingency). Only when all are stable then enter the phase 2 of upgrading OS & NBU.