Hey everyone,
I'm pondering something right now but don't have a test environment
yet to test it. Here is the situation:
- I have an AIX system with two different types of disk sub-systems hooked
up to it via Fibre card/HBA connections - (shark, and EMC disk).
- I want to create two AIX filesystems one on the shark DASD and the other
on the EMC DASD.
- I want to format and restore or production tsm database and tsm recovery
log onto those two filesystems.
- I then will bring each tsm server instance up with those restored
databases and logs respectively.
- Now I have two databases running on the same machine and I will change
the options file and create another server stanza in the client options
file, so each server instance will be listening on different ports of the
same physical AIX server. We'll call the two tsm server instances port1500
and port1600.
- Now Let's say I have client backups goto the server which sits on 1500
one night, then I shutdown that server instance after the backups complete
to that server. Then The next day I want to test backups on the other
server instance 1600, but I don't want to re-restore the tsm server 1500
database the next day so that 1600 see's 1500's backups from the night
before. So I was thinking can I then shutdown the 1600 tsm server instance
and re-point that instance to see 1500's tsm recovery log, then bring that
1600 server instance up, then the question is will 1500's recovery log roll
1600's recovery log forward to restore 1600 to the point in time that 1500
was at after the backups finished last night and when it shutdown?. I know
there's some internal pointers which may cause a problem but not sure. I
know you can do something similar to this in Oracle with cloning a database
and with a standby type of database. I want to apply the recovery log
without restoring the entire tsm database.
I just have to make sure I don't take a database backup on the 1500 server
before I rollforward the 1600 server because then the recovery log will
empty (beacause it's in rollforward mode).
The reason why I want to see if this works and test it. Is because I'm
currently evaluating two different disk sub-systems for performance and
other things. I want to backup the production database which currently
resides on sub-systemA, then restore it to a filesystem which resides on
sub-systemB, then "disable sessions" on the prod and change the server
options file on subsystemB so that the restored tsm database can listen on
the productions port, (this way you don't have to touch every client
system), then restart the restored database system. Now that night we will
run the backups which once ran to production to this restored tsm database
instance. Now remember this restored instance has the productions' tsm tape
drive definitions and since they're on the same physical system there's no
deviation from the AIX "rmt#" definitions. Now once the recovered 1600
server instance is up run production backups to that tsm instance that
evening. Then the night after that night's backup runs shutdown the
restored 1600 tsm server instance, and point the sub-systemA, 1500's tsm
database to point to now shutdown 1600's recovery log then rollforward
1500's database so that it knows about the backups which were backed up to
1600's server the night before, then shutdown 1500 again then run database
backup on 1600 to empty the log and go through the backups again, if we do
that daily then once the evaluation is officially over in a few weeks I can
just rollforward 1500's tsm database to the latest 1600 recovery log file
then we wouldn't have missed anything and we wouldn't have to restore
1600's tsm database back to 1500's subsystemA.
To make a long story short, I want to know if I can rollforward a tsm
database from another tsm server's recovery log if both tsm databases are
based on the same base tsm database either from a restore or whatever?
Thanks in advance!
--Justin Richard Bleistein
Unix/TSM Systems Administrator
|
