I'll let someone else address whether the over-all process is correct for
what you're trying to do, and just hit the key point -- the copy pool
checkouts.

If these are going off-site for any kind of D/R or "cover-your-backside in
the event of environmental problems" (flood, fire, earthquake, other acts of
God or man) then they need to go DAILY, full or not. Along with a TSM
database backup (if you don't have the TSM database you have used but empty
tapes).

If you're concerned about mostly empty tapes off-site, run reclaims. They'll
get logically rebuilt from the originals again, another day's data gets
added, and THESE tapes go off-site. Bring back the now-empty off-site tapes.
Repeat until tapes wear out. My DLT tapes were still going strong after four
years of this sort of treatment.

One of the tapes I checked out this morning was 13.9% full -- and one of my
real long-term archive pools can give me LTO tapes that list as 0.0% or 0.1%
when they first go off-site. As I've said before -- in the great scheme of
things, tape is cheap.

Tom Kauffman
NIBCO, Inc

> -----Original Message-----
> From: Johnn D. Tan [mailto:jdtan AT MED.CORNELL DOT EDU]
> Sent: Friday, September 13, 2002 9:04 AM
> To: ADSM-L AT VM.MARIST DOT EDU
> Subject: archive being done right?
>
>
> Just wanted to double-check that our archives are actually being done
> correctly (I wasn't the one who set it up). This is more conceptual
> help that I need here.
>
>
> We run TSM Server 4.1.2 (yes, I know, it's EOL, but we are upgrading
> in next few weeks) on AIX 4.3.3.
>
> We already have a nightly backup policy which basically retains
> active data for 90 days and inactive data for 365 days. This is fine
> for 95% of our servers.
>
> However, we now have a few servers where we need to do point-in-time
> restores for up to a year back.
>
> So we created a management class which basically retains active data
> for 365 days. For the servers that need it, we still do a nightly
> backup to the 90-day class. But every Sunday, for these small handful
> of servers, we also run an archive to the new 365-day management
> class.
>
> Then, every Sunday evening, the archive pool gets backed up to a
> archive pool copy. Monday morning, we check out all the archive pool
> copy volumes, EVEN if they say they're still "Filling." We then label
> these tapes indicating the week they were for, and take these offsite.
>
> So, this is not a "true" point-in-time restore. But it's good enough
> for us. If we want to restore as of <any given date in past year>, we
> simply restore from the weekly archive tape of that date, which will
> put us to that Sunday before the date. Then we restore the database
> journal files that are backed up in the normal way. Since these
> journal files have a new file (and filename) each and every day, they
> are retained for up to 365 days under the 90-day mgmt class even when
> they are deleted/purged off the system (as stated above, inactive
> data kept for up to a year under the 90-day class). And then we roll
> these journals forward from Sunday to the day we need them.
>
> Is this the correct way to be using archive? I guess my biggest
> concern is that we are pulling out "Filling" tapes and storing them
> away. For our normal nightly backups, we only pull out "Full" volumes.
>
> johnn
>