You didn't complete the final sentence in #2 as to why you want
to do a full backup. It starts with "Hogwash. I do not wish to have my
only copy of 90% of my files on one tape, and then find out that the one
tape". If you mean that you want to do a full copy to make sure you
have an alternative if a tape is damaged, then what you really want to
do is set up copy pools. That is what you'll want to use for your
offsite tapes. Basically, after defining the pools, you'll want to
issue the "backup stg" command for each local storage pool you have. If
you lose a local volume (disk or tape), then you do a "restore vol" and
it will ask for the offsite volume(s) that contain the backup files on
the destroyed volume. Look in the ADSM for AIX V2 Admin Guide, chapter
11, pg 238-245.
If you change the include/exclude file to point to a different
management class, then the previous backups will get rebound to the new
management class and their retention changed.
This should also answer your other questions about offsite
tapes. If you meant something else with your unfinished reasoning on
full copies, please repost your questions/concerns.
Tina Hilton
Thomson Consumer Electronics
----------
From: John Schneider[SMTP:jdschn AT ibm DOT net]
Sent: Friday, March 06, 1998 9:27 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Can you have files with multiple Management Classes?
Greetings,
I am referring to the AIX ADSM Server 2.1 PTF 15 in my
question.
I have a need to do the following:
1) Backup all the files on about a dozen AIX systems, daily
incremental. The /, /usr, /var etc. directories go to a
management
class "backup-os", retention 3 weeks. All other filesystems go
a
management class called "backup-generic", retention. Backups go
directly to DLT tape on a central STK9714 library on AIX server.
2) On a weekly basis (or monthly, depending on how many tapes
this
takes), take a "full incremental" (in other words, all files
whether
they are changed or not) of all the nodes to the same management
classes
as in 1). Same retention. I know people say there is no
reason to
every perform a "full incremental", because ADSM still has the
data.
Hogwash. I do not wish to have my only copy of 90% of my files
on one
tape, and then
find out that the one tape
3) On a weekly basis, take a "full incremental" of all the nodes
to a
different set of tapes in the same DLT library, with a retention
of 6
months. This backup needs to go to a separate tapes because
the tapes
sent offsite.
4) An alternative to 2&3 would be to take a "full incremental",
and
simultaneously send it to two different tape storage pools, one
of which
would stay onsite, and one which would go offsite. A "twin" or
"mirrored" backup.
Step 1) was straightforward, and is working fine. Each
node has it's
own include/exclude file to spell out which filesystems get
backed up to
which management class. But I am not sure how to do step 2)
properly.
I could create a separate schedule for a "selective" backup, but
that
is kludgy because in the options parm you have to specify each
filesystem. Some of the larger systems have a couple dozen
filesystems,
and some get backed up while others don't, and the list changes
frequently, and it just looks like a lot of work to keep it
updated and
correct. If someone forgets to add a filesystem to the list, it
doesn't
make it in to the "full" backup.
The other way to do step 2) would be to change the
management class's
backup copy group definition so that a mode of "absolute" were
specified. I could set up a cron job on the server so that once
a month
the backup copy group definition were changed, and the next
incremental
backup performed for each node would back up ALL. A subsequent
cron job
run 24 hours later would change it back. Any better ideas?
Step 3) is even harder. How do I back up files, and
give them a
different retention period, and send them to different tapes,
just one
day of the week? The ways I could think of all have problems:
1) Code a cron script that modifies the include/exclude list on
each
machine to change the management class. The different
management class
would have a backup copy group with different retention times
and
storage pool. The script could run on the day you plan to take
the
backup, then get changed back the next day. But you would have
to
monitor this, and make sure it was working properly. Each new
node you
set up would have to have this same cron job set up.
But wait! According to pages 133-134 of the Admin
Guide, the day after
you backed up the files with a 6 month retention, ADSM would
rebind them
back to the old management class, and the retention would go
back to 5
weeks. So this would not work, unless I am misreading it.
2) Set up a separate policy set, and give it management classes
that
have the same names as the original policy set. The management
classes
for the new policy set would specify different retention periods
and
storage pool. A cron script would activate the new policy set
on the
day you want, and then switch it back the day after. But would
this
work? The Admin Guide implies that the files would not be
rebound to
the other management class because the same name exists in both
policy
sets.
But after the original policy set gets put back and a
backup runs,
would the files backed up using the other policy set still have
the
those values, or would they revert to the values of the new
files being
backed up? I am not clear on this.
Is there any way to do Step 4 at all? Other backup
products (like
Legato Networker and HP Omniback) allow for what they call a
"twin" or
"mirror" backup, where you back up to two tapes at once. Then
you don't
have to worry about your data disappearing because of one bad
tape, and
you can send tapes offsite without worrying that you will need
them
tomorrow to do a restore, etc. This capability would be very
nice. Any
clever ways to get this done?
Sorry for the long note, and thanks in advance for any advice.
This
list is terrific. I have been reading it for awhile now, but
just
decided to post since I have not seen anything just like it go
by.
Best Regards,
John Schneider
***********************************************************************
* John D. Schneider * Email: jdschn AT ibm DOT net * Phone:
314-349-4556 *
* Lowery Systems, Inc.
***********************************************
* 275 Axminister * Disclaimer: Opinions expressed here
are mine*
* Fenton, MO 63026 * and mine alone. My company is off the
hook.*
***********************************************************************
|