How are the Oracle backups. I had the same problem you are experiencing. I
had the backups for one customer who had a 1 TB database, with the files set
to /u01,/02, etc. That was a nightmare trying to back up that amount of data
in the window we were given. With that structure, backups would constantly
fail because a new database was added to a mount point /U02/* and the backup
group were never notified. This customer had 5 domains spread across a Sun
E10K and the Oracle instances spanned across 3 of these domains, which made
things more difficult. To fix this problem and it work extremely well, one
of our backup admins, Blaine-who is on this list, removed all the /U01,/U02
files sets, and created policies for each instance. So, one policy would
have one, two, or all 3 domains/servers, with only one oracle instance. This
gave the admins control over what was backed up and forced the dba's to
notify us when a new database was being created. With this amount of data,
we wrote to 4-9940A drives with a back window of 2230-0430 and rarely had a
failure. With the restructuring I described above, you can piggy back
databases with many small files with binaries to increase throughput. This
was only one customer and we backed up 4-6 TB's every night. Let me know if
this is something you want to try and I'll can give you more specific
information on this.
-Dwayne
Dwayne J. Brzozowski
Department of Veterans Affairs
Night Shift Supervisor-Unix Group
Austin Automation center
(512)326-6728 work
dwayne.brzozowski < at > mail.va.gov
-----Original Message-----
From: veritas-bu-admin < at > mailman.eng.auburn.edu
[mailto:veritas-bu-admin < at > mailman.eng.auburn.edu]On Behalf Of Dobbertien,
Matthew
Sent: Thursday, January 27, 2005 4:59 PM
To: veritas-bu < at > mailman.eng.auburn.edu
Subject: RE: [Veritas-bu] backing up LARGE oracle databases (looking for
suggestions)
I don't necessarily have any good suggestions, but we have a similar issue
and are hoping to find some resolution. We currently have a 600 GB database
that is expected to grow to over 2 TB in the near future. We are trying to
plan for the additional capacity and make the users aware of timeframes for
backups. Currently we are backing up over Gb Ethernet, which seems to max
out at about 60-70 MB/s, even in vendor (Sun) tests. This is barely enough
to maximize throughput on 1 LTO2 drive, taking into account compression. We
are currently writing to 2 LTO2 drives with an L700e library (8 total LTO2
drives, soon to be 10). We also have a large VTL (about 50 TB), which is
not yet installed, which will serve as the primary location for backups.
Right now, the chokepoint is the GigE connections. I'd like to be able to
use snapshots to backup the data directly to tape, but there doesn't seem to
be a good way to manage the snaphshot process using Netbackup with Clariion
arrays. We also considered making the servers SAN media servers, but since
these applications reside on a Sun15k, the cost is prohibitive, and the
extra complexity of having to involve the system owners for backup
maintenance is not desired.
Any comments on expected backup speed, methods of improving backup speed for
large backups (I understand that the world record is 2 TB an hour, but I
don't think that we can achieve that in our environment) and any other
information would be helpful. If you could convince the database owners to
use RMAN, that would help as well, but I'm not going to address that now.
Matthew Dobbertien
Sr. Systems Analyst
Chicago Tribune
_____
From: veritas-bu-admin < at > mailman.eng.auburn.edu
[mailto:veritas-bu-admin < at > mailman.eng.auburn.edu] On Behalf Of Rob Haig
Sent: Thursday, January 27, 2005 10:14 AM
To: veritas-bu < at > mailman.eng.auburn.edu
Subject: [Veritas-bu] backing up LARGE oracle databases (looking for
suggestions)
So I'm trying to wrap my head around this problem. I'm backing up a 4TB
oracle database with 4 LTO2 drives, connected to the server via 1GB
Ethernet. That's not that bad of a problem. RMAN, lvl0 backups twice a
week, plus daily lvl1 backups plus archive logs. However, when you throw
into the mix that I have 20 of these servers (4TB is the average DB size,
they very from about 200G to 7TB) to back up, it gets more interesting.
I'll be able to spend money eventually, but not right now. My eventual plan
is a few fully populated SL500 libraries with 4 drives each. In the final
plan, I'd like to have 8 drives each library and about 150TB of SAN in
between the DB servers and tape drives with serverless backup from the SAN
to tape. Or, a big honkin' virtual tape library full of disks that can do
serverless backup and has several SAN connections.
But right now, I'm limited to 2, maybe 3, 2-drive LTO2 changers with enough
slots to back up about 8TB of data each before I have to shuffle out full
tapes. I've been pounding my head on this problem for a few days now and
every time I think I have it sorted out so it's not that bad, I remember
that I have 20 of these databases.
Anyone have any suggestions?
--
Rob
