We don't come down for backups... We're 24/7 - the docs tend to want their lab results 
in a timely manner. We come down once a month for PM and that can be (and has been) 
rescheduled because of medical reasons, such as a transplant. We're usually down under 
2 hours per month.

We run HP-UX 11.0, and UV 9.5.  We used to use Logical Volume Manager (LVM) and 
MirrorDisk/UX to handle the drives. To run a backup we'd split the mirrors, backup the 
stale part and re-merge them. There were drawbacks to this method though. Either you 
had to have 3 mirror parts, or be running on a single part when split, and the 
re-merging of the mirrors took a few minutes. We did NOT have to have everyone off the 
system for the splits or the merges. 

When we moved to HP-UX 11.0, we also purchased OnlineJFS. This is a Veritas product HP 
sells. With it I can make drive changes on-the-fly both - both increases and decreases 
provided contiguous free space exists to reduce it. It also has the ability to analyze 
and defrag a filesystem online. But the really nice feature is taking a 'snapshot' of 
a filesystem. When you snapshot you get another virtual filesystem that has an image 
of how the filesystem looks at that moment. All writes to the filesystem actually go 
to the snapshots logical volume. Since only the writes go to the snapshot, it's volume 
only needs to be large enough to hold the writes themselves. The app can read and 
write as normal to it's regular volumes never knowing the difference. Vxfs and 
OnlineJFS handle it without the app knowing. When you undo a snapshot, the stored 
writes get written to the real volume to "catch up". Again this happens transparently 
to the application. We still use MirrorDisk/UX but now we never split the mirrors.

Since we're talking about a journalled filesystem, when the snapshot completes the 
filesystem is in a consistent state. That's not to say that UV couldn't be caching 
something itself that's not on disk. For us that's a smaller issue than taking the 
system down to do backups, something we haven't done in the 9 years I've been here. We 
run 3 production-only backups and a full system (plus Novell and NT/SQL) backups every 
day. I've got 28 backups from the past week, plus every Sunday system backup of the 
last month, and every monthly backup from the past year should I need to restore. To 
handle the backups themselves, we're using HP Data Protector (formerly known as 
OmniBack) though licensing can be a little high depending on backup devices, and what 
you're backing up.  To handle the snapshots at backup, I write shell scripts that does 
it and put them into the pre-exec and post-exec of the jobs. DP knows via return code 
from the scripts whether something went wrong and logs it appropriately.

Robert



>>> [EMAIL PROTECTED] 03/08/04 06:41AM >>>
I doubt there is a method that will allow backup a UV database without
stopping it and getting everybody off the system.

If you have a mirrored system you could split the mirrors, backup the
mirror and then remerge the mirrors but you would need to get
everybody off the system for the split.

My 2 cents?

Louis

----- Original Message ----- 
From: "Armon Group" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Monday, March 08, 2004 8:19 PM
Subject: UV - Database backup

: etc
:
: _________________________________________________________________
: It's fast, it's easy and it's free. Get MSN Messenger today!
: http://www.msn.co.uk/messenger
:
: -- 
: u2-users mailing list
: [EMAIL PROTECTED]
: http://www.oliver.com/mailman/listinfo/u2-users
:

-- 
u2-users mailing list
[EMAIL PROTECTED]
http://www.oliver.com/mailman/listinfo/u2-users
--
u2-users mailing list
[EMAIL PROTECTED]
http://www.oliver.com/mailman/listinfo/u2-users

Reply via email to