Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
![]() |
![]() |
Home -> Community -> Mailing Lists -> Oracle-L -> RE: rman backup to disk, mismatch in amount backed up to each channel
Unless your disk i/o communications pathways are truly separate, I would
actually suggest that you use only one channel and alternate the backup
destinations. I might even do this just for the reliability increase, even
if it takes a bit longer in the case where you really do have i/o pathways
that are non-interfering. Then alternate your channels each backup, which
preserves your ability to even out the storage size and means that if one or
the other destination fails, you still have an online backup. A belt and
suspenders. Now each alternation is going to the receiver about 75% full, so
you only have room for two copies. You'd need a little more space if you
wanted to cover the unlikely condition that "rman2" blows up (idle though it
may be at the time) during a backup to "rman1." You'd need a little more
space to have two complete at all times.
Now I did not really answer your question, which certainly would become relevant if someone had, say, "rman1a", "rman2a", "rman1b", and "rman2b" and they wanted to evenly split rman1a and rman1b on "a" days and evenly split rman1b and rman2b on "b" days. Other than declaring the limit on each to be only half the available space, I'm not sure how to do that, and I hope someone more expert than me at rman operations chimes in.
Regards,
mwf
-----Original Message-----
From: oracle-l-bounce_at_freelists.org
[mailto:oracle-l-bounce_at_freelists.org]On Behalf Of Hitchman, Peter
Sent: Friday, October 22, 2004 8:09 AM
To: oracle-l_at_freelists.org
Subject: rman backup to disk, mismatch in amount backed up to each
channel
Hi,
We back up an Oracle 8.1.7.4 database to disk using rman. Its about 900 Gb.
It does work, but we find that much more data is getting sent to one channel
than the other. So when its done, one filesystem is over 50% full and the
other is only at 25%. I thought that rman automatically evened out the work
done by each channel and that would result in an even (ish) amount of disk
space being used in each filesystem.
What we want is for the amount written to each channel to be about even, that way we can do a backup and then rename the directories the backup is in before we run another one, in case the next backup fails for some reason. We only two file systems available (and we had to beg for these) and managed to get enough space in each in theory to hold two disk backups. But its no good if one of the filesystems goes over 50% with the first backup, because it means the next one will fail because of lack of disk space.
Does anyone know how we can control rman's behaviour in this case? (By the way, for reasons that are beyond me, it is not possible to change the disk space allocation so that the one of the filesystems is 25% bigger that the other).
Here is the rman script:-
connect target 'xxx/xxx_at_XXX';
connect catalog 'xxx/xxx_at_XXX';
run {
allocate channel t1 type disk format '/rman1/backup/cms.%U';
allocate channel t2 type disk format '/rman2/backup/cms.%U';
set limit channel t1 kbytes 10000000;
set limit channel t2 kbytes 10000000;
set command id to 'rman_backup';
backup
incremental level 0 cumulative
skip readonly
skip offline
database;
sql 'alter system archive log current';
change archivelog all crosscheck;
backup
filesperset 64
(archivelog all);
backup
current controlfile
tag = ctrlfile_level0backup;
release channel t1;
release channel t2;
change archivelog all crosscheck;
}
Thanks
Pete
[END]
-- http://www.freelists.org/webpage/oracle-l -- http://www.freelists.org/webpage/oracle-lReceived on Fri Oct 22 2004 - 08:49:37 CDT
![]() |
![]() |