rman maxsetsize
- From: Stephen.Lee@xxxxxxxx
- To: oracle-l@xxxxxxxxxxxxx
- Date: Fri, 18 Jun 2004 09:33:14 -0500
I'm still experimenting with backing up to disk then copying to tape. Since
we don't have control of Networker settings like total parallelism and
target parallelism per tape drive, and sometimes enterprising SA's
experiment with these, a problem can happen (read on).
My script starts up a number of save commands equal to the total parallelism
as determined by
MAX_JOBS=`{
/usr/opt/networker/bin/nsradmin -s $NSR_SERVER <<-XXX
show parallelism
print type:nsr
exit
XXX
} | /usr/bin/sed -n '/parallelism/s/\(.*\)\(parallelism:
*\)\([0-9][0-9]*\)\(.*\)/\3/p'`
The commands are started in the background:
/usr/opt/networker/bin/save -i -c "$NSR_CLIENT" -s "$NSR_SERVER" -e
"$EXPIRE" -b oracle "${DIR}/${i}" >> "$LOGFILE" 2>&1 &
Then the following controls the number of save commands running at any one
time:
while [ `jobs | /usr/bin/wc -l` -ge $MAX_JOBS ]; do
sleep 59
jobs > /dev/null 2>&1
sleep 1
done
Then loop back around to start another save command.
Now suppose we have 4 tape drives and "somebody" has set total parallelism
to 20 and target sessions per drive to 1 (why would I use these figures for
an example?). What can happen is that Networker will grab the first tape
drive that is ready to write (i.e. that drive doesn't need to swap tapes, or
relabel a tape, etc.) and pile all or most of the sessions onto that drive.
If rman's save sets are somewhere around 30 Gb, this results in a bunch of
stuff getting assigned to one drive. As that drive fills up and must do a
tape swap, the sessions have become best buds and will faithfully stick
together even though they must move to one of the other drives.
To limit the number of bytes that can get stacked up on one drive, I had
what I thought was a grand idea to use the maxsetsize parameter to limit the
rman save sets to 4GB. Well, rman decided it didn't like that:
RMAN-00571: ===========================================================
RMAN-00569: =============== ERROR MESSAGE STACK FOLLOWS ===============
RMAN-00571: ===========================================================
RMAN-03002: failure of backup command at 06/17/2004 17:15:43
RMAN-06183: datafile or datafilecopy larger than SETSIZE: file# 10
/u14/oradata/XYZ/vehhst_data_01.dbf
Sooooo....
Would there, perchance, be any among the rman cognoscenti, who could shine a
light on the finer points of the maxsetsize parameter? (or provide a good
URL to said light.) The size of the data file listed above is 32000M or
34168840192.
The version here 9.2.0.4.
By the way: I think there was no 9.1 because of the agreement that Oracle
does the even numbered version, and Microsoft does the odd numbered
versions. (ms/pc-dos)
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to: oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------
Other related posts:
