i am attempting to speed up my rman backups as much as possible. it looks like we are only using 3 of the 72 available cpu cores. why ? how can i optimize my rman backups better? how can i use all 72 available cores? does this matter ?
oracle enterprise database version 12.1.0.2
linux redhat version 6
rman configuration as follows:
- backup optimization on;
- device type disk backup type to compressed backup set parallelism 72;
- channel 1 device type disk format '/pcie01/backups/%d_%T_%U';
- channel 2 device type disk format '/pcie02/backups/%d_%T_%U';
- channel 3 device type disk format '/ssd01/backups/%d_%T_%U';
- channel 4 device type disk format '/ssd02/backups/%d_%T_%U';
- channel 5 device type disk format '/ssd03/backups/%d_%T_%U';
- channel 6 device type disk format '/hd01/backups/%d_%T_%U';
- channel 7 device type disk format '/hd02/backups/%d_%T_%U';
- maxsetsize to unlimited;
- encryption for database off;
- compression algorithm 'medium' as of release 'default' optimize for load false;
rman report schema:
1 10240 system
2 5120 sysaux
3 16223 undotbs1
4 17050 users
5 358400 ssd02
6 256000 pci01
7 256000 ssd03
8 5120 audit_trail
9 256000 hd02
hp hardware has several PCI-E ssd and other SSD storage units as well as typical hard drives. i've tried to separate tablespaces and split up the rman backup output using all available separate physical devices.
our full cold backup script is as follows:
connect target /
run {
shutdown abort;
startup force dba;
shutdown immediate;
startup mount;
backup as compressed backupset database plus archivelog section size 5G tag='weekly_full_backup';
delete noprompt obsolete device type disk;
startup mount;
sql "alter database open";
}
exit($result);
---------------------
during rman backup i monitor the progress.
- the linux top command shows 1-3 oracle user processes at the top with %CPU of 90%+- ... i would have expected many more oracle processes than 3 all using 90%+ of cpu or show the percentage above 100%
- oracle enterprise manager 12c cloud control shows top activity of the rman process using only 3 CPUs ... i expected the em to show at least 12 cpu being used... not just 3. it also shows only CPU+CPU Wait... almost no USER or SYSTEM I/O...
I have attempted to follow guidlines from the following oracle online docs. they have helped only marginally.
579158.1
https://support.oracle.com/epmos/faces/DocumentDisplay?_afrLoop=234824633063902&id=579158.1&_adf.ctrl-state=12te8l153z_1…
-- and --
https://docs.oracle.com/cd/E11882_01/backup.112/e10642/rcmtunin.htm#BRADV011
select sid, serial#, context, sofar, totalwork, round(sofar/totalwork*100,2) "%_complete" from v$session_longops
where opname like 'RMAN%' and opname not like '%aggregate%' and totalwork != 0 and sofar <> totalwork;
any suggestions?
sincerely,
user8946061