RMAN Backup Times Anomaly
- From: "Sam Bootsma" <sbootsma@xxxxxxxxxxxxxx>
- To: <oracle-l@xxxxxxxxxxxxx>
- Date: Wed, 2 Jan 2008 16:28:09 -0500
We have two almost identical databases - both cloned in November from
the same database. While troubleshooting a problem related to the RMAN
backup of one of the databases, I discovered the backup time for one of
the backup pieces was vastly different for the two databases. For one
database, the backup time was 9 minutes, for the other it was 34
minutes. The files backed up in this backup piece were the same and the
sizes were almost exactly the same too. Can anybody suggest why backup
of the same data takes so much longer with DEVL then with TEST, for the
first backup piece? Thanks for any suggestions!
Here are more details
- We have databases B, and C.
- B was cloned from database A on November 7/07.
- C was cloned from database A on November 9/07.
- We do a full RMAN backup of B and C to tape once a week on Sunday (a
low activity period). These backups are generally successful.
- The backups occur at different times and do not overlap
- We consistently receive a "Touchnet" error during the backup of C.
This Touchnet error is NOT an RMAN error, but an error from another
application.
- We believe this Touchnet error is occurring because the RMAN backup of
C is hogging all the CPU leaving the Touchnet related process
CPU-starved.
- We rarely or never encounter the Touchnet error during the full RMAN
backup of B.
- I switched the backup times of B and C, and for about 3 weeks we did
not receive the Touchnet Error
- The past two Sundays we received the Touchnet Error again.
I reviewed the RMAN log files for the most recent RMAN backup and
discovered:
- On C, the first backup (piece 1) using channel 2 took almost 34
minutes
- On B, the first backup (piece 1) using channel 2 took just over 9
minutes
- In total, there were 14 backup pieces in each database
- The time differences between B and C for the other backup pieces are
much closer together.
I suspect the lengthy backup time for the first backup piece on C may be
related to the Touchnet Errors we receive.
I confirmed the files backed up in this piece were the same in both
cases. And the file sizes are almost identical:
select file#, name, bytes/1024/1024 MB, status
from v$datafile
where file# in (18,17,26,27)
system@C> /
FILE# NAME MB STATUS
---------- ---------------------------------------- ---------- -------
17 /san1/oradata/C/help01.dbf 200 ONLINE
18 /san1/oradata/C/index01.dbf 12047 ONLINE
26 /san1/oradata/C/mercury01.dbf 148.75 ONLINE
27 /san1/oradata/C/banaq01.dbf 120 ONLINE
system@B> /
FILE# NAME MB STATUS
---------- ---------------------------------------- ---------- -------
17 /san1/oradata/B/help01.dbf 200 ONLINE
18 /san1/oradata/B/index01.dbf 11883 ONLINE
26 /san1/oradata/B/mercury01.dbf 148.75 ONLINE
27 /san1/oradata/B/banaq01.dbf 120 ONLINE
Can anybody suggest why backup of the same data takes so much longer
with C then with B, for the first backup piece? Thanks for any
suggestions!
Sam Bootsma
Oracle Database Administrator
Information Technology Services
George Brown College
Phone: 416-415-5000 x4933
Fax: 416-415-4836
E-mail: sbootsma@xxxxxxxxxxxxxx <mailto:sbootsma@xxxxxxxxxxxxxx>
Other related posts:
