RMAN - 1) job scheduling 2) incremental CPU usage

[Axr2]

1) How and where do you typically initiate/schedule your backup jobs -a) cron (local server)
b) cron (remote server where your catalog sits)
c) dbms_job on local database
d) From the Netbackup/TDP/other backup admin server?

Why so?

I would think the last ('d') would be most appropriate since if the the backup server is itself down, there is no point initiating a backup in the first place. This way you exercise more control over when it makes sense to initiate the backup


2) Concepts question - Why is CPU usage so much lower on an INCR backups versus a Level0 backup? (both cases - backing up to tape with 2 channels). I would expect a small change..but not something dramatic. A full used 50% on a 4CPU box, while the INCR barely used 15%! I thought the INCR reads the entire d/f as well. Yes, it is proportional to d/b activity..and so not all blocks had to be backed up (and i/o is lower)..but still that doesn't explain the difference. Atleast it isn't something apparent to me!

[marist89]

Originally posted by Axr2

I would think the last ('d') would be most appropriate since if the the backup server is itself down, there is no point initiating a backup in the first place. This way you exercise more control over when it makes sense to initiate the backup

We wrote our own scheduler that has dependencies and pages us when certain jobs fail. The backup job runs on the server in which the database runs.

2) Concepts question - Why is CPU usage so much lower on an INCR backups versus a Level0 backup? (both cases - backing up to tape with 2 channels). I would expect a small change..but not something dramatic. A full used 50% on a 4CPU box, while the INCR barely used 15%! I thought the INCR reads the entire d/f as well. Yes, it is proportional to d/b activity..and so not all blocks had to be backed up (and i/o is lower)..but still that doesn't explain the difference. Atleast it isn't something apparent to me!

I haven't experienced that. I do incrementals to disk and my cpu usage is actually higher during incrementals because there is less to write.

[hacketta1]

At my site RMAN backups are kicked off by NetBackup. One bonus with this is that if it fails it'll try again a few more times unlike cron (unless you code this feature yourself in a shell script).

My method of being alerted to failed backups is less sophisticated than Jeffs. Given the Oracle NetBackup Agent just runs a shell script that runs rman etc. I get the script to mail me when the backup fails. If you look at the sample scripts that appear when you install the oracle agent it'll be obvious how you can achieve this. They already output a success/fail message to the logfile.