RMAN and archivelogs

[BSG]

Hi All,

We are about to implement RMAN instead of brarchive/brbackup. For us the major disadvantage of using RMAN is the way it handles deletion of the archivelogs after backing them up to our TSM storage manager.

When using brarchive a single archivelog is deleted from the archivelog filesystem once it has been backed up to TSM, in this way freeing up space in the archivelog filesystem. When using RMAN the backed up archive logs are only deleted when the complete bunch of "to be backed up archive logs" are actually backed up to TSM. This would mean that if the archive log filesystem fills up (e.g. due to problems with our storage manager or heavy archivelog creation) the space in the filesystem is only freed after the complete RMAN cycle has finished. The database would be hanging untill that time. When using brarchive a single archive log will be deleted once it has been backed up, thereby creating free space in the filesystem immediately and the database to be running normally again.

In order to keep our systems running when the oraarch filesystem is filled up and RMAN is running we can increase the size of the filesystem untill the RMAN cycle has finished and the archivelogs are subsequently deleted. However we are wondering if there is any other way in which we can make sure that the system keeps running. E.g. is it possible to use an 'overflow archivelog filesystem" to which oracle switches once the initial archivelog filesystem fills up? Or any other method?

What's best practice for this?

Thanks.

Regards,

Bart Groot

[BSG · Accepted solution]

Hi Stefan,

Thanks for your reply.

How exactly did you set the FILESPERSET specifically for the archive log backups?

In the init<SID>.sap file I do see the rman_filesperset parameter (which is commented at this moment). If I set it, is there any way to make a difference between archive log backup and database backup?

Thanks.

Regards,

Bart.

[stefan_koehler · Accepted solution]

Hello Bart,

i am sorry, but in your previous post you said that you want to "implement RMAN instead of brarchive/brbackup".

Now you are talking about the init<SID>.sap file again.

If you use native RMAN you have some backup scripts (for full backup, archive log backup, etc.) in which you can specify the FILESPERSET clause. If you still want to use the BR*Tools with RMAN (which uses the init<SID>.sap file) you can specify the FILESPERSET clause globally .. so it is used for archive log files and full backups, etc.

But as i already mentioned that should be no problem, because of in huge environments not all files should be in one backupset.

http://help.sap.com/saphelp_sm32/helpdata/de/82/29e03a00281877e10000000a11402f/content.htm

Regards

Stefan

[BSG]

Hi Stefan,

We will perform some tests to find out which settings of rman_filesperset will do best for us.

Thanks again for sharing your knowledge. Much appreciated.

Regards,

Bart Groot.

[BSG]

Hi Stefan,

Thanks again for your reply.

We were planning to use BRTOOLS to with RMAN. Regarding BRTOOLS, the rman_filesperset is what I was referring to in the init<SID>.sap file. Is there any specific reason why we should native RMAN instead of the "BR*TOOLS" RMAN?

Thanks,

Regards,

Bart.

[stefan_koehler]

Hello Bart,

Is there any specific reason why we should native RMAN instead of the "BR*TOOLS" RMAN?

.. that depends on your kind of infrastructure (TDPO, TDP for mySAP or any other storage agents for direct tape backup) and on your oracle knowledge. If you use native RMAN you are "free" to do what you want without any limits, but you don't get any support by SAP for it ..

Regards

Stefan

[stefan_koehler]

Hello Bart,

if you are using native RMAN you can specify how many archive logs should be included in one backup set .. so you can get the same behavior as with BR*Tools.

For large systems its also recommended not to put all archive log / data files (or mostly 64 files ... depends on OS, RMAN optimization, etc.) in one backupset ... just think about a restore and recovery if you only need one file .. you need to run through the whole backupset to get only that one particular file.

I use a backupset "size" of 5 for archive log files and 3 for data files in our huge environments.

Check the documentation for the parameter FILESPERSET:

http://download.oracle.com/docs/cd/E11882_01/backup.112/e10643/rcmsynta007.htm#CHDEECJA

Regards

Stefan