TDP for Oracle using one channel, not more

[andrescontilde]

Hi all.

I am running an Oracle restore of a database of about 5 TB. My RMAN script is allocating 3 channels for this task. In the generated log I can see that the 3 channels (t1, t2, t3) are allocated but when the actual restore starts it is using just one channel (t1).

Can someone explain why this is happening?. The restore is working "fine", although it's taking a lot of time.

Thanks in advance.

Regards

Andrés

 

[moon-buddy]

Restoring from tapes? Data may have been backed up on one tape and thus can only restore from one tape!

 

[andrescontilde]

Yes, it is a restore from a VTL. I was thinking the same way but looking at the actlog and the tapes mounted it is using more than one tape in the restore process... I mean one at a time as there is just one channel processing the restore. I have 3 tapes mounted but just one drive is in use and the other two are idle. I don't get it.

 

[domino]

which version of oracle is installed?

 

[Trident]

Hi,

Check this

http://publib.boulder.ibm.com/infoc....jsp?topic=/com.ibm.itsmfd.doc/ab5w000345.htm

You need to define three more mg classes to be able to use four streams of data.

 

[andrescontilde]

@domino

The Oracle version is 10g (10.2.0.4).

@Trident

Thanks for the link but I think it's not what I need. I don't need duplex copies of the backup, I need to "force" the utilization of more than one channel in the restore process as the information is located in different tapes. The restore is using only one channel and mounting different tapes in my 3 free drives but reading them in sequence.

 

[domino]

http://www-01.ibm.com/support/docview.wss?uid=swg1IC48436

 

[Trident]

Hi,
Please post
- your tdpo.opt file (general config). Looking for mg_class settings
- q node TDPO_NODE f=d (looking for maxnummp settings). This needs to be higher than 4
- d devcl for relevant devices (Looking for mount limit). This needs to be drives or higher than 4

 

[andrescontilde]

Thank you very much for your help.

What we did to improve the performance was migrate some backupsets to another storage pool located at another library. We restarted the restore and eventually it was using two channels, this way the restore concluded in less time. It took a lot of time but I least the DB could be restored.

Thanks to domino for the link it would be a very good reason to upgrade our TSM infrastucture.

Regards

Andrés