Backing Up Oracle Using Tivoli Storage Management

Appendix B. Troubleshooting 203

The problem is that during the Oracle backup, a session is opened with the

TSM server and the backup transaction is started. After data that exists in the

Oracle file is sent to the TSM server, the backup then goes into a loop

skipping the empty space in the database file (Null values). If there is a lot of

empty space, the TSM server can time-out and close the session before

RMAN can give the TDP client any additional data to send or finish the

transaction.

This issue is normally witnessed if the value for commtimeout on the TSM

server is set to a low value (such as the default of 60 seconds). It is

suggested to set the commtimeout value to 600-1200 seconds when working

with database backups. This value is set in the dsmserv.opt file for the TSM

server and requires the TSM server to be restarted to take effect.

B.2.2 Administrative client queries

If you are encountering errors, there are some things that you should check

on the TSM server using the administrative client and the query command.

Run the command

query node nodename f=d replacing nodename with the

nodename that is used to back up Oracle. The value for Backup Deleted

Allowed should be set to YES. The value for Maximum Mount Points Allowed

should be equal to or less than the number of drives on the TSM server. The

value for Locked? should be set to NO. The value for Policy Domain Name

should be checked to make sure it belongs to the correct domain. This value

determines what management classes are available to the node for managing

the Oracle backups.

Using the value of Policy Domain Name, you can run some queries to see if

the correct management classes for the Oracle backups are available. The

command

query mgmtclass <domain> active will show which is the DEFAULT

management class for the node. The command

query copygroup <domain>

active

will tell you what the retention settings are for the management

classes.

B.3 Isolating the problem

When troubleshooting, you must try to isolate the cause of the problem. The

four main areas in which to try to isolate the problem are: Oracle RMAN, TDP

for Oracle, TSM API, and TSM Server.

COMMTIMEOUT 600