06-06-2002 08:14 PM
the backup method is
do host copy of the datafile.
do this one by one , tablespace.
1. Do we need to backup Roll back segment tablespace and TEMP tablespace.
2. We encouter database performace in past 2 days , what can be the problem. No relevant error message in alert log.
06-06-2002 08:46 PM
1. Treat both tablespaces as if they are ordinary tablespaces. Strictly speaking you do not need the TEMP tablespace. However, if you don't take the backup you'll have to drop/recreate the tablespace at recovery time.
We use an script that first makes a list of all tablespaces and then backups them one by one. Oh, btw, do NOT forget to backup your controlfiles as well !
2. You might want to - if you haven't already - set up measureware groups to monitor the DBapplications in glance. It's very difficult to pinpoint what is wrong exactly if you don't know what is the "normal" view of your environment. Maybe the following document can help : http://h21007.www2.hp.com/dspp/files/unprotected/d
06-06-2002 08:53 PM
Thanks but this database running in NT.
Does anybody used to face this similar kind of problem.
This same script running fine before.
06-06-2002 09:31 PM
Yes, you need to backup ALL tablespaces ( the files, which belong to the tablespaces ).
The problem will occur, if you try to restore the database and the timestamps in the controlfile are not consistent. That is, why we try to do cold backups whenever possible.
If you do hot backups, you really need these tablespaces, because they can contain changed blocks, that might be needed for rollbacks or commits. These things are extremely important, if you want to do a recover until time.
Alexander M. Ermes
06-06-2002 11:27 PM
The performance issue could be a result of almost anything so you have to get inplace som monitoring of your server and try to pinpoint the bottleneck.
Try "Spotlight on Oracle" from Quest. Its a neat tool which displays the database and all processes as well as SGA and data transfer.
Check if you have exhausted our physical RAM and have began swapping, and check if some disks are taking the lot of the work.
Futher check the Oracle cache hits and Liberey chache hits. If these are low then you might have to enlarge your SGA (if you have memory for it).
06-07-2002 12:02 AM
The recomendation from our Oracle DBA is to take a backup of all of the tablespace defined, included the RBS, system and TEMP.
We do it all the nights as a hot backup and one day in a week as a cold backup, in order to use a hot backup you need the archivelogs also and a copy of the controlfile. We do it with OBII and we use the pre-exec and post-exec in order to shutdown/startup the instances and to put the tablespace in backup mode.
About performance, the spootlight software is a good tool in order to investigate the problems, probably you can obtain a trial license.
There are so many posibilities of problems like:
You are using Dynamic SQL in your data loading proceses.
You can have a lot of fragmentation in your tablespace.
You can have the most offten used tables and index in the same fisical disks.
You can not have enought SGA defined.
Hope this helps,
06-07-2002 12:10 AM
This is the www page where you can download the Trial Spootlight for oracle from Quest Software Company:
When you download and installed the Spootlight you must setup your Oracle the Oracle variable TIMED_STATISTICS=true.
06-07-2002 12:54 AM
yes, and after my special recovery thread described here,
I have to say, that it is even worth to schedule additional backups of SYSTEM and Rollbacktablespaces.
1) If you can only restore without index data, you can recreate it
2) If you can only restore with loss of some table data, you might loose data partially, but will be able to get at least a certain percentage
3) ... but without SYSTEM or a datafile containing an ACTIVE rollback segment, you will not be able to open the database at all, or might have the risk to need to go for data corruption!
Hope this helps