Oracle DB fails to shutdown.

[Chan 007]

John,

All you have to do is

1. Ensure that users of off the system

2. Listner is shutdown

3. Then shutdown the database.

if 1 & 2 fails or takes time you may get into problems.

To avoid better add the command: sleep 30 or 60 before shutting down the database.

Also can you ensure that switchlogs are done twice or 3 times before shutting down, so your db is 100% for shutdown or abort.

Chan

[Eric Antunes]

Hi again John,

Since you are using "/usr/bin/su oracle -c /home/oracle/8.1.7/bin/dbshut", how are you sure it is a immediate shutdown? Are you able to exec the following as root:

#/usr/bin/su oracle -c /home/oracle/8.1.7/bin/dbshut immediate

Or is the immediate the only mode that the dbshut script shuts down the db?

Best Regards,

Eric Antunes

[John Flanagan]

The Answer is yes to all of the above.

Yogeeraj,
I have been checking your advice on the user of rman. From my reading of the docs it does not appear go give a consistent backup of an open database. Is there an option to force it to be consistent(similar to that with the exp command).

John.

[Chan 007]

John,

As per oracle, it is not recommended to go for hotbackup.

Only Cold backups works well with 8.x

Where if you have 9 and above you can go for hotbackup.

We have plenty of inst with both combinations.

As put in my prev ans to you we use sleep while shutting the DB.

It works well.

It is upto you to decide which backup you go for.

Chan

[Bill Thorsteinson]

Hot backups are not consistent. However,
you can recover to a point in time.

My backup script writes the commaand to
recover to a point in time (now) to the
backup directory. It then waits a few seconds
and forces a log rotation.

To recover I restore the files and run
the recovery script as part of the recovery.
This requires that the uncompressed archived
redo logs covering the period of the backup
be in the redo directory.

I have automated script to roll-back our
testing databases using this techniqure.

We also clone databases by restoring thme
to a new directory structure, recreating
the control files, and renaming the
database.

[Simon Wickham_6]

Hi John,

check your alert log as well as the bdump directory. Then You will either have to shutdown abort, or have a program kill all oracle connections and then to a shutdown immediate.

1. look at recent alert log entries.
2. Once backup shutdown immediate and look at logs.

Regards,
Simon

[John Flanagan]

I have modified the standard dbshut script to use sqlplus /nolog and not svrmgrl. It worked OK last night. If it fails aganin I will stop the listener and add the wait statement.

I am also looking at using rman.

Thanks everybody,

[Yogeeraj_1]

hi John,

I still believe RMAN is a much better option.

Good that you have started looking into it.

Read more at: http://download-west.oracle.com/docs/cd/B10501_01/server.920/a96566/rcmrecov.htm#1008603


kind regards
yogeeraj

[Steven E. Protter]

Shalom John,

The dbshut that ships with oracle is flawed. We always modify the shutdown to shutdown immediate and have avoided the issues you seem to encounter.

SEP

[John Flanagan]

Thanks for your help everybody. I think we have this one licked.

I am interested in rman. I did a trial backup to disk and it completed in just 8 minutes. I am still unsure about the recovery options but I will read further into the documentation.

Regards,

John.