Re: Using Shutdown Immediate

From: Thomas J Kyte <tkyte_at_us.oracle.com>
Date: 1995/08/19
Message-ID: <413333$msp_at_inet-nntp-gw-1.us.oracle.com>#1/1


tkyte_at_us.oracle.com (Thomas J Kyte) wrote:

>mmyers_at_plano.net (Matt Myers) wrote:

...

>>In the "Oracle Backup & Recovery Handbook" by Rama Velpuri(pg 30), here's what
>>it says about shutdown normal: "The normal shutdown process stops all user
>>access to the database, waits until all users complete their requests and
>>disconnect from the server, purges data buffer and redo log caches and
>>updates data files and online redo logs, drops file locks, completes
>>ongoing transactions, updates file headers, closes thread, drops the database
>>instance lock, and synchronizes control files and data files."
 

>>Fair enough, right?
 

>>Here's what it says about shutdown immediate: "If this option is used
>>to shut down the database, the current SQL statements that are being
>>processed by Oracle are terminated immediately, any uncommitted transactions
>>are rolled back, and the database is shut down. The only disadvantage of
>>using this option is that Oracle doesn't wait for the current users to
>>disconnect. However, the database will be consistent and no recovery is
> ^^^^^^^^^^^^^^ THIS IS REALLY VERY WRONG

                                                    ^^^^^^^ OK, This is wrong

After digging around further on this here is the final scoop.

It is a documentation bug and I read my alert file too quickly.

It is a documentation bug in the DBA guide first documented in 7.0.16 and will be corrected in 7.2.

I read my alert file too quickly and didn't notice that in the example I quoted, there was a system crash during the next database startup. Two startups got in between the shutdown immediate and the crash recovery I reported. There was an actual crash between the two.

>>Hmmmmm.
 

>out (cause it just happen during immeidate shutdowns).
 

>In reviewing my own alert files i have noticed:
 

>shutdown NORMAL *never* has the message "Beginning Crash Recovery of N threads"
>in the next startup sequence.
 

>shutdown ABORT *ALWAYS* has that message in the next startup sequence
 

>after a system CRASH, we always have that message in the next startup.
 

>after a shutdown immediate, *SOMETIMES* i have "Crash Recovery", Sometimes not.
>Just as the manual states (the database MAY have to recover, it may not).

From above, the above is not true. After closer inspection, I determined that in fact the one case I saw was 'bogus'.

>>As most of us know, the Oracle manuals can leave things to be either a
>>little bit vague or slanted toward the cautious side. The book quoted
>>above draws upon the knowledge and experience of several people within
>>Oracle itself.

In this case, they decided to leave it slanted towards caution.

>In this case however the manuals (specifically the Admin Guide) is *very* clear
>(eg: it says DON"T backup these files up after shutdowm immediate, it says
>DATAFILES MAY require recovery).
 

>Last time-- Don't do a cold backup of shutdown immediate data. It needs
>recovery, it is not good enough to back it up. You will get burned someday if
>you do.

Ok, go ahead and do a backup after shutdown immediate, apparently it WON'T make a difference.

>I think the book you have needs to be fixed with respect to this point. The
>book is wrong, and gives bad advice in this case (advice that will lead to
>backups that are useless). They also need to add to the drawbacks for shutdown
>immediate that the database may need recovery and you can't back it up.

The DBA guide is being fixed. The book "Oracle Backup & Recovery Handbook" by Rama Velpuri is NOT incorrect with respect to this feature.

Apologies to all in this thread.

Better safe then sorry I guess.

The DBA HANDBOOK is CORRECT, the ADMIN GUIDE is WRONG.

Learn something new every day....

>>Matt Myers
>>Symmetry Consulting Corp.
 

>Thomas Kyte
>tkyte_at_us.oracle.com
>Oracle Government

Thomas Kyte
tkyte_at_us.oracle.com
Oracle Government Received on Sat Aug 19 1995 - 00:00:00 CEST

Original text of this message