Comments on: What Happens When You Leave an Oracle Database in Backup Mode

By: Alex Gorbachev

Alex Gorbachev — Fri, 15 Jun 2007 14:42:56 +0000

>ended up with ORA-600 during crash recovery

Chocolate ORA-600â€™s ?

It seems to be contagious!

By: Log Buffer #49: a Carnival of the Vanities for DBAs « Coskans Approach to Oracle

Fri, 15 Jun 2007 12:49:55 +0000

[...] (try it I am sure that you wont leave it) script for performance tuning. Alex Gorbachev wrote about what will happen when you leave database in backup mode with a real story. Another stylish DBA of Pythian Group Christo Kutrovsky wrote about his [...]

By: kurt

kurt — Fri, 15 Jun 2007 06:35:54 +0000

>ended up with ORA-600 during crash recovery

Chocolate ORA-600′s ?

By: Alex Gorbachev

Alex Gorbachev — Thu, 14 Jun 2007 15:48:48 +0000

Fuad,
Agree – monitoring is the way to go and we do the same with all databases that we know have a chance to be in backup mode. There are more disadvantages of having a database in backup mode than just problem with crash recovery.

By: Fuad Arshad

Fuad Arshad — Thu, 14 Jun 2007 15:31:52 +0000

Have had that a couple of time but not for such a long period.
in 9i on mount mode you do have
alter database end backup;
I have checks in my shutdown scripts that specifically check for this condition before shutting down adn starting up.
also have reports that come out daily in my daily checklists to validate the fact.
since we use emc bcv’s we have had a number of incidences where the scripts got hung due to bad establish causing us issues of all sort. better be proactive then get stuck finding tapes

By: Alex Gorbachev

Alex Gorbachev — Thu, 14 Jun 2007 15:24:36 +0000

Did you guys really try to recover a database before checking to see if it was still in backup mode?
We had all reasons to believe that RMAN had been used for backups. Actually, placing database in the backup mode wasn’t part of regular backup activity. Looks like someone was trying to perform something extraordinary at that time and left DB in backup mode. In any case, the decision was to postpone actions until the right people are available and to have some time for more investigations. And rightly so.

Making me a little nervous here people.
Heh… Easy to say when you already know the outcome and you can always say – “wasn’t that obvious to any knowledgeable DBA?”. Few times I have faced the situations when I was mad on myself after the issue was found – how did I manage to miss such an obvious detail.
The only reason to be nervous that I can see is that DBA’s were intentionally not involved into backup and recovery procedures and it seems that responsible group didn’t exercise test restore. Other than that, there was no crazy actions done in the middle of the night. Investigation was done and the root cause determined. Existing pieces of the database were preserved. The most appropriate restore strategy was followed.

I do have few concerns and suspicious why we had ORA-600 on crash recovery but I believe I should leave it aside.

By: Alex Gorbachev

Alex Gorbachev — Thu, 14 Jun 2007 14:37:02 +0000

Amit,
Regarding end backup – I should have mentioned that end backup obviously was tried but ended up with ORA-600 during crash recovery.
Wontâ€™t that result in Datafile is too new or datafile is in future kind of error instead of file needs recovery error ?
I recall that the message is confusingly the same “Datafile 1 needs media recovery” during open resetlogs.

Kurt,
Indeed, failover point you made is very good. Cheers.

By: John Hurley

John Hurley — Thu, 14 Jun 2007 14:32:50 +0000

Did you guys really try to recover a database before checking to see if it was still in backup mode?

Making me a little nervous here people.

By: amit poddar

amit poddar — Thu, 14 Jun 2007 11:52:46 +0000

“My immediate guess â€” the database is started with an old copy of the controlfile ”

Wont’t that result in
Datafile is too new or datafile is in future kind of error instead of file needs recovery error ?

By: amit poddar

amit poddar — Thu, 14 Jun 2007 11:50:05 +0000

You can also just say

alter database end backup;

if all the files were in backup mode

at that point oracle will copy the hot backup scn over to checkpoint scn.

When a datafile is put in backup mode the full header is not frozen. Only the checkpoint scn is frozen. There is a hot backup scn in the header which keeps incrementing. When end backup is issued this hot backup scn is copied over to the checkpoint scn.