RE: Controlfile Recovery Requires RESETLOGS

This demonstrates the risk of over-snipping. I was responding to a fragment with no reference to replacing the current controlfile and it appeared that the poster was saying that the result of a shutdown immediate was a recovery requiring open reset logs (ergo my questions 4 and 5, and I'd say the answer to 4 is Yep, I got rid of the current controlfile.  

That prevents automatic instance recovery, and the rest of what you say in context makes total sense to me.  

Regards,  

mwf  

From: jeremypaulschneider_at_gmail.com [mailto:jeremypaulschneider_at_gmail.com] On Behalf Of Jeremy Paul Schneider
Sent: Wednesday, May 09, 2007 3:49 PM
To: oracle-l; Hemant K Chitale; Joel.Patterson_at_crowley.com; Mark W. Farnham Subject: Controlfile Recovery Requires RESETLOGS  

Hey all...  

In response to a bit of discussion last week - I put together the test and output showing how controlfile recovery requires recovery and a RESETLOGS even if the database was closed normally, in a consistent state. I posted the output here:  

http://www.ardentperf.com/2007/05/09/controlfile-recovery-requires-resetlogs /  

Most questions can be answered by looking at the post since I walked through the whole process. But to answer a few specific questions that various people asked...  

On 5/4/07, Mark W. Farnham <mwf_at_rsiz.com> wrote:

1. Did you shut down just the database or the whole machine?

In this test I only shut down the database however it was completely shutdown and I would anticipate the same results with a machine reboot.  

2) Did you let Oracle complete the shutdown immediate, or did you kneecap any processes, especially including LGWR (log writer, whatever name it goes by on your release)?

Yes Oracle completed the shutdown immediate - it was a clean shutdown. I didn't kneecap anything.

3) Is any sort of write cache enabled on the devices supplying the persistent media for your redo logs or controlfiles?

Shouldn't be relevant for this test. The OS and storage subsystem were never interupted; I simply shutdown Oracle (normal) then removed the control files and restarted it.  

4) Did you replace the current controlfile with anything else?

Nope. Just deleted it.  

5) Did you use the recover command with . using backup controlfile?

Yes - oracle required the "using backup controlfile" clause.    

On 5/4/07, Joel.Patterson_at_crowley.com <Joel.Patterson_at_crowley.com <mailto:Joel.Patterson_at_crowley.com> > wrote:

Even if you switched current, or all before shutdown?

In the test I posted online I tried with a log switch before shutdown. I would anticipate that this has no impact either way on the RESETLOGS.      

On 5/4/07, Hemant K Chitale <hkchital_at_singnet.com.sg> wrote:

Question : In your recovery testing, what was the exact sequence ? Was it ?

1. Restore an older controlfile binary backup -- only controlfile restored, datafiles being current
2. Issue RECOVER DATABASE USING BACKUP CONTROLFILE -- did you also specify "UNTIL CANCEL" ?

OR

1. Create Controlfile
2. Issue RECOVER DATABASE ... commands.

The first - I restored a binary backup.  

-- 
Jeremy Schneider
Chicago, IL
http://www.ardentperf.com/category/technical 




--
http://www.freelists.org/webpage/oracle-l