RE: Detecting Controlfile Type
Date: Tue, 10 Feb 2009 08:08:47 -0500
Before Oracle had a product called dataguard the way we prevented accidental opening of the standby recovery database was to move the online logs aside (typically via mv).
I have not tested this whole area of preventing error based context switching of a standby under control of dataguard. At the moment I don't personally have a standby I can risk breaking.
The test would be to move all the members of all the online redolog groups to canonically determined names. (Move is better than having to create empty redologs when you need to open a standby because every second may count.) Prior to dataguard I discovered that the online redologs are not referenced at all until the open phase (at which point they may be used for instance recovery or in context of resetlogs.)
It's possible some of the backport fixes form 10.2.0.3 to 10.2.0.4+ changed the behavior. It would also be interesting to attempting reads on your database to see if "active" read only works at your point of patching or whether updates are possible. I'm not sure what would be the state of your database after such a change, even if you only may changes to a tablespace created only for the purpose of this test and which could then be dropped so that no datafiles are to far in the future and you had preserved a controlfile binary and a copy of the dummy datafile of the right vintage, so it had better be either a test or something you have no qualms about reinstantiating the standby.
I'm not aware of any documentation that precisely documents Oracle's intentions in this matter. But if the "good" expected results happen in your test (database won't open without online redologs), then the OP could use that to tell himself what the intention of the state of his controlfiles is, and act according along the lines of interperting what to do next as suggested by Mr. Wilton.
Or if you have a "Datawarehouse for the DBA" then you can keep your states there (for everything except your "Datawarehouse for the DBA"). That's the most basic type of NP incomplete problem, but let's just agree that the DBA team has to keep track of the status of the DBA datawarehouses and standbys therefore. Then you can just look up what you want automatic restarts to attempt.
Anyway, making roll your own standbys was quite a bit more fragile, and at the time moving the online redologs aside was part of my recommended procedure to manage them without accidents. Without retesting I can't claim it still works. Still, I'm not sure how Oracle even could mess up your database without processing the online redologs first. That's how I hit on that idea as a safety play.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of David Ballester
Sent: Tuesday, February 10, 2009 5:06 AM To: Stefan Knecht
Subject: Re: Detecting Controlfile Type
El mar, 10-02-2009 a las 10:54 +0100, Stefan Knecht escribió:
> Was that 11g ?
> They changed the default behaviour. An 11g database may open read only
> automatically, and turn on log-apply thereafter if you issue
> "STARTUP". Kinda naughty by Oracle me thinks, as this functionality
> requries additional licensing (active data guard option).
Shuted down database ( with the cancel managed sentence before the shutdown )
FYI, 10.2.0.4 with all recommended patches on top of it
D.Received on Tue Feb 10 2009 - 07:08:47 CST