RE: ORA-0214, controlfile updates question

ah yes, once again I fired off to oracle-l when I meant to
do oak table...sorry for the noise.


________________________________

        From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Kevin Closson
        Sent: Wednesday, August 30, 2006 2:15 PM
        To: ORACLE-L
        Subject: RE: ORA-0214, controlfile updates question
        
        
         
        so, admins create multiplexed CF because one is good, so more
must
        be better! :-)
         
         
         
         


________________________________

                From: Tuomas Pystynen
[mailto:tuomas.pystynen@xxxxxxxxxxxx] 
                Sent: Wednesday, August 30, 2006 1:34 PM
                To: Kevin Closson
                Cc: OakTable List
                Subject: Re: ORA-0214, controlfile updates question
                
                
                I agree. After a crash there is no recovery before
controlfiles are opened (and ORA-214 is signalled). Oracle can not write
atomically to multiple controlfiles.
                
                Some important controlfile updates (such as add
datafile) are logged to redo. But it doesn't help here. The redo can not
be applied before controlfiles are opened (redo log names are there). I
believe redo is applied to controlfiles only during media recovery when
using backup controlfile.
                
                Tuomas
                
                Kevin Closson wrote: 

                         >>>Due bugs that show them in extreme
circumstances - why not?
                          

                                (I have to admit though that I haven't
seen any corruption 
                                cases yet where the cause would have
been determined as lost 
                                writes and nothing else).
                                        

                        
                        I'm not saying it it impossible, I'm saying that
in this
                        case it is MUCH more likely that Oracle's
"transaction"
                        areound this type of multiplexing is a bit
questionable.

Other related posts: