I can see a reason if you get down to only one control file currently surviving
and writeable with confirmation.
For example, someone could add a datafile, do something important, lose the
last controlfile, and then hilarity ensues.
A reasonable alternative <doesn’t exist, potential enhancement warning> would
be to have a list of alternative locations, so instead of having a lot of
controlfiles to constantly write, if you lose one, try to start writing one to
places on the list of alternatives and if successful continue.
The current behavior dates back to the start of V6.0 and I doubt anyone has
really given it a tough look since then. Seems to me amendable to a voting
algorithm. But when V6 was being built, folks didn’t often even have simple
duplex mirroring of drives, so it probably seemed like a bad idea to continue
at all after the first write failure to a controlfile.
But it has to do with recovery. When in doubt on that, ask Bill. I doubt he’s
lurking here. Some lurker from Oracle who knows who I am talking about by just
the tokens “recovery” and “Bill” might want to pass the question on.
mwf
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx] On ;
Behalf Of Robert Freeman
Sent: Tuesday, November 15, 2016 3:08 PM
To: Oracle-L@xxxxxxxxxxxxx
Subject: RE: Logic of database crash should you lOse a control file...?
Note to self – LOSE not LOOSE!!
LOL
RF
Robert G. Freeman
Deliverer of Data
Businessolver
Cell: 801-703-3405
“Personally, I think that’s a hell of a bird.”
-- Dr. Who (Peter Capaldi)