During Veritas
Cluster Failover (Solaris8, Oracle9i, EMC) testing we encountered the following
scenario...and it has me stumped!
Database failed over
to secondary server...operates without error database up and
running
Database failed back
to primary server...ORA-214 signalled during ALTER DATABASE
MOUNT
This indicates that
we have a file that is out of sync (as if we had restored an old version and not
done a recovery).
In looking at the
database files, our 3rd control file had a time stamp from the first failover,
where all the other active files (excepting inactive redo logs) had a timestamp
from the second failover. Residing on the same disk as the out of sync control
file was the file to the users tablespace (users01.dbf) that had a timestamp of
the second failover. Once we replaced the bad controlfile with a good one, the
database could be mounted and opened.
So, if Oracle could
not read all of the control files, why did it even update the first one? If it
could not read all 3 controlfiles, why were the individual database files
touched?
My understanding of
the startup stages don't offer any explanation for this behaviour. Perhaps it is
an Oracle bug...