ReCreate a dg when primary is crashed.
At primary(where db is crashed):
drop database
prerequisite:
1, mount database in exclusive mode (cluster_database=false)
2, enable restricted session.
After drop database in RAC, ONLY archived log and DB cluster service resource is left.
Controlfiles, datafiles, and even spfile in ASMDG is removed.
At standby :
alter database activate standby database ;
alter database open ;
At old primary:
rman target sys/oracle@men_dg auxiliary sys/oracle@men
duplicate target database for standby from active database;
.......
Here I find that some file Location is not as expected, So I remove the files , and the trouble is from here.
1, alter database datafile 13 offline drop ; #if you do not do this first, the file is locked and could't be removed.
2, ASMCMD> rm <datafile 13>
Why I get the unexpected file Location ?
The answer is that I used the parameter "db_file_name_convert" , and match file path name multi-times and only the first match is in effect. So the Right order of file path covert is Let the Most-detailed pathname first. Like this:
alter system set db_file_name_convert= '/oracle/oradata/men/MEN_DG/datafile', '+datadg/men/datafile', '/oracle/oradata','+datadg' scope=spfile;
The following statement can be used to create the missing data file . But when I recover it , I get the trouble again.
alter database create datafile 'OLD_NMAE' as 'NEW_NAME' ;
alter database create datafile 16,17,18 as '+datadg','+datadg', '+datadg' ;
I find that the argument of VALID_FOR is taken current database role as consider, not the remote db.
ALTER SYSTEM SET log_archive_dest_3 = "SERVICE=men lgwr async VALID_FOR=(ONLINE_LOGFILES, PRIMARY_ROLE) DB_UNIQUE_NAME=men" ;
The correct steps to recove missing datafile on standby database:
1, connect to both primary and standby db by using rman.
rman target sys/oracle@men_dg auxiliary sys/oracle@men
2, backup datafile (which missing on standby) as copy to standby host.
backup as copy datafile 13 auxiliary format '+datadg' ;
backup as copy datafile 14 auxiliary format '+datadg' ;
backup as copy datafile 16 auxiliary format '+datadg' ;
backup as copy datafile 17 auxiliary format '+datadg' ;
backup as copy datafile 18 auxiliary format '+datadg' ;
3, register datafile copy on standby host:
catalog clone datafilecopy
'+DATADG/men/datafile/users.324.1091503783',
'+DATADG/men/datafile/ts_test1.327.1091503787',
'+DATADG/men/datafile/undo1.328.1091503823',
'+DATADG/men/datafile/undo2.330.1091503831',
'+DATADG/men/datafile/users.323.1091503087' ;
4, Because some error in command "switch clone", use rman connect to standby standalone
At Standby host:
rman target /
switch datafile 13,14,16,17,18 to copy;
5, online datafiles .Knowing that only in mount mode could you online datafiles.
shutdown abort
startup mount
alter database datafile 13,14,16,17,18 online ;
SELECT status, FILE#, 'rm '||NAME name, checkpoint_change# FROM V$DATAFILE_header ; #status is online
SELECT status, FILE#, 'rm '||NAME name, checkpoint_change# FROM V$DATAFILE ; #status is recover
at this time, the status info of datafile head and controlfile is not synchronous until db is open.
alter database open ;
recover managed standby database using current logfile disconnect ;
#Discovery knowledge point:
In view v$dataguard_stats , "apply lag" is not null when using real time recover. OR it is null when archived log is used to recover.
apply lag +00 00:00:00
Now successfully recover all missing datafiles ! have fun !
