Bug 16496896 - standby mrp crashed with ORA-600 [krrgv_scn8]

刘泽宇

Symptoms:
Related To:
ORA-600 [krrgv_scn8]
(None Specified)
Description
This fix prevents media recovery on the standby from failing with
ORA-00600:[krrgv_scn8] following: a crash recovery combined with
a compatible advance on the primary database.


Rediscovery Notes
Look for:
- MRP fails with ORA-00600:[krrgv_scn8]
- RAC primary/standby (ie multiple redo threads)
- LGWR SYNC redo transport mode to the standby
- one or more redo logs have a LowSCN which is higher than their
   NextSCN, ie the following query returns some rows (the same
   rows) on the primary and the standby:
         select thread#,sequence#,first_change#,next_change#
           from v$archived_log
          where first_change# > next_change#;
- for the logs (thread#/seq#) returned by the query above, the
   primary database switched into these logs during a database
   open which performed a thread recovery coinciding with a   
   "compatible" advance, ie look for BOTH the following types
   of message in the primary database alert log as the primary
   database was opened:
     Thread <t> advanced to log sequence <s1> (thread recovery)
     Thread <t> advanced to log sequence <s2> (COMPATIBLE advance)

Workaround
To prevent the problem from happening: if compatible needs to be
changed on the primary, then only do it after a clean shutdown.
If the problem has already happened, ie the media recovery is
failing with ORA-00600:[krrgv_scn8], then there is no workaround,
ie media recovery is not possible using these logs.