Non-archive ORA-00333 causes the database to fail to open

A customer contacted me before today and said that a database could not be started. It took several minutes to remotely read the database. The alert log information is as follows:

Sun Mar 20:27:49 2014Media Recovery StartParallel recovery started with 7 processesSun Mar 20:27:49 2014Recovery of Online Redo Log: Thread 1 Group 3 Seq 7363 Reading mem 0 Mem #0 errs 0:/oracle/data/orclzjs/redo03.logSun Mar 20:28:01 2014Errors in file/opt/app/oracle/admin/orclzjs/udump/orclzjs_ora_21469.trc:ORA-00333: redo log read block 67221 count 6144 ErrorORA-00312: Online log 3 thread 1: '/oracle/data/orclzjs/redo03.log'ORA-27072: file I/O errorsLinux-x86_64 Error: 2: No such file or directoryAdditional information: 4Additional information: 67221Additional information: 669184Sun Mar 20:28:08 2014Errors in file/opt/app/oracle/admin/orclzjs/udump/orclzjs_ora_21469.trc:ORA-00333: redo log read block 65173 count 8192 ErrorORA-00312: Online log 3 thread 1: '/oracle/data/orclzjs/redo03.log'ORA-27091: Unable to queue I/OORA-27072: file I/O errorsLinux-x86_64 Error: 2: No such file or directoryAdditional information: 4Additional information: 67221Additional information: 669184Sun Mar 20:28:19 2014Media Recovery failed with error 333ORA-283 signalled during: alter database recover database...Sun Mar 20:30:16 2014Alter database recover database until cancelSun Mar 20:30:16 2014

This is a typical case of Oracle currnet redo logfile corruption. The database was forcibly shut down. After the host was restarted, the database could not be started normally.

For non-archive databases, before stopping the database, we recommend that you run alter system checkpoint first, switch the redo several times, kill the related process, and then shutdown. Even if it is shutdown abort, the risk is much smaller, at least forcing the open Database to avoid data loss.

Of course, the customer must lose data here. It is easy to open the database through implicit parameters.

Note: in fact, we can try to fix redo, if the two redo log blocks mentioned in alert log are not seriously damaged.