Database Reference
In-Depth Information
Chapter 7
Handling Online Redo Log Failures
I worked for a company that had just implemented an expensive database server with redundancy built into every
component, or so I thought. The server was configured with RAID disks for all database files and the online redo log
groups. The team was confident that there was minimal risk of failure with these disks.
Therefore, I decided not to multiplex the online redo log groups. A few days later, an inexpensive battery that
maintained the cache for a disk controller failed. This caused corruption in the current online redo log group. As a
result, the company lost data, experienced costly downtime, and had to perform an incomplete recovery.
As detailed in Chapter 2, online redo logs are crucial database files that store a record of transactions that have
occurred in your database. Since RMAN doesn't back up online redo log files, you can't use RMAN to restore these
critical files. Given their criticality, I thought it was important to include a chapter on how to deal with failures with
online redo log files.
Media failures with the online redo logs are usually noticed either when the database ceases to work
(all members of a group have experienced media failure) or you notice an error in the alert.log indicating issues,
for example:
ORA-00312: online log 3 thread 1: '/u01/oraredo/O12C/redo02b.rdo'
Once you've discovered an issue, the first step is to determine how to recover from this failure.
■
Use the RMAN
backup database plus archivelog
command to ensure your current online redo log files
(of all the threads) are switched and archived before and after the backup of the database.
Tip
Determining a Course of Action
If you've experienced a problem with your online redo log files and need to determine what shape they are in and
what action to take. Follow these steps when dealing with online redo log file failures:
1.
Inspect the
alert.log
file to determine which online redo log files have experienced a
media failure.
2.
Query
V$LOG
and
V$LOGFILE
to determine the status of the log group and degree of
multiplexing.
