ORA-00600: 内部错误代码, 参数: [4194], [34], [8], [], [], [], [], []

报错信息：

Thu Aug 27 18:49:17 2009
Errors in file /opt/app/oracle/admin/odps/bdump/odps_pmon_8453.trc:
ORA-00471: DBWR process terminated with error
Thu Aug 27 18:50:12 2009
Recovery of Online Redo Log: Thread 1 Group 3 Seq 975 Reading mem 0
Mem# 0 errs 0: /oradata/odps/redo03.log
Block recovery completed at rba 975.64.16, scn 0.59976334
Doing block recovery for file 2 block 9
Block recovery from logseq 975, block 62 to scn 59976373
Thu Aug 27 18:50:12 2009
Recovery of Online Redo Log: Thread 1 Group 3 Seq 975 Reading mem 0
Mem# 0 errs 0: /oradata/odps/redo03.log
Block recovery completed at rba 975.126.16, scn 0.59976374

===============

最近不知道是RP不好，遇到若干600错误，近乎崩溃。

这个错误是由于红旗HA两端时间不一致，导致访问共享卷出现异常，导致UNDOLOG读不一致。

首先 cd $ORACLE_HOME/dbs
vi initodps.ora

添加:*.undo_management='MANUAL'

然后

strings system01.dbf | grep _SYSSMU | cut -d $ -f 1 | sort -u > /tmp/a.txt

查询是否有交易，如果没有问题就简单的多了

offline掉UNDOTBS1，新建UNDOTBS2，并修改参数重新将UNDOLOG指认到UNDOTBS2

删除UNDOTBS1

[@more@]