一次enq: CF - contention 导致数据库宕机的故障分析

最新推荐文章于 2024-04-10 15:13:51 发布

原创

最新推荐文章于 2024-04-10 15:13:51 发布 · 2.1k 阅读

0 ·

CC 4.0 BY-SA版权

本文分析了一次由于enq: CF - contention 导致的数据库宕机故障，指出Oracle在等待CF锁超时时会自动杀死持有该锁的进程，该行为由参数&_kill_controlfile_enqueue_blocker控制。故障发生前，redo log切换频繁，可能原因包括慢速I/O子系统、频繁的日志切换等。建议调整redo logfile大小和I/O性能，以避免类似问题。

数据库意外宕机，alert日志信息如下：

Sun Jul 27 01:02:48 2014
System State dumped to trace file /oracle/app/oracle/product/1020/admin/jcgl/bdump/jcgl2_diag_569650.trc
Sun Jul 27 01:03:48 2014
Killing enqueue blocker (pid=721256) on resource CF-00000000-00000000
by killing session 3289.1
Killing enqueue blocker (pid=721256) on resource CF-00000000-00000000
by terminating the process
LMD0: terminating instance due to error 2103
Sun Jul 27 01:03:48 2014
System state dump is made for local instance
Sun Jul 27 01:03:48 2014
Errors in file /oracle/app/oracle/product/1020/admin/jcgl/bdump/jcgl2_lms5_1048666.trc:
ORA-02103: PCC: inconsistent cursor cache (out-of-range cuc ref)
Sun Jul 27 01:03:48 2014
Trace dumping is performing id=[cdmp_20140727010348]
Sun Jul 27 01:03:49 2014
Shutting down instance (abort)
License high water mark = 1575
Sun Jul 27 01:03:57 2014
Termination issued to instance processes. Wa