一条sql语句导致的数据库宕机问题及分析（42天)
一条sql语句导致的数据库宕机问题及分析（42天)
之前分享过一篇博文，是一条sql语句"导致"的数据库宕机，上次是另有原因，这次真碰到一个案例，而且是在重要的环境上，希望大家引以为戒。
数据库是基于Linux64的版本，版本是11.2.0.2.0,已经打了最新的psu.
数据库的访问用户数大约在1000左右，当时查看服务器的cpu已经是100%了，有大约10个进程都是cpu 100%,数据库逻辑读也是超高，一秒钟大约是接近百兆的情况，sga是12G,已用了sga的自动管理(sga_target=0), 查看内存组件时发现buffer_cache已经有shrink的迹象，而且buffer_cache的min_size还是有一点小，就在可用范围内给buffer cache 增大了几百兆的样子，生成了一个ADDM, 报告里第一条就是希望设置sga_target为一个特定的值，性能可能会有一定的提升，当时想,sga_max_size都已经是12G了，设置sga_target=12G也没有问题吧
就按照它的提示做了，
alter system set sga_target=12G;
结果命令提顿了几秒钟，然后就崩出来一个end_of_communicaiton的ora错误，我感觉出问题了，已查看进程，数据库是真down掉了。
查看alert日志，发现时由于resize_sga的ora-600问题导致的，所有的在线进程都被自动给kill掉了。
然后马上和相应的team来协调，把数据库先startup了。再查看具体的信息。
alert日志如下：
Thread 1 advanced to log sequence 14054 (LGWR switch)
Current log# 2 seq# 14054 mem# 0: /dbtestPR1/oracle/TEST01/redolog_A2/redo/redo02A.log
Current log# 2 seq# 14054 mem# 1: /dbtestPR1/oracle/TEST01/redolog_B2/redo/redo02B.log
Wed Apr 09 20:07:10 2014
Archived Log entry 14090 added for thread 1 sequence 14053 ID 0xb8c6d509 dest 1:
Wed Apr 09 20:40:13 2014
Errors in file /opt/app/oracle/dbtestpr1/diag/rdbms/TEST01/TEST01/trace/TEST01_mman_27182.trc (incident=360075):
ORA-00600: internal error code, arguments: [kmgsb_resize_sga_target_1], [0], [768], [4], [], [], [], [], [], [], [], []
Incident details in: /opt/app/oracle/dbtestpr1/diag/rdbms/TEST01/TEST01/incident/incdir_360075/TEST01_mman_27182_i360075.trc
Use ADRCI or Support Workbench to package the incident.
See Note 411.1 at My Oracle Support for error and packaging details.
Errors in file /opt/app/oracle/dbtestpr1/diag/rdbms/TEST01/TEST01/trace/TEST01_mman_27182.trc:
ORA-00600: internal error code, arguments: [kmgsb_resize_sga_target_1], [0], [768], [4], [], [], [], [], [], [], [], []
MMAN (ospid: 27182): terminating the instance due to error 822
Wed Apr 09 20:40:14 2014
opiodr aborting process unknown ospid (25518) as a result of ORA-1092
Wed Apr 09 20:40:14 2014
ORA-1092 : opitsk aborting process
Wed Apr 09 20:40:14 2014
Wed Apr 09 20:40:14 2014
opiodr aborting process unknown ospid (27776) as a result of ORA-1092opiodr aborting process unknown ospid (10547) as a result of ORA-1092
Wed Apr 09 20:40:14 2014
opiodr aborting process unknown ospid (7458) as a result of ORA-1092
Wed Apr 09 20:40:14 2014
Wed Apr 09 20:40:14 2014
ORA-1092 : opitsk aborting process
ORA-1092 : opitsk aborting process
Wed Apr 09 20:40:14 2014
ORA-1092 : opitsk aborting process
Wed Apr 09 20:40:14 2014
opiodr aborting process unknown ospid (30719) as a result of ORA-1092
Wed Apr 09 20:40:14 2014
ORA-1092 : opitsk aborting process
.......
ORA-1092 : opitsk aborting process
Wed Apr 09 20:40:14 2014
System state dump requested by (instance=1, osid=27182 (MMAN)), summary=[abnormal instance termination].
System State dumped to trace file /opt/app/oracle/dbtestpr1/diag/rdbms/TEST01/TEST01/trace/TEST01_diag_27176.trc
Instance terminated by MMAN, pid = 27182
查看metalink，确实有这样的一个bug.
Bug 10173135 - Resize SGA_TARGET crashes instance with ORA-600 [kmgsb_resize_sga_target_1] (Doc ID 10173135.8)
而且影响的版本如下：
Product (Component) | Oracle Server (Rdbms) |
|---|---|
Range of versions believed to be affected | Versions BELOW 12.1 |
Versions confirmed as being affected | 11.2.0.211.2.0.111.1.0.7 |
Platforms affected | Generic (all / most platforms affected) |
- 11.2.0.2
- 11.2.0.1
- 11.1.0.7
Platforms affectedGeneric (all / most platforms affected)
真是撞到枪口上了，查看了下，在11.2.0.3.0中才修复了这个问题。 然后自我总结了下，发现sga的自动管理操作还是需要谨慎，新特性的使用也是如此，一定要有足够的把握才能使用。