一條sql語句導致的資料庫當機問題及分析

531968912發表於2015-12-24
SQL資料庫
之前分享過一篇博文，是一條sql語句"導致"的資料庫當機，上次是另有原因，這次真碰到一個案例，而且是在重要的環境上，希望大家引以為戒。
資料庫是基於Linux64的版本，版本是11.2.0.2.0,已經打了最新的psu.
資料庫的訪問使用者數大約在1000左右，當時檢視伺服器的cpu已經是100%了，有大約10個程式都是cpu 100%,資料庫邏輯讀也是超高，一秒鐘大約是接近百兆的情況，sga是12G,已用了sga的自動管理(sga_target=0), 檢視記憶體元件時發現buffer_cache已經有shrink的跡象，而且buffer_cache的min_size還是有一點小，就在可用範圍內給buffer cache 增大了幾百兆的樣子，生成了一個ADDM, 報告裡第一條就是希望設定sga_target為一個特定的值，效能可能會有一定的提升，當時想,sga_max_size都已經是12G了，設定sga_target=12G也沒有問題吧
就按照它的提示做了，
alter system set sga_target=12G;
結果命令提頓了幾秒鐘，然後就崩出來一個end_of_communicaiton的ora錯誤，我感覺出問題了，已檢視程式，資料庫是真down掉了。
檢視alert日誌，發現時由於resize_sga的ora-600問題導致的，所有的線上程式都被自動給kill掉了。

然後馬上和相應的team來協調，把資料庫先startup了。再檢視具體的資訊。
alert日誌如下：
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

檢視metalink，確實有這樣的一個bug.
Bug 10173135 - Resize SGA_TARGET crashes instance with ORA-600 [kmgsb_resize_sga_target_1] (Doc ID 10173135.8)

而且影響的版本如下：
Product (Component) Oracle Server (Rdbms)
Range of versions believed to be affected Versions BELOW 12.1
Versions confirmed as being affected
Platforms affected Generic (all / most platforms affected)

真是撞到槍口上了，檢視了下，在11.2.0.3.0中才修復了這個問題。
然後自我總結了下，發現sga的自動管理操作還是需要謹慎，新特性的使用也是如此，一定要有足夠的把握才能使用。

來自 “ ITPUB部落格 ” ，連結：http://blog.itpub.net/25462274/viewspace-1958816/，如需轉載，請註明出處，否則將追究法律責任。

相關文章