1.ALERT日志错误
奇怪之处:Query Duration=0 sec,竟然出现了ORA-01555

Tue Feb  7 02:41:34 2012
ORA-01555 caused by SQL statement below (Query Duration=0 sec, SCN: 0x0b2e.efcd78a9):
Tue Feb  7 02:41:34 2012
SELECT "ID_NO","CUST_ID" FROM "DBACCADM"."DCUSTMSG" "C" WHERE "ID_NO"=:1

2.ORA-01555解释
超过了undo_retention时间，undo被覆盖导致ORA-01555

[zwq_acc1:/home/oraeye/check]oerr ora 1555
01555, 00000, "snapshot too old: rollback segment number %s with name \"%s\" too small"
// *Cause: rollback records needed by a reader for consistent read are
//         overwritten by other writers
// *Action: If in Automatic Undo Management mode, increase undo_retention
//          setting. Otherwise, use larger rollback segments

3.数据库版本

SQL> select * from v$version;

BANNER
----------------------------------------------------------------
Oracle9i Enterprise Edition Release 9.2.0.8.0 - 64bit Production
PL/SQL Release 9.2.0.8.0 - Production
CORE    9.2.0.8.0       Production
TNS for IBM/AIX RISC System/6000: Version 9.2.0.8.0 - Production
NLSRTL Version 9.2.0.8.0 - Production

4.undo基本信息
从这里可以发现，两个节点的undo表空间还有很多剩余，缺发生了undo被覆盖从而出现了ORA-01555

SQL> col name for a20
SQL> col value for a15
SQL> SELECT INST_ID, NAME, VALUE
  2    FROM GV$PARAMETER
  3   WHERE UPPER (Name) LIKE '%UNDO%';

   INST_ID NAME                 VALUE
---------- -------------------- ---------------
         1 undo_management      AUTO
         1 undo_tablespace      UNDOTBS1
         1 undo_suppress_errors FALSE
         1 undo_retention       1800
         2 undo_management      AUTO
         2 undo_tablespace      UNDOTBS2
         2 undo_suppress_errors FALSE
         2 undo_retention       1800

8 rows selected.


TABLESPACE_NAME                CURRENT_TOTAL(MB)   USED(MB)   FREE(MB)      FREE% AUT MAX_TOTAL(MB)
------------------------------ ----------------- ---------- ---------- ---------- --- -------------
UNDOTBS1                                   40950    1587.94 39362.0625      96.12 NO          40950
UNDOTBS2                                   57330    1926.31 55403.6875      96.64 NO          57330

SQL> SELECT DISTINCT STATUS ,
  2                  COUNT(*) "EXTENT_NUM",
  3                  SUM(BYTES) / 1024 / 1024 / 1024 "UNDO(G)"
  4    FROM DBA_UNDO_EXTENTS
  5   GROUP BY STATUS;

STATUS    EXTENT_NUM    UNDO(G)
--------- ---------- ----------
ACTIVE           208 .273658752
EXPIRED         7651 2.42865753
UNEXPIRED        941 .752548218

查询MOS[ID 761128.1]，发现可能是Oracle bug导致(BUG:6799685 – ORA-1555 ERROR WITH QUERY DURATION=0 AND UNDO_RETENTION=1800和BUG:5475085 – V$UNDOSTAT.EXPBLKREUCNT IS NEVER INCREMENTED)

5.解决方法
Increase the size of the UNDO tablespace and increase the UNDO_RETENTION parameter value to try to prevent required undo expiring too quickly.
基于本库，因为undo空间还有很大剩余，直接设置UNDO_RETENTION=3600即可(可以从一定程度上缓解整个问题，但是要从根本上解决整个问题，需要升级到10.2.0.4及其以上版本)

一.发现ORA-01555

Mon Dec 26 10:08:22 2011
ORA-01555 caused by SQL statement below (Query Duration=49146 sec, SCN: 0x0b4b.17f5ae42):
Mon Dec 26 10:08:22 2011
SELECT COMPANY_ID,
       COMPANY_MOBILE,
       TO_CHAR(NVL(REG_DATE, SYSDATE - 100), 'yyyymmddhh24miss'),
       BAK_FIELD2
  FROM TAB_XN_COMPANY
 WHERE (COMPANY_STATUS = 1 OR
       (COMPANY_STATUS = 3 AND
       NVL(UNREG_DATE, SYSDATE + 100) >=
       TO_DATE('20111226094500', 'yyyymmddhh24miss')))
   AND NVL(REG_DATE, SYSDATE - 100) <=
       TO_DATE('20111226095959', 'yyyymmddhh24miss')
   AND PAY_TYPE > 0

二.数据库状态

[oracle@ora02 ~]$ sqlplus "/ as sysdba"

SQL*Plus: Release 9.2.0.4.0 - Production on Wed Jan 4 10:48:17 2012

Copyright (c) 1982, 2002, Oracle Corporation.  All rights reserved.


Connected to:
Oracle9i Enterprise Edition Release 9.2.0.4.0 - Production
With the Partitioning, OLAP and Oracle Data Mining options
JServer Release 9.2.0.4.0 - Production

SQL> show parameter undo;

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
undo_management                      string      AUTO
undo_retention                       integer     10800
undo_suppress_errors                 boolean     FALSE
undo_tablespace                      string      UNDOTBS1

SQL> select sum(maxbytes)/1024/1024/1024,
   2 SUM(USER_BYTES)/1024/1024/1024 FROM dba_data_files where tablespace_NAME='UNDOTBS1';

SUM(MAXBYTES)/1024/1024/1024 SUM(USER_BYTES)/1024/1024/1024
---------------------------- ------------------------------
                  61.9999847                     32.6834106

SQL> SELECT DISTINCT STATUS "状态",
  2                  COUNT(*) "EXTENT数量",
  3                  SUM(BYTES) / 1024 / 1024 / 1024 "UNDO大小"
  4    FROM DBA_UNDO_EXTENTS
  5   GROUP BY STATUS;

状态      EXTENT数量   UNDO大小
--------- ---------- ----------
ACTIVE             1 .000976563
EXPIRED         2549 31.2333298
UNEXPIRED          3 .000175476

通过undo_retention保留时间为10800秒，而该sql执行了49146秒，在这49146秒钟，TAB_XN_COMPANY表中的数据被修改，而且被修改的undo数据在10800秒后被覆盖导致，导致原查询语句不能获取到scn小于或者等于查询时候的数据块内容(在undo中)，所以出现ORA-01555。从这里也可以看出来，在undo空间还剩余的情况下，如果超过了undo_retention限制，undo内容还是有可能被覆盖，而不是使用未使用的undo

三.出现ORA-1555原因
The ORA-1555 errors can happen when a query is unable to access enough undo to build a copy of the data at the time the query started. Committed “versions” of blocks are maintained along with newer uncommitted “versions” of those blocks so that queries can access data as it existed in the database at the time of the query. These are referred to as “consistent read” blocks and are maintained using Oracle undo management.
就是一个查询要访问某个数据块，而这个数据块在这个查询执行过程中修改过，那么该查询需要查询undo中数据块，而undo中该数据块已经不存在，从而出现ORA-1555

四.ORA-1555解决方法
Case 1 – Rollback Overwritten
1.缩短sql运行时间
2.增加undo_retention，这个同时需要考虑undo空间大小
3.减少commit(rollback)次数
4.在一条sql中尽量使数据块访问一次
4.1)Using a full table scan rather than an index lookup
4.2)Introducing a dummy sort so that we retrieve all the data, sort it and then sequentially visit these data blocks.

Case 2 – Rollback Transaction Slot Overwritten
这种问题，主要是延迟块清理导致，一般建议在进行大批量的dml操作后，使用全表(全index)扫描执行一遍，或者收集全部统计信息