Oracle undo表空间爆满的解决

1. 启动SQLPLUS，并用sys登陆到数据库。
#su - oracle  
$>sqlplus / as sysdba 

2. 查找数据库的UNDO表空间名,确定当前例程正在使用的UNDO表空间：
Show parameter undo_tablespace。

3. 确认UNDO表空间；
SQL> select name from v$tablespace;  
NAME  
------------------------------  
UNDOTBS1 

4. 检查数据库UNDO表空间占用空间情况以及数据文件存放位置；
select file_name, bytes / 1024 / 1024 / 1024
  from dba_data_files
 where tablespace_name like 'UNDOTBS%';

5. 查看回滚段的使用情况，哪个用户正在使用回滚段的资源，如果有用户最好更换时间（特别是生产环境）。
select s.username, u.name
  from v$transaction t, v$rollstat r, v$rollname u, v$session s
 where s.taddr = t.addr
   and t.xidusn = r.usn
   and r.usn = u.usn
 order by s.username;

查询结果为空的话就能删除。

6. 检查UNDO Segment状态；
select usn,
       xacts,
       rssize / 1024 / 1024 / 1024,
       hwmsize / 1024 / 1024 / 1024,
       shrinks
  from v$rollstat
 order by rssize;

USN  XACTS  RSSIZE/1024/1024/1024  HWMSIZE/1024/1024/1024  SHRINKS
8	0	0.00017547607421875	3.11521148681641	700
3	0	0.00023651123046875	3.22954559326172	632
0	0	0.00035858154296875	0.00035858154296875	0
1	0	0.00206756591796875	3.04867553710938	920
10	0	0.00206756591796875	0.648170471191406	819
7	0	0.00231170654296875	3.94835662841797	730
4	0	0.00304412841796875	2.00011444091797	651
11	0	0.00695037841796875	2.26921844482422	740
9	0	0.00792694091796875	2.07530975341797	773
6	0	0.00792694091796875	1.31906890869141	775
2	0	0.00890350341796875	3.13677215576172	699
5	0	1.96833801269531	3.99906921386719	267

这还原表空间中还存在12个回滚的对象。

7. 创建新的UNDO表空间，并设置自动扩展参数；
create undo tablespace undotbs2 datafile '/u02/pnrdb/undotbs2_01.dbf' size 100m reuse autoextend on next 100m maxsize unlimited;

8. 切换UNDO表空间为新的UNDO表空间 ， 动态更改spfile配置文件；

alter system set undo_tablespace=undotbs2 scope=both;  

9.验证当前数据库的 UNDO表空间
SQL> show parameter undo
NAME                                 TYPE        VALUE
------------------------------------ ----------- --------------
undo_management                      string      AUTO
undo_retention                       integer     900
undo_tablespace                      string      UNDOTBS2
9. 等待原UNDO表空间所有UNDO SEGMENT OFFLINE；
select usn,
       xacts,
       status,
       rssize / 1024 / 1024,
       hwmsize / 1024 / 1024,
       shrinks
  from v$rollstat
 order by rssize;

select usn,
       xacts,
       status,
       rssize / 1024 / 1024,
       hwmsize / 1024 / 1024,
       shrinks
  from v$rollstat
 order by rssize;
select t.segment_name, t.tablespace_name, t.segment_id, t.status
  from dba_rollback_segs t;

SEGMENT_NAME      TABLESPACE_NAME SEGMENT_ID   STATUS
_SYSSMU1$	UNDOTBS1	1	OFFLINE
_SYSSMU2$	UNDOTBS1	2	OFFLINE
_SYSSMU3$	UNDOTBS1	3	OFFLINE
_SYSSMU4$	UNDOTBS1	4	OFFLINE
_SYSSMU5$	UNDOTBS1	5	OFFLINE
_SYSSMU6$	UNDOTBS1	6	OFFLINE
_SYSSMU7$	UNDOTBS1	7	OFFLINE
_SYSSMU8$	UNDOTBS1	8	OFFLINE
_SYSSMU9$	UNDOTBS1	9	OFFLINE
_SYSSMU10$	UNDOTBS1	10	OFFLINE
_SYSSMU11$	UNDOTBS1	11	OFFLINE
_SYSSMU12$	UNDOTBS1	12	OFFLINE
_SYSSMU13$	UNDOTBS1	13	OFFLINE
_SYSSMU14$	UNDOTBS1	14	OFFLINE
_SYSSMU15$	UNDOTBS1	15	OFFLINE
_SYSSMU16$	UNDOTBS1	16	OFFLINE
_SYSSMU17$	UNDOTBS1	17	OFFLINE
_SYSSMU18$	UNDOTBS1	18	OFFLINE
_SYSSMU19$	UNDOTBS1	19	OFFLINE
_SYSSMU20$	UNDOTBS1	20	OFFLINE
_SYSSMU21$	UNDOTBS1	21	OFFLINE
_SYSSMU22$	UNDOTBS1	22	OFFLINE
_SYSSMU23$	UNDOTBS1	23	OFFLINE
_SYSSMU24$	UNDOTBS1	24	OFFLINE
_SYSSMU25$	UNDOTBS1	25	OFFLINE
_SYSSMU26$	UNDOTBS1	26	OFFLINE
_SYSSMU27$	UNDOTBS1	27	OFFLINE
_SYSSMU28$	UNDOTBS1	28	OFFLINE
_SYSSMU29$	UNDOTBS1	29	OFFLINE
_SYSSMU30$	UNDOTBS1	30	OFFLINE
_SYSSMU31$	UNDOTBS1	31	OFFLINE
_SYSSMU32$	UNDOTBS1	32	OFFLINE
_SYSSMU33$	UNDOTBS1	33	OFFLINE
_SYSSMU34$	UNDOTBS1	34	OFFLINE
_SYSSMU35$	UNDOTBS1	35	OFFLINE

上面对应的UNDOTBS1还原表空间所对应的回滚段均为OFFLINE
 
10.到$ORACLE_HOME/dbs/init$ORACLE_SID.ora如下内容是否发生变更：
#cat $ORACLE_HOME/dbs/initddptest.ora
……
*.undo_management=’AUTO’
*.undo_retention=10800
*.undo_tablespace=’UNDOTBS2’
……

如果没有发生变更请执行如下语句：
SQL> create pfile from spfile;
File created.

11. 删除原有的UNDO表空间；

drop tablespace undotbs1 including contents and datafiles;

12. os级别释放undo数据文件；

到root下执行
lsof |grep /u02/pnrdb/undotbs01.dbf

lsof |grep /u02/pnrdb/undotbs01.dbf |awk '{printf"kill -9 "$2"\n"}'