今天接到运维人员反馈最近表空间增长有点过快。按现有的数据增长，预计年底前就需要扩容存储。而从DBA的角度盲目进行硬件扩容是一种不负责任的的表现，我们要搞清究竟是什么导致数据增长的，解决问题根本所在。下面把整个过程分享给各位小伙伴，希望能帮助到您。
问题现象
从运维人员那得知表空间加了没有多久就用完了，客户方原本打算做预算将存储进行扩容。但本人觉得还是要对客户方负责，评估一下到底需不要增加硬件。先看一下表空间使用率，用户数据主要在USERS表空间，大概3700GB。
[root@rac01 ~]# su - oracle
[oracle@rac01 ~]$ sqlplus / as sysdba
SQL*Plus: Release 11.2.0.4.0 Production on Tue Jun 18 14:46:45 2024
Copyright (c) 1982, 2013, Oracle. All rights reserved.
Connected to:
Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
SQL> set linesize 200
col tablespace_name for a20
select df.tablespace_name,
count(*) files,
round(sum(df.bytes) / 1048576) size_mb,
round(sum(free.bytes) / 1048576, 2) free_mb,
100 - round(100.0 * sum(free.bytes) / sum(df.bytes), 2) pct_used,
round(100.0 * sum(free.bytes) / sum(df.bytes), 2) pct_free
from dba_data_files df,
(select tablespace_name,
file_id,
sum(bytes) bytes,
max(bytes) maxbytes
from dba_free_space
group by tablespace_name, file_id) free
where df.tablespace_name = free.tablespace_name(+)
and df.file_id = free.file_id(+)
group by df.tablespace_name
order by pct_used desc
/SQL> SQL> 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
TABLESPACE_NAME FILES SIZE_MB FREE_MB PCT_USED PCT_FREE
-------------------- ---------- ---------- ---------- ---------- ----------
SYSTEM 2 11850 137.63 98.84 1.16
USERS 118 3854888 73229 98.1 1.9
SYSAUX 2 44374 1593.38 96.41 3.59
UNDOTBS1 1 30355 9554.31 68.52 31.48
UNDOTBS2 1 28319 25664.06 9.38 90.62
SQL>
问题分析
通过dba_hist_tbspc_space_usage查询USERS表空间近期增量情况，发现每天平均在2G的数据增量。
SQL> set line 200
col ts_name for a30
col pct_used for a10
SELECT a.snap_id,
c.tablespace_name ts_name,
to_char(to_date(a.rtime, 'mm/dd/yyyy hh24:mi:ss'), 'yyyy-mm-dd hh24:mi') rtime,
round(a.tablespace_size * c.block_size / 1024 / 1024, 2) ts_size_mb,
round(a.tablespace_usedsize * c.block_size / 1024 / 1024, 2) ts_used_mb,
round((a.tablespace_size - a.tablespace_usedsize) * c.block_size / 1024 / 1024,2) ts_free_mb,
round(a.tablespace_usedsize * c.block_size / 1024 / 1024, 2) -
lag(round(a.tablespace_usedsize * c.block_size / 1024 / 1024, 2),1)
over(order by a.tablespace_id,to_date(a.rtime, 'mm/dd/yyyy hh24:mi:ss')) inc_mb,
round(a.tablespace_usedsize / a.tablespace_size * 100, 2) || '%' pct_used
FROM dba_hist_tbspc_space_usage a,
(SELECT tablespace_id,
substr(rtime, 1, 10) rtime,
max(snap_id) snap_id
FROM dba_hist_tbspc_space_usage nb
group by tablespace_id, substr(rtime, 1, 10)) b,
dba_tablespaces c,
SQL> v$tablespace d
where a.snap_id = b.snap_id
and a.tablespace_id = b.tablespace_id
and a.tablespace_id=d.TS#
and d.NAME=c.tablespace_name
and d.NAME = '&tbs_name'
and to_date(a.rtime, 'mm/dd/yyyy hh24:mi:ss') >=sysdate-30
order by a.tablespace_id,to_date(a.rtime, 'mm/dd/yyyy hh24:mi:ss') desc;
SQL> SQL> 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Enter value for tbs_name: old 23: and d.NAME = '&tbs_name'
new 23: and d.NAME = ''
no rows selected
SQL> /
Enter value for tbs_name: USERS
old 23: and d.NAME = '&tbs_name'
new 23: and d.NAME = 'USERS'
SNAP_ID TS_NAME RTIME TS_SIZE_MB TS_USED_MB TS_FREE_MB INC_MB PCT_USED
---------- ------------------------------ ---------------- ---------- ---------- ---------- ---------- ----------
39581 USERS 2024-06-18 23:45 3854887.86 3780090.17 74797.69 1913.81 98.06%
39485 USERS 2024-06-17 23:45 3854887.86 3778176.36 76711.5 1710.5 98.01%
39389 USERS 2024-06-16 23:45 3854887.86 3776465.86 78422 832 97.97%
39293 USERS 2024-06-15 23:45 3854887.86 3775633.86 79254 1255.19 97.94%
39197 USERS 2024-06-14 23:45 3854887.86 3774378.67 80509.19 2311.1 97.91%
39101 USERS 2024-06-13 23:45 3854887.86 3772067.57 82820.29 1705 97.85%
39005 USERS 2024-06-12 23:45 3854887.86 3770362.57 84525.29 971.31 97.80%
38909 USERS 2024-06-11 23:45 3854887.86 3769391.26 85496.6 1564.5 97.78%
38813 USERS 2024-06-10 23:45 3854887.86 3767826.76 87061.1 1322.75 97.74%
38717 USERS 2024-06-09 23:45 3854887.86 3766504.06 88383.8 1089.2 97.70%
10 rows selected.
SQL>
继续查是哪个对象增长那么块占用USERS表空间，通过dba_segments查出USERS占用空间前十的对象，发现总共3700Gb的库有一个LOB对象居然占用了1600Gb多的空间(大概1/2的空间)。这就很像https://asktom.oracle.com/ords/asktom.search?tag=reclaimreuse-lob-space类似情况，那么这个SYS_LOB0000262715C00018$$段就成为了严重怀疑对象，继续定位这个LOB字段是哪个表，通过dba_lobs找到是PRINT_YJD_SD，统计表里记录数大概在69万。
SQL> col segment_name for a40
SQL> with t as (select t.owner,t.segment_name,t.segment_type,t.tablespace_name,bytes/1024/1024/1024 GB from dba_segments t where t.tablespace_name='USERS' order by 5 desc) select * from t where rownum<=10;
OWNER SEGMENT_NAME SEGMENT_TYPE TABLESPACE_NAME GB
------------------------------ ---------------------------------------- ------------------ -------------------- ----------
CMSDYPT SYS_LOB0000262715C00018$$ LOBSEGMENT USERS 1612.38773
GKSD6 WMS_INOUT_LOT_HISTORY TABLE USERS 117.87793
SAVEFEGKSD6 CMSFE_MSG_LOG_BAK TABLE USERS 112.451172
GKSD6 SCM_INOUT_LOT_HISTORY TABLE USERS 106.714844
GKSD6 PK_WMS_INOUT_LOT_HISTORY INDEX USERS 57.2041016
GKSD6 BPM_EVENTS_LOG TABLE USERS 56.7021484
SAVEFEGKSD6 SYS_LOB0000146929C00006$$ LOBSEGMENT USERS 53.2021484
FRUSER SYS_LOB0000348753C00013$$ LOBSEGMENT USERS 42.6552734
GKSD6 PK_SCM_INOUT_LOT_HISTORY INDEX USERS 39.8720703
GKSD6 SCM_INOUT_HISTORY TABLE USERS 39.2128906
10 rows selected.
SQL> col COLUMN_NAME for a40
SQL> set timing on
SQL> SELECT table_name, column_name, segment_name, a.bytes/1024/1024/1024 Gb FROM dba_segments a JOIN dba_lobs b USING (owner, segment_name) WHERE b.segment_name = 'SYS_LOB0000262715C00018$$';
TABLE_NAME COLUMN_NAME SEGMENT_NAME GB
------------------------------ ---------------------------------------- ---------------------------------------- ----------
PRINT_YJD_SD BASE SYS_LOB0000262715C00018$$ 1612.3877
Elapsed: 00:00:03.42
SQL> select count(1) from CMSDYPT.PRINT_YJD_SD;
COUNT(1)
----------
693385
Elapsed: 00:00:02.06
SQL> select sum(dbms_lob.getlength (BASE))/1024/1024/1024 GB from CMSDYPT.PRINT_YJD_SD;
GB
----------
64.0668227
Elapsed: 00:00:10.09
SQL>
我们也可以在AWR报告中的Segments by Direct Physical Writes部分也能定位到有问题表和LOB字段
处理过程
和客户方沟通这个表的BASE字段占用空间问题，客户方表示停业务可以清理这个表。由于可以停机操作，我们就放心大胆删除了部分数据后，通过SHRINK SPACE方式释放LOB空间，但释放过程中发现时间过长，4个多小时选择取消操作，改用expdp-drop-impdp的方式，expdp后我们看到最终dmp文件大小在129Gb。然后我们drop,create重建PRINT_YJD_SD表，再创建一个HUIFU的用户作为临时恢复过度用，最后精简到2024年有效数据。表空间释放出1600GB空间正常了。
SQL> ALTER TABLE CMSDYPT.PRINT_YJD_SD MODIFY LOB (BASE) (SHRINK SPACE);
ALTER TABLE CMSDYPT.PRINT_YJD_SD MODIFY LOB (BASE) (SHRINK SPACE)
*
ERROR at line 1:
ORA-00028: your session has been killed
ORA-00028: your session has been killed
Elapsed: 04:43:24.89
SQL>
SQL> exit
Disconnected from Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
[oracle@rac01 ~]$ expdp \'/ as sysdba\' directory=dmpdir dumpfile=20240619.dmp logfile=20240619.log tables=CMSDYPT.PRINT_YJD_SD
Export: Release 11.2.0.4.0 - Production on Wed Jun 19 14:07:24 2024
Copyright (c) 1982, 2011, Oracle and/or its affiliates. All rights reserved.
Connected to: Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
Starting "SYS"."SYS_EXPORT_TABLE_01": "/******** AS SYSDBA" directory=dmpdir dumpfile=20240619.dmp logfile=20240619.log tables=CMSDYPT.PRINT_YJD_SD
Estimate in progress using BLOCKS method...
Processing object type TABLE_EXPORT/TABLE/TABLE_DATA
Total estimation using BLOCKS method: 1612. GB
Processing object type TABLE_EXPORT/TABLE/TABLE
Processing object type TABLE_EXPORT/TABLE/INDEX/INDEX
Processing object type TABLE_EXPORT/TABLE/INDEX/STATISTICS/INDEX_STATISTICS
Processing object type TABLE_EXPORT/TABLE/STATISTICS/TABLE_STATISTICS
. . exported "CMSDYPT"."PRINT_YJD_SD" 128.2 GB 479394 rows
Master table "SYS"."SYS_EXPORT_TABLE_01" successfully loaded/unloaded
******************************************************************************
Dump file set for SYS.SYS_EXPORT_TABLE_01 is:
/u01/app/ora11g/20240619.dmp
Job "SYS"."SYS_EXPORT_TABLE_01" successfully completed at Wed Jun 19 16:33:55 2024 elapsed 0 02:26:27
[oracle@rac01 ~]$ ll -h /u01/app/ora11g/20240619.dmp
-rw-r----- 1 oracle oinstall 129G 6月 19 16:33 /u01/app/ora11g/20240619.dmp
[oracle@rac01 ~]$ sqlplus CMSDYPT
SQL*Plus: Release 11.2.0.4.0 Production on Wed Jun 19 16:35:56 2024
Copyright (c) 1982, 2013, Oracle. All rights reserved.
Enter password:
Connected to:
Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
SQL> drop table PRINT_YJD_SD;
Table dropped.
SQL> create table PRINT_YJD_SD
(
2 id VARCHAR2(200),
3 4 phid VARCHAR2(200),
spid VARCHAR2(200),
yjdh VARCHAR2(255),
hzid VARCHAR2(200),
wjlj VARCHAR2(255),
wjm VARCHAR2(255),
mark VARCHAR2(255),
pcdh VARCHAR2(255),
ddzh VARCHAR2(255),
khid VARCHAR2(255),
timestamps VARCHAR2(255),
dycs NUMBER default 0 not null,
dybz VARCHAR2(255) default 0 not null,
5 6 7 hz VARCHAR2(255),
8 9 khbm VARCHAR2(255),
khmc VARCHAR2(255),
base CLOB,
ddid VARCHAR2(30),
10 11 12 13 14 ddzhid VARCHAR2(50),
bt VARCHAR2(255),
15 16 xh VARCHAR2(255),
printflag VARCHAR2(255) default 0 not null,
yjdid VARCHAR2(50)
)
tablespace USERS
pctfree 10
initrans 1
17 18 maxtrans 255
storage
(
initial 64K
next 1M
minextents 1
maxextents unlimited
); 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
Table created.
SQL> create index IDX_YJ_PCDH on PRINT_YJD_SD (PCDH, DDZH)
tablespace USERS
pctfree 10
initrans 2
maxtrans 255
storage
(
initial 64K
next 1M
minextents 1
maxextents unlimited
); 2 3 4 5 6 7 8 9 10 11 12
Index created.
SQL> create user huifu identified by huifu default tablespace users;
User created.
SQL> grant dba to huifu;
Grant succeeded.
SQL> exit
Disconnected from Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
[oracle@rac01 ~]$ impdp huifu/huifu directory=dmpdir dumpfile=20240619.dmp logfile=imp20240619.log full=y remap_schema=CMSDYPT:HUIFU
Import: Release 11.2.0.4.0 - Production on Wed Jun 19 16:39:34 2024
Copyright (c) 1982, 2011, Oracle and/or its affiliates. All rights reserved.
Connected to: Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
Master table "HUIFU"."SYS_IMPORT_FULL_01" successfully loaded/unloaded
Starting "HUIFU"."SYS_IMPORT_FULL_01": huifu/******** directory=dmpdir dumpfile=20240619.dmp logfile=imp20240619.log full=y remap_schema=CMSDYPT:HUIFU
Processing object type TABLE_EXPORT/TABLE/TABLE
Processing object type TABLE_EXPORT/TABLE/TABLE_DATA
. . imported "HUIFU"."PRINT_YJD_SD" 128.2 GB 479394 rows
Processing object type TABLE_EXPORT/TABLE/INDEX/INDEX
Processing object type TABLE_EXPORT/TABLE/INDEX/STATISTICS/INDEX_STATISTICS
Processing object type TABLE_EXPORT/TABLE/STATISTICS/TABLE_STATISTICS
Job "HUIFU"."SYS_IMPORT_FULL_01" successfully completed at Wed Jun 19 18:42:28 2024 elapsed 0 02:02:53
[oracle@rac01 ~]$ sqlplus CMSDYPT
SQL*Plus: Release 11.2.0.4.0 Production on Wed Jun 19 18:43:24 2024
Copyright (c) 1982, 2013, Oracle. All rights reserved.
Enter password:
Connected to:
Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production
With the Partitioning, Real Application Clusters, Automatic Storage Management, OLAP,
Data Mining and Real Application Testing options
SQL> insert into PRINT_YJD_SD select * from huifu.PRINT_YJD_SD where wjlj like '%2024%';
116533 rows created.
SQL> commit;
Commit complete.
SQL> conn / as sysdba
Connected.
SQL> drop user huifu cascade;
User dropped.
SQL> set linesize 200
col tablespace_name for a20
select df.tablespace_name,
count(*) files,
round(sum(df.bytes) / 1048576) size_mb,
round(sum(free.bytes) / 1048576, 2) free_mb,
100 - round(100.0 * sum(free.bytes) / sum(df.bytes), 2) pct_used,
round(100.0 * sum(free.bytes) / sum(df.bytes), 2) pct_free
from dba_data_files df,
(select tablespace_name,
file_id,
sum(bytes) bytes,
max(bytes) maxbytes
from dba_free_space
group by tablespace_name, file_id) free
where df.tablespace_name = free.tablespace_name(+)
and df.file_id = free.file_id(+)
group by df.tablespace_name
order by pct_used desc
/SQL> SQL> 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
TABLESPACE_NAME FILES SIZE_MB FREE_MB PCT_USED PCT_FREE
-------------------- ---------- ---------- ---------- ---------- ----------
SYSTEM 2 11850 138 98.84 1.16
SYSAUX 2 44374 1398.19 96.85 3.15
USERS 118 3854888 1604830.44 58.37 41.63
UNDOTBS1 1 30355 24383.25 19.67 80.33
UNDOTBS2 1 28319 24739.06 12.64 87.36
SQL>
总结
表空间增长是DBA必须关注的，Oracle 11.2.0.4版本特别要注意对有BasicFiles LOB类型字段的表，频繁插入和删除DML操作的话，LOB空间一旦分配，就不会回收，除非表被删除或截断。如果要对LOB字段进行回收空间使用ALTER TABLE <table_name> MODIFY LOB (<column_name>) (SHRINK SPACE);
最后修改时间：2024-06-24 10:02:21
「喜欢这篇文章，您的关注和赞赏是给作者最好的鼓励」
关注作者
【版权声明】本文为墨天轮用户原创内容，转载时必须标注文章的来源（墨天轮），文章链接，文章作者等基本信息，否则作者和墨天轮有权追究责任。如果您发现墨天轮中有涉嫌抄袭或者侵权的内容，欢迎发送邮件至：contact@modb.pro进行举报，并提供相关证据，一经查实，墨天轮将立刻删除相关内容。
