概述

生产环境中,经常会遇到表由于数据不断插入,导致空间越来越大,由于前期配置问题,没有做分区或者其他优化,而且生产数据实时向表插入。要删除历史数据来释放空间。所以DBA一般都需要定期去对Oracle表碎片做整理,简单整理表碎片整理流程如下:
1、定位存在碎片的对象
使用如下脚本,检查需要进行碎片整理的对象:
- --all tables(partition_tables + non_partition_tables )
 - select a.owner,
 - a.table_name,
 - a.num_rows,
 - a.avg_row_len,
 - round(a.avg_row_len * a.num_rows / 1024 / 1024, 2) real_bytes_MB,
 - round(b.seg_bytes_mb, 2) seg_bytes_mb,
 - decode(a.num_rows,
 - 0,
 - 100,
 - (1 - round(a.avg_row_len * a.num_rows / 1024 / 1024 /
 - b.seg_bytes_mb,
 - 2)) * 100) || '%' frag_percent
 - from dba_tables a,
 - (select owner, segment_name, sum(bytes / 1024 / 1024) seg_bytes_mb
 - from dba_segments
 - group by owner, segment_name) b
 - where a.table_name = b.segment_name
 - and a.owner = b.owner
 - and a.owner not in
 - ('SYS', 'SYSTEM', 'OUTLN', 'DMSYS', 'TSMSYS', 'DBSNMP', 'WMSYS',
 - 'EXFSYS', 'CTXSYS', 'XDB', 'OLAPSYS', 'ORDSYS', 'MDSYS', 'SYSMAN')
 - and decode(a.num_rows,
 - 0,
 - 100,
 - (1 - round(a.avg_row_len * a.num_rows / 1024 / 1024 /
 - b.seg_bytes_mb,
 - 2)) * 100) > 30
 - order by b.seg_bytes_mb desc;
 
2、统计信息检查
2.1 统计信息检查
查看统计信息收集日期,确保碎片查询结果准确:
- select owner,table_name,last_analyzed from dba_tables Where owner='
 ' AND table_name=' '; 
2.2 统计信息收集
如果统计信息过旧,则重新收集统计信息:
- exec dbms_stats.gather_table_stats(ownname=>'
 ', tabname =>' '); 
3、表碎片整理
3.1 打开行移动
- alter table
 enable row movement ; 
3.2 进行表收缩
- alter table
 shrink space cascade ; 
3.3 失效对象编译
语句可能会造成引用表 的对象(如存储过程、包、视图等)变为无效。
运行如下脚本,重新编译失效对象。
- @?/rdbms/admin/utlrp.sql
 
4、对象收缩后的结果检查
运行如下脚本,确认对象空间是否已经完成收缩。
- --all tables(partition_tables + non_partition_tables )
 - select a.owner,
 - a.table_name,
 - a.num_rows,
 - a.avg_row_len,
 - round(a.avg_row_len * a.num_rows / 1024 / 1024, 2) real_bytes_MB,
 - round(b.seg_bytes_mb, 2) seg_bytes_mb,
 - decode(a.num_rows,
 - 0,
 - 100,
 - (1 - round(a.avg_row_len * a.num_rows / 1024 / 1024 /
 - b.seg_bytes_mb,
 - 2)) * 100) || '%' frag_percent
 - from dba_tables a,
 - (select owner, segment_name, sum(bytes / 1024 / 1024) seg_bytes_mb
 - from dba_segments
 - group by owner, segment_name) b
 - where a.table_name = b.segment_name
 - and a.owner = b.owner
 - and a.owner not in
 - ('SYS', 'SYSTEM', 'OUTLN', 'DMSYS', 'TSMSYS', 'DBSNMP', 'WMSYS',
 - 'EXFSYS', 'CTXSYS', 'XDB', 'OLAPSYS', 'ORDSYS', 'MDSYS', 'SYSMAN')
 - and decode(a.num_rows,
 - 0,
 - 100,
 - (1 - round(a.avg_row_len * a.num_rows / 1024 / 1024 /
 - b.seg_bytes_mb,
 - 2)) * 100) > 30
 - order by b.seg_bytes_mb desc;
 
5、性能监控
监控数据库会话,是否存在异常等待事件:
- select inst_id ,sid,serial#,sql_id,event,machine,module,program,seconds_in_wait from gv$session ;
 - --看会话在做什么操作
 - select sid, sql_text
 - from v$session a, v$sql b
 - where sid in(85,160)
 - and(b.sql_id = a.sql_id or b.sql_id = a.prev_sql_id);