Oracle清理SYSAUX表空间

11g的数据库因为BUG，导致与AWR相关的分区表不会自动split，这样一来，oracle自动清理策略不会生效，导致AWR数据越来越大，撑满SYSAUX表空间。 12c不存在这个问题。 检查快照实际保留时间与设定值的差异 14:51:28 SQL> col ash for a30 14:51:41 SQL> col SNAP for a30 14:51:46 SQL> col RETENTION for a30 14:51:51 SQL> select sysdate - a.sample_time ash, 14:51:53   2  sysdate - s.begin_interval_time snap, 14:51:53   3  c.RETENTION 14:51:53   4  from sys.wrm$_wr_control c, 14:51:53   5  ( 14:51:53   6  select db.dbid, 14:51:53   7  min(w.sample_time) sample_time 14:51:53   8  from sys.v_$database db, 14:51:54   9  sys.Wrh$_active_session_history w 14:51:54  10  where w.dbid = db.dbid group by db.dbid 14:51:54  11  ) a, 14:51:54  12  ( 14:51:54  13  select db.dbid, 14:51:54  14  min(r.begin_interval_time) begin_interval_time 14:51:54  15  from sys.v_$database db, 14:51:54  16  sys.wrm$_snapshot r 14:51:54  17  where r.dbid = db.dbid 14:51:54  18  group by db.dbid 14:51:54  19  ) s 14:51:54  20  where a.dbid = s.dbid 14:51:54  21  and c.dbid = a.dbid; ASH                            SNAP                           RETENTION ------------------------------ ------------------------------ ------------------------------ +000000851 21:45:18.673        +000000008 17:01:35.550        +00008 00:00:00.0 1 row selected. ASH保留了851天，但是设定值是8天。 SYSAUX表空间的使用情况 15:04:31 SQL> COL Item FOR A30 15:05:18 SQL> COL Schema FOR A30 15:05:49 SQL>   SELECT occupant_name "Item", 15:05:50   2           space_usage_kbytes / 1048576 "Space Used (GB)", 15:05:50   3           schema_name "Schema", 15:05:50   4           move_procedure "Move Procedure" 15:05:51   5      FROM v$sysaux_occupants 15:05:51   6    ORDER BY 2 desc ; Item                           Space Used (GB) Schema                         Move Procedure ------------------------------ --------------- ------------------------------ ---------------------------------------------------------------- SM/AWR                              10.2494507 SYS SM/OPTSTAT                          .566101074 SYS XDB                                 .123962402 XDB                            XDB.DBMS_XDB.MOVEXDB_TABLESPACE LOGMNR                              .073669434 SYSTEM                         SYS.DBMS_LOGMNR_D.SET_TABLESPACE SDO                                 .072509766 MDSYS                          MDSYS.MOVE_SDO SM/ADVISOR                          .067810059 SYS EM                                  .047058105 SYSMAN                         emd_maintenance.move_em_tblspc AO                                   .03729248 SYS                            DBMS_AW.MOVE_AWMETA XSOQHIST                             .03729248 SYS                            DBMS_XSOQ.OlapiMoveProc JOB_SCHEDULER                        .02557373 SYS ORDIM/ORDDATA                       .013244629 ORDDATA                        ordsys.ord_admin.move_ordim_tblspc SM/OTHER                             .00769043 SYS XSAMD                               .005004883 OLAPSYS                        DBMS_AMD.Move_OLAP_Catalog SMON_SCN_TIME                       .004150391 SYS EXPRESSION_FILTER                   .003540039 EXFSYS TEXT                                .003540039 CTXSYS                         DRI_MOVE_CTXSYS WM                                  .002380371 WMSYS                          DBMS_WM.move_proc SQL_MANAGEMENT_BASE                 .001647949 SYS PL/SCOPE                            .001525879 SYS STREAMS                             .001342773 SYS LOGSTDBY                            .001342773 SYSTEM                         SYS.DBMS_LOGSTDBY.SET_TABLESPACE EM_MONITORING_USER                  .000854492 DBSNMP ORDIM                               .000427246 ORDSYS                         ordsys.ord_admin.move_ordim_tblspc AUTO_TASK                           .000305176 SYS ORDIM/SI_INFORMTN_SCHEMA                     0 SI_INFORMTN_SCHEMA             ordsys.ord_admin.move_ordim_tblspc TSM                                          0 TSMSYS AUDIT_TABLES                                 0 SYS                            DBMS_AUDIT_MGMT.move_dbaudit_tables ULTRASEARCH_DEMO_USER                        0 WK_TEST                        MOVE_WK ULTRASEARCH                                  0 WKSYS                          MOVE_WK ORDIM/ORDPLUGINS                             0 ORDPLUGINS                     ordsys.ord_admin.move_ordim_tblspc STATSPACK                                    0 PERFSTAT 31 rows selected. Elapsed: 00:00:00.84 SM/AWR的含义是Server Manageability - Automatic Workload Repository，如果它的值大那么表示AWR信息容量大。 SM/OPTSTAT的含义是Server Manageability - Optimizer Statistics History，如果它的值大那么表示优化器统计信息容量大。 查询具体哪些表占用量大 15:06:00 SQL>   select * from ( 15:06:41   2   select segment_name,SEGMENT_TYPE,sum(bytes)/1024/1024 total_mb from dba_segments where tablespace_name = 15:06:41   3  'SYSAUX' group by segment_name,SEGMENT_TYPE order by 3 desc) 15:06:41   4   where rownum <=20; SEGMENT_NAME                                                                      SEGMENT_TYPE         TOTAL_MB --------------------------------------------------------------------------------- ------------------ ---------- WRH$_ACTIVE_SESSION_HISTORY                                                       TABLE PARTITION     2951.0625 WRH$_EVENT_HISTOGRAM_PK                                                           INDEX PARTITION      768.0625 WRH$_EVENT_HISTOGRAM                                                              TABLE PARTITION      736.0625 WRH$_LATCH                                                                        TABLE PARTITION      640.0625 WRH$_SQLSTAT                                                                      TABLE PARTITION      528.0625 WRH$_SYSSTAT_PK                                                                   INDEX PARTITION      432.0625 WRH$_SYSSTAT                                                                      TABLE PARTITION      408.0625 WRH$_PARAMETER_PK                                                                 INDEX PARTITION      384.0625 WRH$_LATCH_PK                                                                     INDEX PARTITION      376.0625 WRH$_LATCH_MISSES_SUMMARY                                                         TABLE PARTITION      336.0625 WRH$_PARAMETER                                                                    TABLE PARTITION      320.0625 WRH$_LATCH_MISSES_SUMMARY_PK                                                      INDEX PARTITION      320.0625 WRH$_ACTIVE_SESSION_HISTORY_PK                                                    INDEX PARTITION      304.0625 WRH$_SEG_STAT                                                                     TABLE PARTITION      246.0625 WRH$_SERVICE_STAT_PK                                                              INDEX PARTITION      168.0625 WRH$_SYSTEM_EVENT                                                                 TABLE PARTITION      168.0625 WRH$_SQLSTAT_PK                                                                   INDEX PARTITION      160.0625 SYS_LOB0000006409C00004$$                                                         LOBSEGMENT                144 I_WRI$_OPTSTAT_H_OBJ#_ICOL#_ST                                                    INDEX                     136 WRH$_SEG_STAT_PK                                                                  INDEX PARTITION      112.0625 20 rows selected. Elapsed: 00:00:00.09 查询快照信息： 15:06:41 SQL> SELECT MIN(SNAP_ID),MAX(SNAP_ID) FROM DBA_HIST_SNAPSHOT; MIN(SNAP_ID) MAX(SNAP_ID) ------------ ------------        20147        20354 1 row selected. Elapsed: 00:00:00.00 15:07:22 SQL> SELECT MIN(SNAP_ID),MAX(SNAP_ID) FROM DBA_HIST_ACTIVE_SESS_HISTORY; MIN(SNAP_ID) MAX(SNAP_ID) ------------ ------------            1        20354 1 row selected. Elapsed: 00:00:03.83 可以用下面的方法清理指定的快照数据。但是一般情况下，表空间告警的时候，WRH$_ACTIVE_SESSION_HISTORY已经很大了，下面的方法其实就是delete，奇慢无比。 EXEC dbms_workload_repository.drop_snapshot_range(low_snap_id=>1, high_snap_id=>3); 上面的存储过程最主要的其实是下面的语句，可以通过v$session和v$sql抓取出来。 delete from WRH$_ACTIVE_SESSION_HISTORY tab where (:beg_snap <= tab.snap_id and         tab.snap_id <= :end_snap and         dbi d = :dbid)    and not exists (select 1 from WRM$_BASELINE b                where (tab.dbid = b.dbid) and       (tab.snap_id >= b.start_snap_id) and     (tab.snap_id <= b.end_snap_id)) 下面的命令可以展示每个分区最大最小的snap id 15:29:35 SQL> set serveroutput on 15:29:43 SQL> declare 15:29:43   2  CURSOR cur_part IS 15:29:43   3  SELECT partition_name from dba_tab_partitions 15:29:43   4  WHERE table_name = 'WRH$_ACTIVE_SESSION_HISTORY'; 15:29:43   5   15:29:43   6  query1 varchar2(200); 15:29:43   7  query2 varchar2(200); 15:29:43   8   15:29:43   9  TYPE partrec IS RECORD (snapid number, dbid number); 15:29:43  10  TYPE partlist IS TABLE OF partrec; 15:29:43  11   15:29:43  12  Outlist partlist; 15:29:43  13  begin 15:29:43  14  dbms_output.put_line('PARTITION NAME SNAP_ID DBID'); 15:29:43  15  dbms_output.put_line('--------------------------- ------- ----------'); 15:29:43  16   15:29:43  17  for part in cur_part loop 15:29:43  18  query1 := 'select min(snap_id), dbid from sys.WRH$_ACTIVE_SESSION_HISTORY partition ('||part.partition_name||') group by dbid'; 15:29:43  19  execute immediate query1 bulk collect into OutList; 15:29:44  20   15:29:44  21  if OutList.count > 0 then 15:29:44  22  for i in OutList.first..OutList.last loop 15:29:44  23  dbms_output.put_line(part.partition_name||' Min '||OutList(i).snapid||' '||OutList(i).dbid); 15:29:44  24  end loop; 15:29:44  25  end if; 15:29:44  26   15:29:44  27  query2 := 'select max(snap_id), dbid from sys.WRH$_ACTIVE_SESSION_HISTORY partition ('||part.partition_name||') group by dbid'; 15:29:44  28  execute immediate query2 bulk collect into OutList; 15:29:44  29   15:29:44  30  if OutList.count > 0 then 15:29:44  31  for i in OutList.first..OutList.last loop 15:29:44  32  dbms_output.put_line(part.partition_name||' Max '||OutList(i).snapid||' '||OutList(i).dbid); 15:29:44  33  dbms_output.put_line('---'); 15:29:44  34  end loop; 15:29:44  35  end if; 15:29:44  36   15:29:44  37  end loop; 15:29:44  38  end; 15:29:44  39  / PARTITION NAME SNAP_ID DBID --------------------------- ------- ---------- WRH$_ACTIVE_3245728642_0 Min 1 3245728642 WRH$_ACTIVE_3245728642_0 Max 20354 3245728642 --- PL/SQL procedure successfully completed. Elapsed: 00:00:09.44 下面展示一种清理方法。总体来说，就是保留最近的数据，将表清空，再将数据插回去。 除此之外，前面提到过此问题最关键的问题是因为分区不能自动split，下面的命令可以将所有与AWR有关的分区表全部split出一个新的分区出来。 alter session set "_swrf_test_action" = 72; 这样就可以自动等待上一个分区数据过期，自动清理，而且还会清理所有的相关分区表。下面我展示的方法，只会清理最大的那张表，其他表中的过期数据还是存在，但是占用量不大，所以也可以接受。      1. 创建临时表，保存近8天的数据 create table mingshuo.ash_bak_20190610 as select * from sys.WRH$_ACTIVE_SESSION_HISTORY where SAMPLE_TIME>=sysdate-9; SQL> select count(*)  from mingshuo.ash_bak_20190610;   COUNT(*) ----------      12648 1 row selected. Elapsed: 00:00:00.01 2. 禁用AWR exec dbms_workload_repository.modify_snapshot_settings(interval => 0); 3. truncate WRH$_ACTIVE_SESSION_HISTORY TRUNCATE TABLE sys.WRH$_ACTIVE_SESSION_HISTORY; 4. truncate 完成后，重建 WRH$_ACTIVE_SESSION_HISTORY表的索引，并从备份表恢复数据 insert into sys.WRH$_ACTIVE_SESSION_HISTORY select * from mingshuo.ash_bak_20190610; commit; 5. 查看索引状态 set line 300 pages 200 col owner for a20 col index_name for a30 select index_owner, index_name, partition_name, status, tablespace_name, last_analyzed from dba_ind_partitions where index_name in (select index_name from dba_indexes where table_name in ('WRH$_ACTIVE_SESSION_HISTORY') and table_owner = 'SYS'); 6. 如果索引失效重建索引（测试的时候是没有失效的，索引为本地索引） alter index sys.WRH$_ACTIVE_SESSION_HISTORY_PK rebuild parallel 8 nologging; alter index sys.WRH$_ACTIVE_SESSION_HISTORY_PK noparallel; 7. 启动AWR exec dbms_workload_repository.modify_snapshot_settings(interval => 60); 8. 测试awr和ash可能正常生成 ---手动生成snapshot exec dbms_workload_repository.create_snapshot; @?/rdbms/admin/awrrpt @?/rdbms/admin/ashrpt 9. 删除备份表 drop table mingshuo.ash_bak_20190610 purge;