tzwweixiao 2019-06-25
我们在维护业务系统时,可能会建立很多索引,那么这些索引的使用到底怎么样,是否有些索引一直都没有用到过,那么oracle 是如何监控索引的使用状况,是否可以清除它们?
一般有两种方式:
1、直接监控索引的使用情况
(1)设置所要监控的索引:ALTER INDEX IDX_T_XX MONITORING USAGE;
(2)查看该索引有没有被使用:SELECT * FROM V$OBJECT_USAGE;
(3)关闭监控:ALTER INDEX IDX_T_XX NOMONITORING USAGE;
2、schema级别索引监控
如果我们想在系统中监控所有的索引,可以通过下面脚本实现监控数据库所有的索引。注意我们要排除一些系统表的索引、以及LOB indexes。原因有下面两个:
1:LOB indexes不能修改,否则会报ORA-22864错误(ORA-22864: cannot ALTER or DROP LOB indexes)。
2:ORA-00701: object necessary for warmstarting database cannot be altered
今天主要也是讲schema级别这块的索引监控,下面介绍下三个脚本。
直接执行脚本来开启索引监控,当然监控索引时长非常重要,太短的话有可能导致查询出来的数据有问题,一般建议监控一周后即可,OLAP系统则需要适当延长监控的时间。
SELECT 'ALTER INDEX ' || owner || '.' || index_name || ' MONITORING USAGE;' enable_monitor, 'ALTER INDEX ' || owner || '.' || index_name || ' NOMONITORING USAGE;' disable_monitor FROM dba_indexes WHERE INDEX_TYPE != 'LOB' and owner IN (SELECT username FROM dba_users WHERE account_status = 'OPEN') AND owner NOT IN ('SYS', 'SYSTEM', 'PERFSTAT', 'MGMT_VIEW', 'MONITOR', 'SYSMAN', 'DBSNMP') AND owner not like '%SYS%';
监控一个月就大概可以知道那些是无用的索引了。
虽然v$object_usage表能记录索引监控和使用的状态,但它不能统计索引被使用的次数和频率,只记录了在开启索引监控的时间段索引是否被使用过,这一点要值的注意。
另外需要注意的2点:
① 10g在收集统计信息时会导致索引被监控、这并非SQL语句产生、而在11g则不会出现这种情况了
② 外键索引不会因为主表的DML操作而被监控到、不要因为该索引没用而将它给删了。
查询V$OBJECT_USAGE就可以知道数据库对索引的使用情况了。通过一段时间的监控,就可以确定哪些是无用的索引。
另外,为了避免使用V$OBJECT_USAGE只能查询到当前用户下索引的监控情况,可以使用如下语句查询数据库中所有被监控索引的使用情况:
SELECT U.NAME OWNER, IO.NAME INDEX_NAME, T.NAME TABLE_NAME, DECODE(BITAND(I.FLAGS, 65536), 0, 'NO', 'YES') MONITORING, DECODE(BITAND(OU.FLAGS, 1), 0, 'NO', 'YES') USED, OU.START_MONITORING START_MONITORING, OU.END_MONITORING END_MONITORING FROM SYS.USER$ U, SYS.OBJ$ IO, SYS.OBJ$ T, SYS.IND$ I, SYS.OBJECT_USAGE OU WHERE I.OBJ# = OU.OBJ# AND IO.OBJ# = OU.OBJ# AND T.OBJ# = I.BO# AND U.USER# = IO.OWNER#;
其实也可以从视图DBA_HIST_SQL_PLAN中获取到数据库中所有索引的扫描次数情况,然后根据扫描次数和开发人员沟通是否需要保留索引。
WITH TMP1 AS (SELECT I.OWNER INDEX_OWNER, I.TABLE_OWNER, TABLE_NAME, INDEX_NAME, INDEX_TYPE, (SELECT NB.CREATED FROM DBA_OBJECTS NB WHERE NB.OWNER = I.OWNER AND NB.OBJECT_NAME = I.INDEX_NAME AND NB.SUBOBJECT_NAME IS NULL AND NB.OBJECT_TYPE = 'INDEX') CREATED, (SUM(S.BYTES) / 1024 / 1024) INDEX_MB, (SELECT COUNT(1) FROM DBA_IND_COLUMNS DIC WHERE DIC.INDEX_NAME = I.INDEX_NAME AND DIC.TABLE_NAME = I.TABLE_NAME AND DIC.INDEX_OWNER = I.OWNER) COUNT_INDEX_COLS FROM DBA_SEGMENTS S, DBA_INDEXES I WHERE I.INDEX_NAME = S.SEGMENT_NAME AND I.OWNER = S.OWNER AND S.OWNER NOT LIKE '%SYS%' GROUP BY I.OWNER, I.TABLE_OWNER, TABLE_NAME, INDEX_NAME, INDEX_TYPE HAVING SUM(S.BYTES) > 1024 * 1024), TMP2 AS (SELECT INDEX_OWNER, INDEX_NAME, PLAN_OPERATION, (SELECT MIN(TO_CHAR(NB.BEGIN_INTERVAL_TIME, 'YYYY-MM-DD HH24:MI:SS')) FROM DBA_HIST_SNAPSHOT NB WHERE NB.SNAP_ID = V.MIN_SNAP_ID) MIN_DATE, (SELECT MAX(TO_CHAR(NB.END_INTERVAL_TIME, 'YYYY-MM-DD HH24:MI:SS')) FROM DBA_HIST_SNAPSHOT NB WHERE NB.SNAP_ID = V.MAX_SNAP_ID) MAX_DATE, COUNTS FROM (SELECT D.OBJECT_OWNER INDEX_OWNER, D.OBJECT_NAME INDEX_NAME, D.OPERATION || ' ' || D.OPTIONS PLAN_OPERATION, MIN(H.SNAP_ID) MIN_SNAP_ID, MAX(H.SNAP_ID) MAX_SNAP_ID, COUNT(1) COUNTS FROM DBA_HIST_SQL_PLAN D, DBA_HIST_SQLSTAT H WHERE D.OPERATION LIKE '%INDEX%' AND D.SQL_ID = H.SQL_ID GROUP BY D.OBJECT_OWNER, D.OBJECT_NAME, D.OPERATION, D.OPTIONS) V) SELECT A.TABLE_OWNER, A.TABLE_NAME, A.INDEX_OWNER, A.INDEX_NAME, A.CREATED, A.INDEX_TYPE, A.INDEX_MB, A.COUNT_INDEX_COLS, B.PLAN_OPERATION, CASE WHEN MIN_DATE IS NULL THEN (SELECT MIN(TO_CHAR(NB.BEGIN_INTERVAL_TIME, 'YYYY-MM-DD HH24:MI:SS')) FROM DBA_HIST_SNAPSHOT NB) ELSE MIN_DATE END AS MIN_DATE, CASE WHEN MAX_DATE IS NULL THEN (SELECT MAX(TO_CHAR(NB.BEGIN_INTERVAL_TIME, 'YYYY-MM-DD HH24:MI:SS')) FROM DBA_HIST_SNAPSHOT NB) ELSE MAX_DATE END AS MAX_DATE, COUNTS FROM TMP1 A LEFT OUTER JOIN TMP2 B ON (A.INDEX_OWNER = B.INDEX_OWNER AND A.INDEX_NAME = B.INDEX_NAME);
后面会分享更多devops和DBA方面的内容,感兴趣的朋友可以关注一下~
另外一部分,则需要先做聚类、分类处理,将聚合出的分类结果存入ES集群的聚类索引中。数据处理层的聚合结果存入ES中的指定索引,同时将每个聚合主题相关的数据存入每个document下面的某个field下。