ITPub博客

首页 > 数据库 > Oracle > AWR TOP SQL实现

AWR TOP SQL实现

原创 Oracle 作者:raysuen 时间:2019-07-08 11:25:49 0 删除 编辑

1 按解析次数排序
select a.*,
       to_char(substr(b.sql_text,1,4000))
from
    (select dhs.sql_id,
       sum(parse_calls_delta) parse,
       sum(executions_delta) exec_nums,
       dhs.MODULE
    from dba_hist_sqlstat dhs
    where
        snap_id  > 22438
        and snap_id <= 22440
    group by dhs.sql_id,MODULE) a,
    dba_hist_sqltext b
    where a.sql_id=b.sql_id order by a.parse desc;
2 按执行时间排序
select a.*,
       to_char(substr(b.sql_text,1,4000))
from
    (select dhs.sql_id,
       round(sum(elapsed_time_delta)/1000/1000,2) "elapsed_time(s)",
       sum(executions_delta) execs,
       round(sum(elapsed_time_delta)/1000/1000/sum(executions_delta),2)  elapsed_time_per,
       dhs.MODULE
    from dba_hist_sqlstat dhs
    where
        snap_id  > 22438
        and snap_id <= 22440
    group by dhs.sql_id,MODULE) a,
    dba_hist_sqltext b
    where a.sql_id=b.sql_id order by a."elapsed_time(s)" desc;
3 按CPU时间排序
select a.*,
       to_char(substr(b.sql_text,1,4000))
from
    (select dhs.sql_id,
       round(sum(cpu_time_delta)/1000/1000,2) "cpu_time",
       sum(executions_delta) execs,
       round(sum(cpu_time_delta)/1000/1000/sum(executions_delta),2)  cpu_time_per,
        round(sum(elapsed_time_delta)/1000/1000,2) "elapsed_time(s)",
       dhs.MODULE
    from dba_hist_sqlstat dhs
    where
        snap_id  > 22438
        and snap_id <= 22440
    group by dhs.sql_id,MODULE) a,
    dba_hist_sqltext b
    where a.sql_id=b.sql_id order by a."cpu_time" desc;
4 按User I/O wait排序
select a.*,
       to_char(substr(b.sql_text,1,4000))
from
    (select dhs.sql_id,
       round(sum(iowait_delta)/1000/1000,2) "iowait_time(s)",
       sum(executions_delta) execs,
       round(sum(iowait_delta)/1000/1000/sum(executions_delta),2)  iowait_time_per,
        round(sum(elapsed_time_delta)/1000/1000,2) "elapsed_time(s)",
       dhs.MODULE
    from dba_hist_sqlstat dhs
    where
        snap_id  > 22438
        and snap_id <= 22440
    group by dhs.sql_id,MODULE) a,
    dba_hist_sqltext b
    where a.sql_id=b.sql_id order by a."iowait_time(s)" desc;
5 按逻辑读(gets)排序
select a.*,
       to_char(substr(b.sql_text,1,4000))
from
    (select dhs.sql_id,
       round(sum(buffer_gets_delta),2) "buffer_ges",
       sum(executions_delta) execs,
       round(sum(buffer_gets_delta)/sum(executions_delta),2)  iowait_time_per,
        round(sum(elapsed_time_delta)/1000/1000,2) "elapsed_time(s)",
       dhs.MODULE
    from dba_hist_sqlstat dhs
    where
        snap_id  > 22438
        and snap_id <= 22440
    group by dhs.sql_id,MODULE) a,
    dba_hist_sqltext b
    where a.sql_id=b.sql_id order by a."buffer_ges" desc;
7 按物理读(physical read)排序
select a.*,
       to_char(substr(b.sql_text,1,4000))
from
    (select dhs.sql_id,
       round(sum(DISK_READS_DELTA),2) "physical_read",
       sum(executions_delta) execs,
       round(sum(DISK_READS_DELTA)/sum(executions_delta),2)  iowait_time_per,
        round(sum(elapsed_time_delta)/1000/1000,2) "elapsed_time(s)",
       dhs.MODULE
    from dba_hist_sqlstat dhs
    where
        snap_id  > 22438
        and snap_id <= 22440
    group by dhs.sql_id,MODULE) a,
    dba_hist_sqltext b
    where a.sql_id=b.sql_id order by a."physical_read" desc;

8 按执行次数排序
select a.*,
       to_char(substr(b.sql_text,1,4000))
from
    (select dhs.sql_id,
       round(sum(executions_delta),2) "exec_num",
       sum(ROWS_PROCESSED_DELTA) row_process,
       round(sum(ROWS_PROCESSED_DELTA)/sum(executions_delta),2)  rows_per_exec,
        round(sum(elapsed_time_delta)/1000/1000,2) "elapsed_time(s)",
       dhs.MODULE
    from dba_hist_sqlstat dhs
    where
        snap_id  > 22438
        and snap_id <= 22440
    group by dhs.sql_id,MODULE) a,
    dba_hist_sqltext b
    where a.sql_id=b.sql_id order by a."exec_num" desc;

来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/28572479/viewspace-2649803/,如需转载,请注明出处,否则将追究法律责任。

请登录后发表评论 登录
全部评论
擅长oracle,Linux。精通shell,python。

注册时间:2016-08-23

  • 博文量
    138
  • 访问量
    179110