Comments on: SQL execution times from ASH

By: Kyle Hailey

Kyle Hailey — Mon, 03 Jun 2013 03:57:11 +0000

MX AV and MIN are the maximum, average and minimum execution times for the SQL statements in seconds, but seconds based on ASH sampling, so it’s not exact but statistically sufficient for queries that take more than a second.

By: lmwang

lmwang — Mon, 03 Jun 2013 01:55:06 +0000

Hi Kyle,
Do MX AV MIN mean in seconds?
THanks
lmwang

By: Drill down to sql_id execution details in ASH | bdt's oracle blog

Drill down to sql_id execution details in ASH | bdt's oracle blog — Fri, 19 Apr 2013 13:17:59 +0000

[...] metrics as avg, min, max execution time: You can see some good examples of sql_exec_id usage into Kyle Hailey’s blog post or Karl Arao’s [...]

By: dboptimizer.com | xfang

dboptimizer.com | xfang — Fri, 28 Dec 2012 14:38:10 +0000

[...] http://dboptimizer.com/2011/05/04/sql-execution-times-from-ash/ [...]

By: DB Optimizer » Tuning Blog Entries

DB Optimizer » Tuning Blog Entries — Thu, 20 Oct 2011 17:11:57 +0000

[...] SQL execution times from ASH – using ASH to see SQL execution times and execution time variations [...]

By: Kyle Hailey

Kyle Hailey — Wed, 20 Jul 2011 18:47:06 +0000

@Kyle Hailey SOmething like this should work with SASH for queries over 5 minutes:

select * from (
   select
        count(*),
        sql_id,
        fixed_table_sequence
   from sash
   where
          sql_id > 0
     and  sql_id is not null
   group by
            sql_id,
            fixed_table_sequence,
   having count(*) > 300 
/

By: DB Optimizer » SQL timings for ASH II

DB Optimizer » SQL timings for ASH II — Thu, 19 May 2011 04:08:52 +0000

[...] recent commentsDB Optimizer » SQL execution times from ASH on SQL ASH timings IIIMark Entwistle on ToolsKyle Hailey on SQL ASH timings IIIKyle Hailey on SQL [...]

By: Kyle Hailey

Kyle Hailey — Fri, 13 May 2011 04:33:21 +0000

@coskan gundogar Hi Coskan, Do you have any example output? Or better yet an export of the AWR that I could import and run queries on? THanks Kyle

By: coskan gundogar

coskan gundogar — Wed, 11 May 2011 11:54:43 +0000

Hi Kyle,

Script is very useful and sql_exec_id is very wise to use.

One thing to bear in mind, I don’t know if it is a bug on our 11.2.0.1 DB but I can see a sql which runs 2 seconds has got sql_exec_start_date back in 15 days ago which can cause misleading conclusions with the sql you shared.

By: Kyle Hailey

Kyle Hailey — Fri, 06 May 2011 23:56:28 +0000

If you use SASH you can do similar queries using fixed_table_sequence: This shows the top twolongest running queries every 15 minute period. The higher the count the longer they have been running (you can't do this with 10g ASH)

   START_TIME     COUNT(*)     SQL_ID FIXED_TABLE_SEQUENCE       TOPN
   ------------ ---------- ---------- -------------------- ----------
   JUN-01 02:00         29  795857215                34254          1
   JUN-01 02:00         29  795857215                34547          1
   JUN-01 02:00         29  986013188                34305          1
   JUN-01 02:15        105  452165714                35961          1
   JUN-01 02:15        102  866152581                35999          2
   JUN-01 02:30        119 2586830956                39017          1
   JUN-01 02:30        109  452165714                39130          2

break on  start_time
clear breaks
column f_secs new_value v_secs
column f_mins new_value v_mins
select 900 f_secs from dual;
select &v_secs/60 f_mins from dual;

select * from (
   select
        to_char(trunc(sample_time,'HH')+
             (trunc((sample_time-trunc(sample_time,'HH'))*24/(&v_mins/60))+1)/(60/&v_mins)/24 ,'MON-DD HH:MI') start_time,
        count(*),
        sql_id,
        fixed_table_sequence,
        rank()
          over (
            partition by
                    trunc(sample_time,'HH')+
             (trunc((sample_time-trunc(sample_time,'HH'))*24/(&v_mins/60))+1)/(60/&v_mins)/24
            order by count(*) desc ) topn
   from sash
   where
          sql_id > 0
     and  sql_id is not null
     and  sample_time > sysdate - (30)/(24*60)
   group by
            sql_id,
            fixed_table_sequence,
            trunc(sample_time,'HH')+
            (trunc((sample_time-trunc(sample_time,'HH'))*24/(&v_mins/60))+1)/(60/&v_mins)/24
)
where  topn < 3
/
clear breaks