----------------------------------------

--SQL Activity Chart.

--Version 1.3.0

--Copyright (C) 2022. This program is licensed under the LGPLv3.

--

--How to use:

-- 1. Change sample times and time_chunks in the "configuration" table below

-- (If there was a database refresh or upgrade, you may need to change the DBID.)

-- 2. Run the query and view results in a fixed-width font.

--This query may take a while to finish, depending on the time period.

----------------------------------------

with configuration as

(

select

systimestamp - 1 sample_start_time,

systimestamp sample_end_time,

--Specify absolute times like this:

--timestamp '2022-04-01 00:00:01' sample_start_time,

--timestamp '2022-04-01 03:00:00' sample_end_time,

1440 time_chunks,

--If this isn't the right DBID, check in DBA_HIST_SNAPSHOT.)

(select dbid from v$database) dbid

--1307358397 dbid

from dual

)

----------------------------------------

--(Do not change anything below this line)

select chart

from

(

--#1: Header.

select -7 interval_number, 'SQL Activity chart.' chart from dual union all

select -6 interval_number, 'https://github.com/jonheller1/SQL_Activity_Chart' chart from dual union all

select -5 interval_number, '' chart from dual union all

select -4 interval_number, 'Generated for '||(select global_name from global_name)||' on '||to_char(sysdate, 'YYYY-MM-DD HH24:MI')||'.' chart from dual union all

select -3 interval_number, '' chart from dual union all

select -2 interval_number, ' Sample Period SQL (one case-sensitive letter per active session, see SQL Key at bottom)' chart from dual union all

select -1 interval_number, ' ---------------------------------------------------------------' chart from dual

union all

--#2: Chart.

select time_intervals.interval_number, to_char(begin_time, 'YYYY-MM-DD HH24:MI') || ' | ' || letters chart

from

(

--Time intervals.

select

min_begin, max_end

,min_begin + ((level-1) * time_interval) begin_time

,min_begin + ((level) * time_interval) - interval '1' second end_time

,level interval_number, samples_per_interval

from

(

--Times and chunk size.

select

max(end_interval_time) max_end,

min(begin_interval_time) min_begin,

(max(end_interval_time) - min(begin_interval_time)) / max((select time_chunks from configuration)) time_interval,

(cast(max(end_interval_time) as date) - cast(min(begin_interval_time) as date))

/ max((select time_chunks from configuration)) /*chunks*/ *24*60*60 /*convert to seconds*/ / 10 /*sample every 10 seconds*/ samples_per_interval,

max((select time_chunks from configuration)) time_chunks

from dba_hist_snapshot

where begin_interval_time between (select sample_start_time from configuration) and (select sample_end_time from configuration)

) times

connect by level <= time_chunks

) time_intervals

left join

(

--Events per interval.

select interval_number, listagg(letter) within group (order by letter) letters

from

(

--Combine letters and OTHER.

select

interval_number, min_begin, max_end, begin_time, end_time, username, sql_id, sql_text, counts_per_user_and_sql, event_counts,

case

when is_other = 0 then round(avg_events_per_sample)

when is_other = 1 then unrepresented_sessions

end avg_events_per_sample,

case

when is_other = 0 then letter

when is_other = 1 then '~'

end letter

from

(

--Add "OTHER" logic.

select

interval_number, min_begin, max_end, begin_time, end_time, average_events.username, average_events.sql_id, sql_text, counts_per_user_and_sql, event_counts,

letter,avg_events_per_sample,

--Total activity minus activity that is represented by a letter and has more than 0.5 average events per sample.

round

(

sum(avg_events_per_sample) over (partition by interval_number)

-

sum(case when avg_events_per_sample >= 0.5 and letter is not null then round(avg_events_per_sample) else 0 end) over (partition by interval_number)

) unrepresented_sessions,

case when avg_events_per_sample < 0.5 or letter is null then 1 else 0 end is_other,

row_number() over (partition by interval_number order by case when avg_events_per_sample < 0.5 or letter is null then 0 else 1 end, 1) other_rownumber

--DEBUG

--average_events.*, sql_key.*

from

(

--Average events per sample.

select min_begin, max_end, begin_time, end_time, interval_number, username, sql_id

,count(*) / samples_per_interval avg_events_per_sample

from

(

--Time intervals.

select /*+ cardinality(50) */

min_begin, max_end

,min_begin + ((level-1) * time_interval) begin_time

,min_begin + ((level) * time_interval) - interval '1' second end_time

,level interval_number, samples_per_interval

from

(

--Times and chunk size.

select

max(end_interval_time) max_end,

min(begin_interval_time) min_begin,

(max(end_interval_time) - min(begin_interval_time)) / max((select time_chunks from configuration)) time_interval,

(cast(max(end_interval_time) as date) - cast(min(begin_interval_time) as date))

/ max((select time_chunks from configuration)) /*chunks*/ *24*60*60 /*convert to seconds*/ / 10 /*sample every 10 seconds*/ samples_per_interval,

max((select time_chunks from configuration)) time_chunks

from dba_hist_snapshot

where begin_interval_time between (select sample_start_time from configuration) and (select sample_end_time from configuration)

) times

connect by level <= time_chunks

) time_intervals

left join

(

--Events

select username, sql_id, sample_time

from dba_hist_active_sess_history

join dba_users on dba_hist_active_sess_history.user_id = dba_users.user_id

where dbid = (select dbid from configuration) --For the partition pruning.

and snap_id in

(

select /*+ cardinality(24) */ snap_id

from dba_hist_snapshot

where begin_interval_time between (select sample_start_time from configuration) and (select sample_end_time from configuration)

)

and sql_id is not null

) events

on events.sample_time between time_intervals.begin_time and time_intervals.end_time

group by min_begin, max_end, begin_time, end_time, interval_number, username, sql_id, samples_per_interval

order by interval_number

) average_events

left join

(

--SQL Key - sums and key for the top 52 SQL IDs.

--Top user and SQL, with text, counts, and a KEY letter

select rownumber, letter, username, count_per_user_sql_w_rownum.sql_id, cast(substr(sql_text, 1, 100) as varchar2(100)) sql_text, counts_per_user_and_sql, event_counts

from

(

--Count per user and SQL, with rownum

select username, sql_id, event_counts, counts_per_user_and_sql, rownum rownumber

from

(

--Count per user and SQL.

select username, sql_id

,listagg(event||' ('||event_count||')', ',') within group (order by event_count desc) event_counts

,sum(event_count) counts_per_user_and_sql

from

(

--Count per user, SQL, and event.

select username, sql_id, nvl(event, 'CPU') event, count(*) event_count

from dba_hist_active_sess_history

join dba_users on dba_hist_active_sess_history.user_id = dba_users.user_id

where dbid = (select dbid from configuration)

and snap_id in

(

select /*+ cardinality(24) */ snap_id

from dba_hist_snapshot

where begin_interval_time between (select sample_start_time from configuration) and (select sample_end_time from configuration)

)

and sql_id is not null

group by username, sql_id, event

) count_per_user_sql_event

group by username, sql_id

order by counts_per_user_and_sql desc

) count_per_user_sql

) count_per_user_sql_w_rownum

join

(

--Letters.

select /*+ cardinality(52) */

level letter_number,

chr(64+case when level <= 26 then level else level + 6 end) letter

from dual

connect by level <= 52

) letters

on count_per_user_sql_w_rownum.rownumber = letters.letter_number

left join

(

select sql_id, sql_text

from dba_hist_sqltext

where dbid = (select dbid from configuration)

) sqltext

on count_per_user_sql_w_rownum.sql_id = sqltext.sql_id

) sql_key

on average_events.username = sql_key.username

and average_events.sql_id = sql_key.sql_id

order by interval_number

) add_others

--Include the regular rows and only one of the "OTHER" rows.

where is_other = 0 or (is_other = 1 and other_rownumber = 1)

) combine_letter_and_other

join

(

--Duplicator.

--H: Isn't that your transmogrifier?

--C: It WAS. But I made some modifications. See, the box is on its side now. It's a duplicator!

select level duplicator_number

from dual

--Realistically there will never be more than 200 events per sample.

connect by level <= 200

)

on avg_events_per_sample >= duplicator_number

group by interval_number

) events_per_interval

on time_intervals.interval_number = events_per_interval.interval_number

----------------------------------------

union all

--#3: Key header.

select 1000001 interval_number, ' ---------------------------------------------------------------' chart from dual union all

select 1000002 interval_number, '' chart from dual union all

select 1000003 interval_number, 'SQL Key - Statements are ordered by most active first' chart from dual union all

select 1000004 interval_number, '==============================================================================================================================================================================================================================================================================' chart from dual union all

select 1000005 interval_number, '| ID| Username | SQL_ID | SQL Text | Samples | Sample counts per event (ordered by most common events first) |' chart from dual union all

select 1000006 interval_number, '==============================================================================================================================================================================================================================================================================' chart from dual

----------------------------------------

union all

--#4: Key.

select

2000000 + rownumber interval_number

,'| '||letter||' | '||rpad(username, 30, ' ')||' | '||sql_id||' | '||rpad(replace(replace(replace(nvl(sql_text, ' '), chr(10), null), chr(13), null), chr(9), ' '), 50, ' ')||' | '||nvl(lpad(counts_per_user_and_sql, 7, ' '), ' ')||' | '||rpad(event_counts, 150, ' ')||' |' chart

from

(

--SQL Key - sums and key for the top 52 SQL IDs.

--Top user and SQL, with text, counts, and a KEY letter

select rownumber, letter, username, count_per_user_sql_w_rownum.sql_id, cast(substr(sql_text, 1, 1000) as varchar2(1000)) sql_text, counts_per_user_and_sql, event_counts

from

(

--Count per user and SQL, with rownum

select username, sql_id, event_counts, counts_per_user_and_sql, rownum rownumber

from

(

--Count per user and SQL.

select username, sql_id

,listagg(event||' ('||event_count||')', ',') within group (order by event_count desc) event_counts

,sum(event_count) counts_per_user_and_sql

from

(

--Count per user, SQL, and event.

select username, sql_id, nvl(event, 'CPU') event, count(*) event_count

from dba_hist_active_sess_history

join dba_users on dba_hist_active_sess_history.user_id = dba_users.user_id

where dbid = (select dbid from configuration)

and snap_id in

(

select /*+ cardinality(24) */ snap_id

from dba_hist_snapshot

where begin_interval_time between (select sample_start_time from configuration) and (select sample_end_time from configuration)

)

and sql_id is not null

group by username, sql_id, event

) count_per_user_sql_event

group by username, sql_id

order by counts_per_user_and_sql desc

) count_per_user_sql

) count_per_user_sql_w_rownum

join

(

--Letters.

select /*+ cardinality(52) */

level letter_number,

chr(64+case when level <= 26 then level else level + 6 end) letter

from dual

connect by level <= 52

) letters

on count_per_user_sql_w_rownum.rownumber = letters.letter_number

left join

(

select sql_id, sql_text

from dba_hist_sqltext

where dbid = (select dbid from configuration)

) sqltext

on count_per_user_sql_w_rownum.sql_id = sqltext.sql_id

--Add "OTHER" statements.

union all

select 999 rownumber, '~' letter, 'other' username, 'other ' sql_id

,'activity not caused by one of the Top N queries' SQL_TEXT, null counts_per_user_and_sql, ' ' event_counts

from dual

) sql_key

----------------------------------------

union all

--#5: Key footer.

select 3000000 interval_number, '==============================================================================================================================================================================================================================================================================' chart from dual

)

order by interval_number;