Monday, February 28, 2011

How to ... redo logs generation?

About Me

All around Internet there are many articles which says how to narrow redo log generation in Oracle. However I still found out my self in position to ask what is the real cause of redo log
generation. In this post I'll try to answer all aspects of the cause redo log generation. This subject is more interesting for those who are running stand by databases which redo log transport on
remote location may caused many problems.

Damir Vadas
Who is generating redo logs now?
LinkedIn profile
One of the first question, which cover firefighter is: "Who is generating redo logs in this moment". Idea is to determine action on ad hoc basis. For that I use following script, which I call oDesk Cert. Contractor
top_redo.sql: View my complete profile

/* ----------------------------------------------------------------------------- ?
Filename: top_redo.sql

CR/TR# : Search This

Purpose : Shows current redo logs generation info (RAC-non RAC environment) Blog
Date : 12.08.2008.
Search
Author : Damir Vadas, [email protected]

Remarks : run as privileged user

Changes (DD.MM.YYYY, Name, CR/TR#):

--------------------------------------------------------------------------- */
col machine for a15
col username for a10
col redo_MB for 999G990 heading "Redo |Size MB"
column sid_serial for a13;

select b.inst_id,
lpad((b.SID || ',' || lpad(b.serial#,5)),11) sid_serial,
b.username,
machine,
b.osuser,
b.status,
a.redo_mb Followers
from (select n.inst_id, sid,
 round(value/1024/1024) redo_mb
from gv$statname n, gv$sesstat s Followers (28) Next
where n.inst_id=s.inst_id
and n.name = 'redo size'
and s.statistic# = n.statistic#
order by value desc
) a, Follow by
gv$session b Email
where b.inst_id=a.inst_id
and a.sid = b.sid
and rownum <= 30 Email address Submit
;

PROMPT Top 30 from gv$sesstat view according generated redo logs

Result is something like:

Popular Posts
SQL> @top_redo ?
Redo The macro cannot be found or
INST_ID SID_SERIAL USERNAME MACHINE OSUSER STATUS Size MB has been disabled (BI Publisher
---------- ------------- ---------- --------------- ------------------------------ -------- -------- and Office 2010)
1 788, 1 iis1 oracle ACTIVE 2,073 Recently I have upgraded MS
4 788, 1 iis4 oracle ACTIVE 1,928 Office from 2007 to 2010. As I
1 792, 1 iis1 oracle ACTIVE 1,168 work on some BI projects, part
1 791, 1 iis1 oracle ACTIVE 1,149 of mine set of oracle programs
is oracle BI Publis...
3 788, 1 iis3 oracle ACTIVE 1,111
4 792, 1 iis4 oracle ACTIVE 1,092 Retrieve Oracle password from
1 785, 1 iis1 oracle ACTIVE 1,064 Toad for Oracle
4 791, 1 iis4 oracle ACTIVE 1,064 One of the oldest feature Dell
3 792, 1 iis3 oracle ACTIVE 757 Toad has is saving login
3 791, 1 iis3 oracle ACTIVE 738 passwords. This is accomplish
3 785, 1 iis3 oracle ACTIVE 436 easy with enabling check box
4 785, 1 iis4 oracle ACTIVE 411 "Save passwords"...
1 764, 4 SYS iis1 oracle ACTIVE 340
1 737,61477 DBSNMP iis1 oracle ACTIVE 117 How to ... redo logs generation?
3 703,33361 DBSNMP iis3 oracle ACTIVE 113 All around Internet there are
4 677,30159 DBSNMP iis4 oracle ACTIVE 86 many articles which says how
4 795, 1 iis4 oracle ACTIVE 81 to narrow redo log generation in
1 795, 1 iis1 oracle ACTIVE 77 Oracle. However I still found out
4 794, 1 iis4 oracle ACTIVE 76 my self in pos...
3 795, 1 iis3 oracle ACTIVE 75
1 794, 1 iis1 oracle ACTIVE 74 How to see index usage
(without "alter index ...
3 794, 1 iis3 oracle ACTIVE 70
monitoring usage")
1 645, 5393 ANPI USR\APINTARIC apintaric INACTIVE 50
Recent Oracle online course
3 758, 5 iis3 oracle ACTIVE 39
"Understanding Explain Plans &
3 755, 2 iis3 oracle ACTIVE 24 Index Utilization" by Dan Hotka
3 754, 2 iis3 oracle ACTIVE 22 , reminded me to write some
3 756, 2 iis3 oracle ACTIVE 21 thought...
3 757, 3 iis3 oracle ACTIVE 21
4 774, 5835 JAGO CLT\JGOLUZA jgoluza INACTIVE 10 How to find correct SCN?
1 619,61303 LIMI NIO\LMIHALIC lmihalic INACTIVE 9 Recent online help to certain
Mr. Muthu, who had problems
30 rows selected. with database recover, force me
to write this blog topic. Even
Top 30 from gv$sesstat view according generated redo logs thought there are m...
SQL>

If you want to concentrate on real oracle users (avoid core Oracle processes in result) place next condition in outer where clause:

and b.username is not null ? Blog Archive

December 2018 ( 1 )

November 2017 ( 2 )
When and how many redo logs generation occurred?
October 2016 ( 1 )
Beside current analyze in many times wider analyze/compare is even more interesting. So questions like:
When do we have most of redo log generation? September 2015 ( 1 )
 Where was the peak of log generation? August 2015 ( 4 )
Did we have any "strange" redo log generation? July 2014 ( 1 )

need a little different approach-query v$log_history view. It holds historic data which retention period is initially controlled with MAXLOGHISTORY, defined while creating database (fixed not March 2014 ( 1 )
changeable without recreation of control file) and CONTROL_FILE_RECORD_KEEP_TIME which is changeable. In mine case it was set to 31 days (exact number of days for longest month): October 2013 ( 1 )
September 2013 ( 4 )
SQL> show parameter CONTROL_FILE_RECORD_KEEP_TIME ?
July 2013 ( 2 )
NAME_COL_PLUS_SHOW_PARAM TYPE VALUE_COL_PLUS_SHOW_PARAM
June 2013 ( 2 )
------------------------------ ----------- ---------------------------------------------
control_file_record_keep_time integer 31 February 2013 ( 1 )
January 2013 ( 2 )
Script to gather data through mentioned period looks like.
December 2012 ( 1 )
/* ----------------------------------------------------------------------------- ?
November 2012 ( 1 )
Filename: rl.sql
October 2012 ( 4 )
CR/TR# :
September 2012 ( 3 )
Purpose : redo logs distribution per hours on each day ... July 2012 ( 4 )
June 2012 ( 2 )
Date : 07.08.2007.
Author : Damir Vadas, [email protected] May 2012 ( 4 )
April 2012 ( 3 )
Remarks : run as privileged user
March 2012 ( 2 )
Changes (DD.MM.YYYY, Name, CR/TR#):
February 2012 ( 3 )
--------------------------------------------------------------------------- */
set pagesize 120; January 2012 ( 3 )
set linesize 200;
col day for a8; December 2011 ( 3 )
spool rl.txt November 2011 ( 1 )
PROMPT Archive log distribution per hours on each day ...
October 2011 ( 4 )
select September 2011 ( 3 )
to_char(first_time,'YY-MM-DD') day,
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'00',1,0)),'999') "00", August 2011 ( 1 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'01',1,0)),'999') "01", July 2011 ( 2 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'02',1,0)),'999') "02",
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'03',1,0)),'999') "03", April 2011 ( 3 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'04',1,0)),'999') "04", March 2011 ( 4 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'05',1,0)),'999') "05",
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'06',1,0)),'999') "06", February 2011 ( 4 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'07',1,0)),'999') "07", January 2011 ( 6 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'08',1,0)),'999') "08",
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'09',1,0)),'999') "09", December 2010 ( 2 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'10',1,0)),'999') "10", November 2010 ( 4 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'11',1,0)),'999') "11",
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'12',1,0)),'999') "12", September 2010 ( 4 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'13',1,0)),'999') "13", August 2010 ( 5 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'14',1,0)),'999') "14",
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'15',1,0)),'999') "15", July 2010 ( 4 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'16',1,0)),'999') "16", June 2010 ( 7 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'17',1,0)),'999') "17",
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'18',1,0)),'999') "18", May 2010 ( 5 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'19',1,0)),'999') "19", February 2010 ( 3 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'20',1,0)),'999') "20",
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'21',1,0)),'999') "21", January 2010 ( 13 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'22',1,0)),'999') "22",
December 2009 ( 8 )
to_char(sum(decode(substr(to_char(first_time,'HH24'),1,2),'23',1,0)),'999') "23",
COUNT(*) TOT November 2009 ( 6 )
from v$log_history
October 2009 ( 6 )
group by to_char(first_time,'YY-MM-DD')
order by day
;

Result looks like:

 SQL>@rl
Archive log distribution per hours on each day ... My Blog List
DAY 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 TOT
-------- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- -----
11-01-14 0 0 23 16 17 16 16 16 22 39 23 18 22 18 18 18 22 18 19 16 19 16 16 17 425 Dimitri Gielis Blog
11-01-15 24 23 17 18 21 16 16 16 22 18 20 19 19 18 18 21 20 18 20 16 18 16 16 17 447 (Oracle Application
11-01-16 40 39 43 24 17 16 16 16 22 18 18 21 21 18 19 19 22 18 19 17 18 16 16 16 509
11-01-17 25 23 18 16 17 16 16 16 22 18 21 18 20 18 18 18 24 18 18 17 20 16 16 17 446 Express - APEX)
11-01-18 25 23 21 16 18 17 16 17 21 43 18 19 26 18 20 40 21 18 20 16 18 16 16 17 500 Free Oracle Cloud: 15. The
11-01-19 24 24 22 16 17 16 16 16 23 18 19 18 23 19 18 19 19 18 22 16 21 16 16 17 453
11-01-20 24 24 19 16 17 16 16 16 22 18 19 18 24 18 18 19 70 18 19 18 17 16 16 17 495 request could not be
11-01-21 24 23 20 16 17 16 16 16 22 18 18 20 48 25 46 57 22 18 18 17 20 16 16 17 546 mapped to any database
11-01-22 24 27 22 16 17 16 16 16 22 18 18 19 21 18 18 21 20 18 18 17 20 16 16 17 451 1 week ago
11-01-23 24 19 19 16 17 16 16 16 22 18 18 18 21 18 20 18 21 18 20 17 18 16 16 17 439
11-01-24 24 21 19 16 17 16 16 16 22 18 20 18 21 18 21 18 21 18 19 17 17 16 16 17 442
11-01-25 24 23 18 16 17 16 16 16 43 18 18 18 25 18 115 41 23 18 19 17 20 16 16 18 589
11-01-26 23 22 19 16 17 16 16 16 22 18 21 18 30 18 43 18 20 18 22 16 21 16 16 16 478 Johns Blog
11-01-27 21 23 23 16 17 16 16 16 22 18 18 20 22 20 40 97 155 145 155 93 109 116 164 214 1556 utPLSQL integration with
11-01-28 93 24 23 16 17 16 16 16 29 18 19 19 30 101 19 18 75 30 19 16 20 16 22 16 688 Oracle SQLcl
11-01-29 21 16 16 16 17 16 27 20 22 18 18 18 31 18 18 19 23 18 19 18 18 16 16 17 456
11-01-30 24 22 17 16 17 16 16 16 22 18 18 19 23 18 19 18 23 18 18 17 20 16 16 17 444 6 months ago
11-01-31 24 19 21 16 17 16 16 16 23 18 19 18 22 19 18 20 20 18 18 16 21 16 16 17 444
11-02-01 24 25 22 16 17 16 16 16 23 18 18 20 24 18 18 19 21 18 18 17 19 16 16 17 452
11-02-02 24 22 20 16 17 16 16 16 23 18 18 40 23 18 39 18 21 18 22 16 21 16 16 17 491 Oracle SQL Tuning
11-02-03 24 23 22 16 18 16 16 16 22 18 19 18 24 18 20 18 20 18 19 18 19 16 16 17 451
11-02-04 24 23 23 16 17 16 16 16 22 18 19 19 48 18 20 20 23 18 18 17 19 16 16 17 479 Notepad
11-02-05 24 23 17 17 17 16 16 16 22 18 20 19 23 18 19 20 22 18 18 17 20 16 16 17 449 Query to get SQL_ID from
11-02-06 24 23 17 16 18 16 16 16 22 18 19 19 21 18 19 18 24 18 20 16 21 16 16 17 448 DBA_SQL_PLAN_BASELIN
11-02-07 24 22 18 16 18 16 16 16 22 18 52 18 44 25 29 24 21 18 18 16 19 16 16 17 519
11-02-08 24 23 19 16 17 16 16 16 22 18 20 19 28 18 19 18 22 18 19 17 34 32 32 31 514 ES
11-02-09 36 39 35 31 37 31 32 32 32 34 34 34 39 34 34 34 36 34 38 32 37 31 32 33 821 10 months ago
11-02-10 37 36 38 31 33 32 31 32 37 34 34 34 41 34 35 35 37 34 34 31 34 32 31 33 820
11-02-11 39 35 38 31 33 32 31 32 37 34 34 34 38 34 34 34 35 34 34 32 34 32 31 32 814
11-02-12 40 34 34 33 35 32 31 32 37 34 34 34 36 34 35 34 36 34 34 31 36 32 31 32 815
11-02-13 40 34 35 31 34 32 31 32 37 34 34 34 38 34 34 34 37 34 34 32 36 32 31 32 816 Marko Sutic's Oracle
11-02-14 40 34 33 31 35 32 32 32 37 66 34 35 38 34 34 34 37 34 34 31 36 32 31 33 849 Blog
11-02-15 48 50 50 44 48 32 32 32 35 34 34 34 38 34 37 34 39 34 34 32 36 31 32 33 887
11-02-16 39 34 33 32 33 31 32 32 37 53 37 34 38 38 34 2 0 0 0 0 0 0 0 0 539 Galera Cluster Schema
Changes, Row Based
34 rows selected. Replication and Data
SQL> Inconsistency
1 year ago

Redo logs generation is grouped by hours where last column (TOT) is sum of all redo logs in one day. According this it is more then obvious where redo log generation was highest, so our interest
may be focused on presented point in time. Charles Hooper's Oracle
Notes
Presentation – Working with
How much is that in Mb? Oracle Database in VB.Net
with ODP.Net and Visual
Total redo logs size (and according that, archived log size) cannot be computed from previous query because not all redo log switches occur when redo log was full. For that you might want to use API Toolkit; Enhancing
this very easy query: Visual Manufacturing 9.0.0
2 years ago
SQL> select sum(value)/1048576 redo_MB from sys.gv_$sysstat where name = 'redo size'; ?
Dan McGhan's Oracle
REDO_MB
Blog
----------
Tutorial: Creating a
1074623.75 Component Similar to Super
LOV Using Native
SQL> Components
3 years ago
If you want to calculate on instance grouping, then use this:
Coskan's Approach to
SQL> select inst_id, sum(value)/1048576 redo_MB from sys.gv_$sysstat where name = 'redo size' ?
2 group by inst_id; Oracle
oratop from MOS
6 years ago
INST_ID REDO_MB
---------- ----------
1 370325.298
2 4712.567
4 405129.283
3 294457.100

SQL>

Both queries works on single instances as well.

Which segments are generating redo logs?
After we found out our point of interest, in mine case where were most of the redo logs generation, it is very useful to find out which segments (not tables only) are causing redo log generation.
For that we need to use "dba_hist" based tables, part of "Oracle AWR (Automated Workload Repository)", which usage I have described in topic Automated AWR reports in Oracle 10g/11g. For
this example I'll focus on data based on time period: 11-01-28 13:00-11-01-28 14:00. Query for such a task should be:

SELECT to_char(begin_interval_time,'YY-MM-DD HH24') snap_time, ?

dhso.object_name,
sum(db_block_changes_delta) BLOCK_CHANGED
FROM dba_hist_seg_stat dhss,
dba_hist_seg_stat_obj dhso,
dba_hist_snapshot dhs
WHERE dhs.snap_id = dhss.snap_id
AND dhs.instance_number = dhss.instance_number
AND dhss.obj# = dhso.obj#
AND dhss.dataobj# = dhso.dataobj#
AND begin_interval_time BETWEEN to_date('11-01-28 13:00','YY-MM-DD HH24:MI')
AND to_date('11-01-28 14:00','YY-MM-DD HH24:MI')
GROUP BY to_char(begin_interval_time,'YY-MM-DD HH24'),
dhso.object_name
HAVING sum(db_block_changes_delta) > 0
ORDER BY sum(db_block_changes_delta) desc ;

Reduced result from previously shown query would be:

SNAP_TIME OBJECT_NAME BLOCK_CHANGED ?

----------- ------------------------------ -------------
11-01-28 13 USR_RACUNI_MV 1410112
11-01-28 13 TROK_TAB_RESEAU_I 734592
11-01-28 13 TROK_VOIE_I 638496
11-01-28 13 TROK_DATUM_ULAZA_I 434688
11-01-28 13 TROK_PAIEMENT_I 428544
11-01-28 13 D_DPX_VP_RAD 351760
11-01-28 13 TROK_SVE_OK_I 161472
11-01-28 13 I_DATPBZ_S002 135296
11-01-28 13 IDS2_DATUM_I 129904
11-01-28 13 IDS2_PZNBR 129632
11-01-28 13 IDS2_IDS1_FK_I 128848
11-01-28 13 IDS2_DATTRAN_I 127440
11-01-28 13 IDS2_DATSOC_I 127152
11-01-28 13 IDS2_VRSTA_PROD_I 122816
...

Let us focus on first segment "USR_RACUNI_MV", segment with highest number of changed blocks (what mean directly highest redo log generation). Just for information, this is
MATERIALIZED VIEW.

What SQL was causing redo log generation

Now when we know when, how much and what, time is to find out how redo logs are generated. In next query "USR_RACUNI_MV" and mentioned period are hard codded, because we are
focused on them. Just to point that SQL that start with "SELECT" are not point of our interest because they do not make any changes.

SELECT to_char(begin_interval_time,'YYYY_MM_DD HH24') WHEN, ?

dbms_lob.substr(sql_text,4000,1) SQL,
dhss.instance_number INST_ID,
dhss.sql_id,
executions_delta exec_delta,
rows_processed_delta rows_proc_delta
FROM dba_hist_sqlstat dhss,
dba_hist_snapshot dhs,
dba_hist_sqltext dhst
WHERE upper(dhst.sql_text) LIKE '%USR_RACUNI_MV%'
AND ltrim(upper(dhst.sql_text)) NOT LIKE 'SELECT%'
AND dhss.snap_id=dhs.snap_id
 AND dhss.instance_number=dhs.instance_number
AND dhss.sql_id=dhst.sql_id
AND begin_interval_time BETWEEN to_date('11-01-28 13:00','YY-MM-DD HH24:MI')
AND to_date('11-01-28 14:00','YY-MM-DD HH24:MI')
;

Result is like:

WHEN SQL inst_id sql_id exec_delta rows_proc_delta ?

------------- ------------------------------------------------- ------- ------------- ---------- ---------------
2011_01_28 13 DECLARE 1 duwxbg5d1dw0q 0 0
job BINARY_INTEGER := :job;
next_date DATE := :mydate;
broken BOOLEAN := FALSE;
BEGIN
dbms_refresh.refresh('"TAB"."USR_RACUNI_MV"');
:mydate := next_date;
IF broken THEN :b := 1;
ELSE :b := 0;
END IF;
END;
2011_01_28 13 delete from "TAB"."USR_RACUNI_MV" 1 5n375fxu0uv89 0 0

For both of examples it was impossible to find out number of rows changed according operation that was performed. Let us see output of another example (NC_TRANSACTION_OK_T table)
where we can meet with DDL that generate redo logs!

WHEN SQL inst_id sql_id exec_delta rows_proc_delta ?

------------- ------------------------------------------------- ------- ------------- ---------- ---------------
2011_01_28 13 alter table TAB.NC_TRANSACTION_OK_T 4 g5gvacc8ngnb8 0 0
shrink space cascade

If you are focused on pure number of changes, then you might to perform query where inst_id and sql_id are irrelevant (excluded from query). Here is a little modified previous example, for
"Z_PLACENO" segment (pure oracle table):

SELECT when, sql, SUM(sx) executions, sum (sd) rows_processed ?

FROM (
SELECT to_char(begin_interval_time,'YYYY_MM_DD HH24') when,
dbms_lob.substr(sql_text,4000,1) sql,
dhss.instance_number inst_id,
dhss.sql_id,
sum(executions_delta) exec_delta,
sum(rows_processed_delta) rows_proc_delta
FROM dba_hist_sqlstat dhss,
dba_hist_snapshot dhs,
dba_hist_sqltext dhst
WHERE upper(dhst.sql_text) LIKE '%Z_PLACENO%'
AND ltrim(upper(dhst.sql_text)) NOT LIKE 'SELECT%'
AND dhss.snap_id=dhs.snap_id
AND dhss.instance_Number=dhs.instance_number
AND dhss.sql_id = dhst.sql_id
AND begin_interval_time BETWEEN to_date('11-01-25 14:00','YY-MM-DD HH24:MI')
AND to_date('11-01-25 15:00','YY-MM-DD HH24:MI')
GROUP BY to_char(begin_interval_time,'YYYY_MM_DD HH24'),
dbms_lob.substr(sql_text,4000,1),
dhss.instance_number,
dhss.sql_id
)
group by when, sql;

Result is like:

WHEN SQL exec_delta rows_proc_delta ?

------------- ---------------------------------------------------------------------- ---------- ---------------
2011_01_25 14 DELETE FROM Z_PLACENO 4 7250031
2011_01_25 14 INSERT INTO Z_PLACENO(OBP_ID,MT_SIFRA,A_TOT) 4 7250830
SELECT P.OBP_ID,P.MT_SIFRA,SUM(P.OSNOVICA)
FROM (SELECT OPI.OBP_ID,
 OPO.MT_SIFRA,
SUM(OPO.IZNKN) OSNOVICA
WHERE OPI.OBP_ID = OPO.OPI_OBP_ID
AND OPI.RBR = OPO.OPI_RBR
AND NVL(OPI.S_PRETPOREZA,'O') IN ( 'O','N','A','Z','S')
GROUP BY OPI.OBP_ID,OPO.MT_SIFRA
)

Here you can see directly number executions and number of involved rows.

Query based on segment directly

Sometimes you do not want to focus on period, so your investigation may start with segment as starting point. For such a tasks I use next query. This is small variation of previous example where
"USR_RACUNI_MV" segment is hard codded.

SELECT to_char(begin_interval_time,'YY-MM-DD HH24') snap_time, ?

sum(db_block_changes_delta)
FROM dba_hist_seg_stat dhss,
dba_hist_seg_stat_obj dhso,
dba_hist_snapshot dhs
WHERE dhs.snap_id = dhss.snap_id
AND dhs.instance_number = dhss.instance_number
AND dhss.obj# = dhso.obj#
AND dhss.dataobj# = dhso.dataobj#
AND dhso.object_name = 'USR_RACUNI_MV'
GROUP BY to_char(begin_interval_time,'YY-MM-DD HH24')
ORDER BY to_char(begin_interval_time,'YY-MM-DD HH24');

Reduced result is:

SNAP_TIME SUM(DB_BLOCK_CHANGES_DELTA) ?
----------- ---------------------------
...
11-01-28 11 1224240
11-01-28 12 702880
11-01-28 13 1410112
11-01-28 14 806416
11-01-28 15 2008912
11-01-28 16 1103648
...

As you can see in accented row, the numbers are the same as at the begging of topic.

The End
Regardless which approach you use, redo logs generation has important effects in proper management of Oracle RMAN backups as well as stand by databases.

Cheers!

at Monday, February 28, 2011

60 comments :

Anonymous March 12, 2011 at 1:43 AM

Regardless there was many topic on this subject, this one seems to be most complete in covering the whole subject at once. Nice work.

Reply
 Replies

Anonymous March 4, 2015 at 2:02 AM

100% agree.
Great work,

Thanks,
zafar

Reply

Anonymous June 8, 2011 at 4:02 PM

Great Job!!!! Thanks

Reply

Damir Vadas (aka Funky) June 8, 2011 at 8:30 PM

Thank you all!
Reply

Anonymous June 23, 2011 at 9:50 PM

Excellent Men!!!.... thank you very much!!!!....
Reply

Ray September 16, 2011 at 3:39 AM

Yeaaahhh, Excellent topic.

Thx !
Reply

Damir Vadas (aka Funky) September 16, 2011 at 4:20 AM

Glad to announce that this subject will be one of mine two seminars on HROUG 2011, Croatian Oracle Conference Group.
Cheers!
Reply

Bob November 14, 2011 at 6:08 PM

thanx ,very helpful

Reply
Damir Vadas November 14, 2011 at 7:56 PM
Glad to help you Bob!
Reply

Anonymous December 4, 2011 at 8:13 AM

Great work.very Useful scripts.
Reply

Damir Vadas December 4, 2011 at 12:06 PM

Glad to help you.
Reply

Jyothish December 16, 2011 at 1:51 PM

Thank you very much Damir

Reply

Anonymous December 21, 2011 at 3:14 PM

Great and very useful scripts
Reply

Anonymous April 18, 2012 at 5:30 PM

Just came across this post, this is an excellent set of scripts for monitoring redo generation.
thanks!! Darryl
Reply

Replies

Damir Vadas April 18, 2012 at 8:22 PM

Glad to help you ...
Damir

Reply

Anonymous April 20, 2012 at 3:50 AM

Hi Damir

Is using V$log_history ok even in RAC or should we use gv$log_history?

Many thanks for great post!

S
Reply

Replies

Damir Vadas April 20, 2012 at 9:29 AM

Hi Anonymous,

querying v$log_history will give you result based on instance you are logged in.
querying gv$log_history will give you result based whole RAC (all instances). Look in this RAC example:
[code]
SQL> select count(*) from v$log_history;

COUNT(*)
----------
39944

1 row selected.

Elapsed: 00:00:02.46
SQL> select count(*) from gv$log_history;

COUNT(*)
----------
159776

1 row selected.

Elapsed: 00:00:00.56
SQL>
[/code]

Reply

Prise Mike May 28, 2012 at 8:28 AM

Thank you Damir.

Indeed the most complete work I've seen on the subject.

Although I knew which was the Redo generator, the complete work you've done here helped me indeed.

Thanks again,
Mike
Reply

newbie01.oracle May 29, 2012 at 7:26 AM

Hi,

Very good post. Just have one question though.

Is it safe to assume whichever session/process is causing the REDOLOG, this is the same one that is generating the FLASHBACK log ?

Thanks.
Reply

Damir Vadas May 29, 2012 at 7:54 AM

Correct.
Flash back logs are pure redologs just saved in another place for online restore/recover.
Reply

Anonymous July 31, 2012 at 2:30 PM

Heу very interеsting blog!
My web blog - Authentic Talisman
Reply

Anonymous October 2, 2012 at 8:43 PM

Thank you so much compiling such a useful list of scripts and for explaining their usefulness that both the technical and business managers can both understand and appreciate.

Great job here.

Reply

Replies

Damir Vadas October 2, 2012 at 8:57 PM

Glad to help you.

Always happy when breaking distance between tech and managers ...
:-)

Reply

Anonymous October 20, 2012 at 8:11 AM

gv$log_history vs v$log_history
GV$ Always gives the double(11g R1)

If i query v$ from both nodes, it shows the same REDO Logs.

Even in RAC, should we use v$log_history instead of gv$ ??

Reply

Replies

Damir Vadas October 20, 2012 at 2:43 PM

No not at all!
Please look in example posted on "April 20, 2012 9:29 AM" on this page.

Reply

Rob October 23, 2012 at 3:29 PM

My friend, a really awesome post. Helped me with a work issue. cheers and thanks, rob
Reply

Replies

Damir Vadas February 5, 2013 at 7:16 PM

You are welcome

Reply

Shaan December 4, 2012 at 2:05 PM

Super article, it would have been nice if you could please share historical redo generation using dba_hist tables
Reply

Replies

Damir Vadas February 5, 2013 at 7:19 PM

Do not understand what is dba_hist tables view?

Reply

Unknown January 23, 2013 at 9:56 AM

woww ... Damir you made is soo easy .. thanks ..

can you provide me the script to capture the no of redos generated per hour and their size per hour in the last one month ..

Thanks ,
Sameer
Reply

Replies
 Damir Vadas February 5, 2013 at 7:17 PM

It is already there, look in "rl.sql" script.

Reply

Anonymous February 5, 2013 at 11:36 AM

Great Job , thanks for sharing
Reply

Replies

Damir Vadas February 5, 2013 at 7:16 PM

You are welcome

Reply

Anonymous March 6, 2013 at 2:21 PM

Hi Damir

This is a very useful document and i am following this document for a long time to get the information of redo generation, currently i am struggling with FND_LOB table and SYSLOB
segment, there are high redo generation on these tables by when i check the attachments they are hardly 2 to 3 MB attachment uploaded during high redo generation but it created 400 to
500 archivelog file that is very strange.

Can you please help me troubleshoot the issue, any queries related to this will help

Thank you

Regards
shahrukh
Reply

Anonymous March 9, 2013 at 4:39 AM

You nailed it. Excellent

Thank
Reply

Anonymous June 26, 2013 at 1:59 PM

Does "redo size" equate or proportioonal to db_block_changes_delta for all database operations including NOLOGGING operations?
Reply

Replies
 Damir Vadas June 26, 2013 at 2:02 PM
Do not know. Let me know if you find out.

BTW, it depends on Oracle version.

Reply

Anonymous July 3, 2013 at 12:11 PM

very GREat
Reply

Unknown July 15, 2013 at 7:00 AM

Hats off....Great Work..Really Helpful...

Reply

Replies

Damir Vadas July 15, 2013 at 4:46 PM

THX

Reply

Unknown August 27, 2013 at 6:07 PM

Great Work
Reply

Thomas Hu September 11, 2013 at 2:20 AM

GREAT WORK! It helps me a lot.

Is there problem with this script:

If you are focused on pure number of changes, then you might to perform query where inst_id and sql_id are irrelevant (excluded from query). Here is a little modified previous example,
for "Z_PLACENO" segment (pure oracle table):
SELECT when, sql, SUM(sx) executions, sum (sd) rows_processed
...
Reply

Replies

Damir Vadas September 11, 2013 at 3:24 AM

 Please explain a little bit more ... do not follow you.

Thomas Hu September 11, 2013 at 6:58 AM

If you run this query, it can't find 'sx' or 'sd' column.

Damir Vadas December 10, 2013 at 6:06 PM

Yap,
Change first line of query to:
SELECT when, sql, SUM(exec_delta) executions, sum (rows_proc_delta) rows_processed

Hpe now is ok....

Mohammadimran Uddin December 28, 2016 at 6:00 PM

Hi Damir,
Thanks a lot for this queries. The query SELECT when, sql, SUM(exec_delta) executions, sum (rows_proc_delta) rows_processed .....(full query) when i execute it is getting
executed but not giving any output. What might be the reason? The database is 11.2.0.4.

Regards,
Imran

Reply

newbie01.oracle December 10, 2013 at 5:28 AM

Hi Damir,

Your blogs are now officially my favorite. Would be nice if your scripts can be downloaded in one zip file though. :-) [ wishful thinking ]

Anyway, can you please advise how to modify the rl.sql script so that it shows DAY as the header and HOUR as the column.

Example as below:

HOUR 2011-01-14 2011-01-15 .... and so on

00 0 24
01 0 23
02
... and so on

Reply

Replies

Damir Vadas December 10, 2013 at 6:03 PM

this will be rows to columns (or crosstab) query... so grep on google for such an olution
 Reply

Anonymous April 22, 2014 at 12:09 PM

Great, thank you!

Reply

Anonymous October 29, 2014 at 7:06 PM

thanks

Reply

Mohammadkhalilullah Balsing December 11, 2014 at 4:24 PM

Dear Damir,

excellent work

please let me know %Z_PLACENO% in query

Reply

Replies

Damir Vadas December 12, 2014 at 3:41 PM

It is explained ... with it's sql ... just grep by thius string

Reply

Anonymous February 4, 2015 at 9:18 AM

Really cool scripts.
Shared with all of my friends.
Reply

Anonymous March 7, 2015 at 2:33 AM

Hi Damir,
Again its very good work .Looks like you need to make some correction Inbelow query which i have taken from your above post. Please look at first line , looks like the variables used are
not what inner query refer.(SUM(sx) executions, sum (sd) rows_processed )

SELECT when, sql, SUM(sx) executions, sum (sd) rows_processed

FROM (
SELECT to_char(begin_interval_time,'YYYY_MM_DD HH24') when,
dbms_lob.substr(sql_text,4000,1) sql,
dhss.instance_number inst_id,
dhss.sql_id,
sum(executions_delta) exec_delta,
sum(rows_processed_delta) rows_proc_delta
FROM dba_hist_sqlstat dhss,
dba_hist_snapshot dhs,
dba_hist_sqltext dhst
WHERE upper(dhst.sql_text) LIKE '%Z_PLACENO%'
AND ltrim(upper(dhst.sql_text)) NOT LIKE 'SELECT%'
AND dhss.snap_id=dhs.snap_id
AND dhss.instance_Number=dhs.instance_number
AND dhss.sql_id = dhst.sql_id
AND begin_interval_time BETWEEN to_date('11-01-25 14:00','YY-MM-DD HH24:MI')
AND to_date('11-01-25 15:00','YY-MM-DD HH24:MI')
GROUP BY to_char(begin_interval_time,'YYYY_MM_DD HH24'),
dbms_lob.substr(sql_text,4000,1),
dhss.instance_number,
dhss.sql_id
)
group by when, sql;

Thanks
Zafar
Reply

mansoorm December 22, 2015 at 6:41 PM

thank you ,excellent work

Reply

Anonymous June 6, 2017 at 12:01 AM

Great work and thanks for sharing with us.
Reply

Marco Marobin September 27, 2017 at 9:36 AM

Excellent scripts, very good work.
Reply

Replies

Damir Vadas September 27, 2017 at 9:43 AM

thx

Reply
 Anonymous September 19, 2019 at 6:04 PM
This is the best redo/archivelog investigation manual!

Reply

Replies

Damir Vadas September 19, 2019 at 6:27 PM

Thank you

Reply

Enter your comment...

Comment as: Google Accoun

Publish Preview

Links to this post

Create a Link

Newer Post Home Older Post

Subscribe to: Post Comments ( Atom )

Zagreb u srcu!

Copyright © 2009-2018 Damir Vadas

All rights reserved.

Sign by Danasoft - Get Your Sign