Tales From A Lazy Fat DBA

Its all about Databases & their performance, troubleshooting & much more …. ¯\_(ツ)_/¯

Archive for January, 2018

Visualize your database performance statistics using Tableau.

Posted by FatDBA on January 28, 2018

Hi Everyone,

Today’s post is all about producing some interactive data visualizations your system/database statistics using one of the most popular BI tool Tableau.
There are lot’s of other tools which can be used for your data analysis i.e. Excel, Microstrategy Analytics, Domo, QlikView but i always find Tableau easiest way to do such things, specially during all my database audit, 360 health reviews and troubleshooting task-forces.

Tableau offers a suite of tools that include an online, desktop and server version. All of these versions provide a easy-to-use drag and drop interface that can help you quickly turn your data into business insights. Like many other data analytics and visualization tools, Tableau can connect to local or remote data of many different formats.

Okay now after that short introduction of the tool, time to do some tests using the tool.
I have divided process in to three step activity and are discussed below.

First: Data Collection
You can collect your AWR reports in TEXT format and which will be later on parsed to create a CSV file.
There are many tools/scripts available online to generate multiple AWR reports of your database. I see an awesome work was already done by FlashDBA (Download). You can use his script to generate batch AWR reports and yes in TEXT format ONLY!

Example:

[oracle@dixitlab AWR]$ ls -ltr
total 12852
-rw-r--r--. 1 oracle oinstall 225031 Jan 27 21:25 awrrpt_1_445_446.txt
-rw-r--r--. 1 oracle oinstall 255010 Jan 27 21:26 awrrpt_1_446_447.txt
-rw-r--r--. 1 oracle oinstall 253217 Jan 27 21:26 awrrpt_1_447_448.txt
-rw-r--r--. 1 oracle oinstall 253217 Jan 27 21:26 awrrpt_1_448_449.txt
-rw-r--r--. 1 oracle oinstall 244229 Jan 27 21:26 awrrpt_1_449_450.txt
........
.................

Second: Data Fold or Compression.
Now next you need a tool to fold your AWR reports in to a CSV. Here once again FlashDBA did a marvelous job, he wrote one fabulous script to parse your AWR text files and generate a final AWR report.
You can download the script from his Github (Download Link)

As far as the script, you need to pass the format of your files and direct output to a CSV as shown below.

[oracle@dixitlab AWR]$ ./awr-parser.sh awr*.txt > tunedbperftests.csv

Info : Parsing file awrrpt_1_445_446.txt at 2018-01-27 21:32:49
Info : Parsing file awrrpt_1_446_447.txt at 2018-01-27 21:32:53
Info : Parsing file awrrpt_1_447_448.txt at 2018-01-27 21:32:56
Info : Parsing file awrrpt_1_448_449.txt at 2018-01-27 21:33:01
Info : Parsing file awrrpt_1_449_450.txt at 2018-01-27 21:33:07
Info : Parsing file awrrpt_1_450_451.txt at 2018-01-27 21:33:15
Info : Parsing file awrrpt_1_451_452.txt at 2018-01-27 21:33:21
....
........
Info : Parsing file awrrpt_1_499_500.txt at 2018-01-27 21:36:56
Info : No more files found
Info :
Info : ______SUMMARY______
Info : Files found : 55
Info : Files processed : 55
Info : Processing errors : 0
Info :
Info : Completed with 0 errors
[oracle@dixitlab AWR]$

With that you are done with the parsing of reports and have got the final CSV which we will be using to play around within Tableau.
Contents inside the parsed file.

Filename	Database Name	Instance Number	Instance Name	Database Version	Cluster	Hostname	Host OS	Num CPUs	Server Memory (GB)	DB Block Size	Begin Snap	Begin Time	End Snap	End Time	Elapsed Time (mins)	DB Time (mins)	Average Active Sessions	Busy Flag	Logical Reads/sec
awrrpt_1_445_446.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	445	1/26/2018 21:57	446	1/26/2018 23:00	62.24	6.33	0.1	N	2629.5
awrrpt_1_446_447.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	446	1/26/2018 23:00	447	1/27/2018 0:00	60.19	12.18	0.2	N	13973.4
awrrpt_1_447_448.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	447	1/27/2018 0:00	448	1/27/2018 1:00	60.15	13.52	0.2	N	14055.8
awrrpt_1_448_449.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	448	1/27/2018 1:00	449	1/27/2018 2:00	60.15	10.13	0.1	N	11597.4
awrrpt_1_449_450.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	449	1/27/2018 2:00	450	1/27/2018 3:00	60.16	0.03	0	N	65.4
awrrpt_1_450_451.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	450	1/27/2018 3:00	451	1/27/2018 4:00	60.12	0.02	0	N	70.3
awrrpt_1_452_453.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	452	1/27/2018 5:00	453	1/27/2018 6:00	60.13	0.69	0	N	189.8
awrrpt_1_453_454.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	453	1/27/2018 6:00	454	1/27/2018 7:00	60.13	2.88	0	N	2439.1
awrrpt_1_454_455.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	454	1/27/2018 7:00	455	1/27/2018 8:00	60.14	12.57	0.2	N	14027.3
awrrpt_1_455_456.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	455	1/27/2018 8:00	456	1/27/2018 9:00	60.14	10.11	0.1	N	13916.6
awrrpt_1_456_457.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	456	1/27/2018 9:00	457	1/27/2018 10:00	60.14	10.26	0.1	N	13941.5
awrrpt_1_457_458.txt	TUNEDB	PRIMARY	tunedb	EE	1	dixitlab.localdo	Linux x86 64-bit	1	2.77	8K	457	1/27/2018 10:00	458	1/27/2018 10:50	50.25	18.67	0.3	N	14118.9

Third: Data Representation using Tableau.
Okay so here we have the final parsed CSV of all those TEXT AWR reports named ‘tunedbperftests.csv’ and we are ready to play around and learn.

Immediately after launching you will see couple of options available for Data Sources on the left. Choose TEXT as the source and browse the CSV to load.

Next you will see all rows of your data source (tunedbperftests.csv in our case) file.

Next click on Worksheet, your personal area to play.

Tableau then divides the data in two main types: dimensions and measures. Dimensions are usually those fields that cannot be aggregated; measures, as its name suggests, are those fields that can be measured, aggregated, or used for mathematical operations. Dimension fields are usually used for row or column headings; measures are usually used for plotting or giving values to the sizes of markers.

             

Next tab is for Analytical functions, i.e. If you want to add a constant, average, mean, median averages or any reference lines to your graph/chart.

All good now, so we all all set yto plot out performance charts for that DB historical information that we have collected in the form of CSV and loaded to Tableau. Lets, plot for average hard parse per/second, Average DB Time, Average Pareses, Average Transactions happened against Time (Hourly rate of BEGIN TIME as a measure).

So, using above Area Graph you have plotted the average metric usages on the database during a time period.

Next, i will visualize one of the most prominent db wait event observed in the database during the probe (data collection) period ‘DPR’ or ‘Direct Path Reads’ and will plot the Bar graph against the TIME (Hourly BEGIN TIME).

Some more stats visualisations, this time ‘top 5 waits‘ and their hourly frequency.

There are lot’s of other things that you can do with your statistics, i mean you can plot your data in the form of Square, side by side circle, polygon, pie char, polygons, gantt bar, line graph, area graphs, box-and-whisker plots, highlight tables and many more.

So, imagine and you can visualize your database statistics using Tableau!

Questions are welcome. Happy reading! 🙂 🙂

Hope It Helps
Prashant Dixit

Advertisement

Posted in Advanced, troubleshooting | Tagged: , , | Leave a Comment »

AAS or Average Number Of Active Sessions – The first thing to look in an AWR & its Uses.

Posted by FatDBA on January 27, 2018

Hi All,

Today’s post is all about answering the question ‘What is the very first thing that one should look out for in an AWR report ?‘. I have got this question so many times in the past about the first statistic i prefer to look at when troubleshooting a performance problem so though of answering this by writing this post with some real-time examples.

And the answer is ‘AAS‘ or ‘Average Number of Active Sessions’ is the first thing that i always look out for while reading AWR reports because It gives you a quick idea about how busy the system is and about the workload happening.

Okay so first lets understand what is an ‘Active Session’ : In simple words It is a session currently spending time in the database (i.e. from v$session where status=’ACTIVE’).
Now, what exactly is AAS – It’s the ratio or rate of change of DB time over clock time. The value of this metric is calculated by using a standard formula of (DB Time/Elapsed Time).

Lets calculate the value for one of the system.

Host Name	        Platform	       CPUs   
dixitLab1.fatdba.com	Linux x86 64-bit	16

AAS In this case : 1024.72/60.04 => 17.067 of average active sessions during the snap interval of an hour.

Let’s further decode the magical Figures of AAS.
We always use CPU Count as a standard for comparing the AAS. Few rule of thumbs while doing this comparison are give below.
– If the AAS is higher than the number of CPU you have then there is a problem. i.e In above example we have an AAS value of 17 and CPU
count 16, hence we could have performance problems and needs investigation.
– If the value is very high than the number of CPUs then there is a choke-point in the database.

You could also use the AAS to plot your graphs, lines and Manhattan’s as one of the axis to compare it with CPU consumption and quickly pin point the pain areas and time slots. Let understand and use it through a scenario.

Assume one fine day you got a call from monitoring team that they have observed huge spikes in system resource usage and many of the other metrics set on the dashboard are in red. And as usual lot’s of fingers and eyes started pointing towards you and the DBA team.

Now you as a DBA quickly generated the AWR for that specific time frame to understand the system behavior and performance and observed a huge workload is happening on the database with AAS of 305 (For a 2 Node RAC database with 128 CPUs collectively) and some huge peaks for Application class (i.e. row lock contentions etc.), User IO classified waits (i.e. DBF Sequential Reads, read by other session etc.) and some Network class waits (i.e SQL*Net message from dblink waits) in your database.

Now you want to understand the trend for wait classes for the database during last few days. Here you can use DBA_HIST_ACTIVE_SESSION_HISTORY view to collect historical statistics for the database which you will use to plot charts using excel, tableau etc.

I have collected similar stats using ASH view and have plotted a graph using few of my data representation tools to understand this transient variation in system performance.

Here you see a sudden spike in DB wait Classes (Specially User IO, Cluster, Application and Network) on March 5th with average number of active sessions (AAS) stacked for both of RAC nodes was close around 305. Which if compare it with number of total CPUs (64+64=128) is extremely high.

After further investigation you understand that it’s application class wait ‘enq: TX – row lock contention‘ which is the primary cause of this high system resources utilization.

Below graph is a representation of AAS Waiting on Application class event ‘enq: TX – row lock contention’ per Instance on the database where we can the same happening. A constant then a sudden raise in row locking contentions.

And you have identified the major sources contributing towards this row locking during the probe period of last 7 days till now. You can do a join on dba_hist_active_sess_history and dba_hist_snapshot to get this historical information — Read my previous article on how to get this past information from AWR repository.

Now when you have narrowed down the problem and have identified the problematic SQLs with their total contribution, you can now start the query optimization/tuning to fix the issue.
There are lot of other data representations you can do by using AAS as one of the graph axis i.e. AAS on CPU and Top Wait Events and will discuss in my further posts.

Hope It Helps
Prashant Dixit

Posted in Advanced, troubleshooting | Tagged: , , | 3 Comments »

Parse CPU to Parse Elapsd % – Lets clear the Perplex!

Posted by FatDBA on January 26, 2018

Happy Republic Day!

 

Hi Mates,

I see a lot of confusion, mix-ups and perplexity in between the DBAs on few of the metrics under ‘Instance Efficiency % section’ in AWR reports. Specially for one of the metric “Parse CPU to Parse Elapsd %”. In today’s post i will try to explain the metrics in detail which will help you to understand it in depth and clear the muddiness.

Few of the lines you might have read about this metric i.e.
“we should always look for as low as possible numbers for this metric …”
“Try to achieve the impossible value of zero for this one “
– Even some of the Metalink notes are misleading too.

Scenario:

Below is the snippet from one of my test box.

Note: Just taking a look at the instance efficiency ratios can be very dangerous and i advise to first start with Load Profile, top 5 waits and there on …

Okay, so my definition of this statistic differ from what you judge after reading the name of the metric – This datum signals the delay/wait in parsing of SQL queries during the snap interval.
In our example the value is 1.37% this means that for every CPU second spend parsing we spent about 72.99 (100/1.37) Seconds of clock time. It can happen due to various reasons i.e Latch or any contention between the sessions etc.

The ideal value for this stat should also be 100% like rest of the ratios (Yes, that’s correct!!).

Let’s see how this value was calculated, what all it considers while deducing that final figure of 1.37%.
It takes the “parse time cpu/parse time elapsed * 100” to get the ‘Parse CPU to Parse Elapsed’ figure.

 

Statistic	                              Total	per Second     per Trans
------------------------------------------------------------------------------------------------
parse time cpu                                  398	      0.11	0.01
parse time elapsed                           29,055	      8.07	0.81

 

So, in short each time there is a drop in this metric from 100%, means the database was waiting for something which slowed down the parse times.
If you want to dig in deep then you should trace the session using 10046 tracing with level 8 to see where the other % of parse time is being spent and leaving this for readers to test.

 

Hope It Helps
Prashant Dixit

Posted in Advanced, troubleshooting | Tagged: , , | Leave a Comment »

Shared Pool Management in 12c, What’s new!

Posted by FatDBA on January 17, 2018

Hi Everyone,

Few days while working on Shared Pool issue where we were getting ORA-4031 on one of the 11gR2 database i discovered something interesting and new regarding the SGA duration management. Here i wont discuss the problem that i faced and how we fixed but would try to show the architectural changes that has happened with 12c that can now fix these errors/issues.

Let me explain what are ‘Durations’ first. The shared pool is made up of a number of granules. The shared pool then split into sub-pools if you have a large enough SGA, and each sub-pool consists of a number of non-overlapping granules. In 11g each sub-pool also split into four sub-sub-pools known as durations.

What was there before 12c arrived —
Starting from Oracle 10g each sub-pool in SGA was divided in to four durations.
Let’s check the distribution by generating the Heap Dump for shared pool, here i used oradebug with level 2 (This provides you the full SGA Summary or you can try with level 2050 to get full summary with contents).

SQL> oradebug setmypid
Statement processed.
SQL> oradebug dump heapdump 2
Statement processed.
SQL> oradebug tracefile_name
/u01/app/oracle/diag/rdbms/orcl/orcl/trace/orcl_ora_8127.trc

.....
******************************************************
HEAP DUMP heap name="sga heap(1,0)"  desc=380030610
Total heap size    =218102664
Total free space   =  1066928
Total reserved free space   =  8439520
Unpinned space     = 38812528  rcr=11971 trn=17906
Permanent space    =208595160
HEAP DUMP heap name="sga heap(1,1)"  desc=380031e68
Total heap size    = 67108512
Total free space   =  2912528
Total reserved free space   =  1382816
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(1,2)"  desc=3800336c0
Total heap size    =167771280
Total free space   = 92743480
Total reserved free space   =  3852856
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(1,3)"  desc=380034f18
Total heap size    =268434048
Total free space   = 74547592
Total reserved free space   = 13497472
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(2,0)"  desc=380039e38
Total heap size    =201325536
Total free space   =    17200
Total reserved free space   =  8435920
Unpinned space     = 26474112  rcr=7934 trn=8094
Permanent space    =192871456
HEAP DUMP heap name="sga heap(2,1)"  desc=38003b690
Total heap size    = 83885640
Total free space   = 48723768
Total reserved free space   =  1035792
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(2,2)"  desc=38003cee8
Total heap size    =369096816
Total free space   =258674312
Total reserved free space   = 16982464
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(2,3)"  desc=38003e740
Total heap size    =218102664
Total free space   = 17202608
Total reserved free space   = 10966696
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(3,0)"  desc=380043660
Total heap size    =184548408
Total free space   =    13008
Total reserved free space   =  5061928
Unpinned space     = 26943408  rcr=4930 trn=9425
Permanent space    =179472608
HEAP DUMP heap name="sga heap(3,1)"  desc=380044eb8
Total heap size    = 67108512
Total free space   = 27568352
Total reserved free space   =     4744
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(3,2)"  desc=380046710
Total heap size    =352319688
Total free space   =233302736
Total reserved free space   = 15981216
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
HEAP DUMP heap name="sga heap(3,3)"  desc=380047f68
Total heap size    =385873944
Total free space   =143746536
Total reserved free space   = 19402616
Unpinned space     =        0  rcr=0 trn=0
Permanent space    =        0
.....
******************************************************

So above stats shows that we have three sub-pools [Sub pool (1,0), (1,1), (1,2), (1,3) ….. (3,0), (3,1), (3,2), (3,3)] of SGA Heaps with Four Durations each heap. And every duration has its own size, free space and reserved free space. This type of distribution possibly causes the ORA 4031 even when you have enough free space in other durations and this is what the actual cause in my earlier case, but here we won’t discuss how we fixed that.

Okay so now lets do the same with 12c database. Lets generate the Heap Dump for SGA in 12c database and see the distributions of durations here. Once again we will use the oradebug to dump heapdump with Level 2.

SQL> oradebug setmypid
Statement processed.
SQL> oradebug dump heapdump 2
Statement processed.
SQL> oradebug tracefile_name
/u01/app/oracle/diag/rdbms/tunedb/tunedb/trace/tunedb_ora_11054.trc


******************************************************
HEAP DUMP heap name="sga heap(1,0)"  desc=0x60103678
 extent sz=0xfe0 alt=304 het=32767 rec=9 flg=0x82 opc=0
 parent=(nil) owner=(nil) nex=(nil) xsz=0x1000000 heap=(nil)
 fl2=0x24, nex=(nil), idx=1, dsxvers=1, dsxflg=0x0
 dsx first ext=0x73000000
 dsx empty ext bytes=0  subheap rc link=0x730000c0,0x730000c0

******************************************************
HEAP DUMP heap name="sga heap(1,3)"  desc=0x60107f80
 extent sz=0xfe0 alt=304 het=32767 rec=9 flg=0x82 opc=0
 parent=(nil) owner=(nil) nex=(nil) xsz=0x1000000 heap=(nil)
 fl2=0x24, nex=(nil), idx=1, dsxvers=1, dsxflg=0x0
 dsx first ext=0x7e000000
 dsx empty ext bytes=0  subheap rc link=0x7e0000c0,0x7e0000c0

Alright, so here we only have two groups of SGA durations – ‘Sub pool 1, duration 0’ and ‘Sub Pool 1 duration 3’ for improved sharability and to avoid ORA 4031 errors.

Hope It helps
Prashant Dixit

Posted in Advanced | Tagged: , , , | Leave a Comment »

How to troubleshoot, understand HW events & measure performance using PERF (Linux Profiler)! – Part 2

Posted by FatDBA on January 13, 2018

Repeat: I think I’ve mistakenly deleted the post, so re-posting the same.

Hey Folks,

Back with second edition of my previous post on troubleshooting performance issues using Linux in-built profiler named ‘perf’. Many of the users requested to write about few use cases situations on when and how to use this tool.

So, this post is all about discussing some cases that i have faced while working on few performance tuning projects and few of them are from the tests that I’ve performed on my Lab systems.

Okay coming back to the reason on why we need this, perf i mean ?
The answer is sometimes Oracle wait interface is not enough and you need to dig deeper inside the system to understand the problem. That point you have to use some third party, in-built dedicated tools for performance investigation. There perf might help you to understand what your resource intensive query is doing on OS layers.
So, using the tool you can monitor your process on what’s its doing!

And in case if you are using perf on Virtualised system, you might get error “perf.data file has no samples“.
In order to fix it try with “-e cpu-clock” arguments to collect the sample data and then interpret the file.

Okay so one fine day, you saw a spike in server’s CPU consumption using TOP, Oratop, OEM, Scheduled scripts or by any possible monitoring techniques and you have identified the process and its other attributes.
Using the PID you have reached the SID, SQL_ID, SQL_TEXT and other statistics. So now you might want to analyze and understand the oracle’s execution.

 
Global Information
------------------------------
 Status              :  EXECUTING
 Instance ID         :  1
 Session             :  DIXIT (1:53089)
 SQL ID              :  71aa5ju8pwtf2
 SQL Execution ID    :  16777216
 Execution Started   :  01/09/2018 06:10:32
 First Refresh Time  :  01/09/2018 06:10:32
 Last Refresh Time   :  01/09/2018 06:13:03
 Duration            :  152s
 Module/Action       :  SQL*Plus/-
 Service             :  SYS$USERS
 Program             :  sqlplus@dixitlab.localdomain (TNS V1-V3)

Global Stats
========================================================
| Elapsed |   Cpu   |    IO    | Buffer | Read | Read  |
| Time(s) | Time(s) | Waits(s) |  Gets  | Reqs | Bytes |
========================================================
|     163 |     150 |      115 |     5M | 648K |   5GB |
========================================================

SQL Plan Monitoring Details (Plan Hash Value=12102956)
===================================================================================================================================================================
| Id   |      Operation       | Name |  Rows   | Cost  |   Time    | Start  | Execs |   Rows   | Read | Read  | Activity |       Activity Detail       | Progress |
|      |                      |      | (Estim) |       | Active(s) | Active |       | (Actual) | Reqs | Bytes |   (%)    |         (# samples)         |          |
===================================================================================================================================================================
| -> 0 | SELECT STATEMENT     |      |         |       |       150 |     +2 |     1 |        0 |      |       |          |                             |          |
| -> 1 |   SORT AGGREGATE     |      |       1 |       |       150 |     +2 |     1 |        0 |      |       |          |                             |          |
| -> 2 |    TABLE ACCESS FULL | T294 |   96523 | 56397 |       151 |     +1 |     1 |      306 | 648K |   5GB |    99.34 | Cpu (69)                    |      31% |
|      |                      |      |         |       |           |        |       |          |      |       |          | db file sequential read (6) |          |
|      |                      |      |         |       |           |        |       |          |      |       |          | direct path read (126)      |          |
===================================================================================================================================================================

Above results points to high CPU Time (150 Seconds). Now to look more deeper in to things and to understand what exactly the session is doing.

[root@dixitlab ~]#  perf top -e cpu-clock -p 3505 

   PerfTop:     349 irqs/sec  kernel:42.7%  exact:  0.0% [1000Hz cpu-clock],  (target_pid: 3505)
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

             samples  pcnt function                    DSO
             _______ _____ ___________________________ __________________________________________________

             1981.00 46.2% _raw_spin_unlock_irqrestore [kernel.kallsyms]
              906.00 21.1% _intel_fast_memcmp          /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
              270.00  6.3% kole_simple_string_match    /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle 
               96.00  2.2% copy_user_generic_unrolled  [kernel.kallsyms]
               96.00  2.2% kcbgtcr                     /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
               54.00  1.3% __intel_new_memset          /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle 
               37.00  0.9% __intel_ssse3_rep_memcpy    /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
               30.00  0.7% kghfrf                      /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
               28.00  0.7% kghalf                      /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
               27.00  0.6% kcbldio                     /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
               24.00  0.6% kdxbrs1                     /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
               24.00  0.6% kspgvc                      /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle
               21.00  0.5% kksMapCursor                /u01/app/oracle/product/12.2.0/dbhome_1/bin/oracle

Results shows system was mostly busy processing kernel calls (kernel.kallsyms) and its function ‘_raw_spin_unlock_irqrestore‘ with 46% of its time spend using CPU cycles. It’s coming with huge overhead and with large samples and in general irq_restore shows up because re-enabling interrupts is costly, but this is not the real CPU consumption but is how the tool, Interrupt and system works in few of the Virtual systems – I was testing it on my Lab VM.

Here i would like to thank Tanel Poder for reviewing the document and highlighting the issue with the perf top command when executed inside a VM environment.

“…… perf top that concluded _raw_spin_unlock_irqrestore taking lots of CPU time. I’ve been through this myself in past – and this is likely not actual CPU usage but rather how perf, OS and interrupts work in some VMs (basically measurement bias/errors by these tools). I assume that you tested in a VM? ……..“

Next in the list is function/object ‘_intel_fast_memcmp‘ called by oracle with 21.1% which i believe is for fast memory compilations.

So the conclusion of the analysis:
Most of the CPU by process was spend processing kernel calls and for for fast memory compilations.

Purpose Revisit: This gives you a glimpse of what happens with the process calls and monitors a cpu-bound, database process is pass its time.

Hope It Helps
Prashant Dixit

Posted in Advanced, troubleshooting | Tagged: , | Leave a Comment »

Using Flame Graphs to analyze performance & workloads: Part 1

Posted by FatDBA on January 13, 2018

Hi Guys,

Now after my last two posts on Perf tool/profiler, i guess this the right time to move ahead with interpretation on stack trace data.
Back with another post, this time as promised about ‘Flame Graphs’!
So, would like to first start with the very standard question – Why, what this Flame Graphs are ?

Flame Graphs is tool developed by Kernel/System Performance maestro Brendan Gregg to project or visualize strack traces for both user/all processes and kernel-level stacks for CPU, Off-CPU etc. You can use it on stacks/traces collected by various profiler i.e. perf, Dtrace, SystemTap etc.

In this post we will be using PERF for statistics collection and project them using Flame Graphs and will understand the complexity and the code paths of many Oracle’s internal functions. This being the very first edition for this subject we will start with basics and will first discuss about CPU graphs and try to understand why and where CPUs are busy using stack traces and identify hot code-paths.This can be a really helpful tool for fast identification of performance problems where conventional oracle performance tools failed.

Okay, let’s jump on how you will do it. First you have to download the toolkit (Download Link)
This tool generates results in SVG format using below three steps.

  • Stats Collection using any of the profiler e.g. Perf, Dtrace, Systemtap etc.
  • Compress or Fold your stats
  • Finally use the core (flamegraph.pl) script to generate the SVG file.

Okay, so now after enough description on tool, let’s start out first case – “CPU Flame Graphs for an expensive SQL Statement”.

Scenario: We have a resource intensive query “COUNT on one of the big table “ running from session with SPID 19280
So, we will be collecting process strack traces while the SQL statement was in run.

SQL> explain plan for select count(*) from dwh.**********_*** where ****TIMESTAMP > '23-JUN-16' and ******KEY > 789999999;

Explained.

SQL> SELECT * FROM TABLE(dbms_xplan.display);

PLAN_TABLE_OUTPUT
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Plan hash value: 1112565023

------------------------------------------------------------------------------------------
| Id  | Operation          | Name                | Rows  | Bytes | Cost (%CPU)| Time     |
------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                     |     1 |    14 |  2604K  (1)| 08:40:53 |
|   1 |  SORT AGGREGATE    |                     |     1 |    14 |            |          |
|*  2 |   TABLE ACCESS FULL| **************_**** |    39M|   524M|  2604K  (1)| 08:40:53 |
------------------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - filter("*****KEY">789999999 AND "*****TIMESTAMP">'23-JUN-16')

14 rows selected.

SQL>

Now collecting perf statistics for the process at low-frequency of 997 Hertz.

[root@dselimw5862 perftest]# perf record -a -g -F997 -p 19280 

perf record: Woken up 3926 times to write data ]
[ perf record: Captured and wrote 981.809 MB perf.data (9751081 samples) ]

This will create perf data file under the same directory.

[root@dselimw5862 perftest]# ls -ltrh
-rw-------. 1 root root 982M Jan 12 14:58 perf.data

Next, collapse the stacks using below

[root@dselimw5862 perftest]# /root/perftest/FlameGraph-master/stackcollapse-perf.pl

Finally, time to render the SVG file.

[root@dselimw5862 perftest]# /root/perftest/FlameGraph-master/flamegraph.pl --title "Flame Graph: Selective data search on dwh.W6AUDITHISTORY_FULL Table" > ImageAuditHistory.svg

Let’s open the SVG file in browser and see what’s it’s got!

Important Note: The x-axis reflects stack profile population, and the y-axis represents the stack depth

Here you will see many of the familiar Oracle specific functions/processes opiodr, opitsk, opiino, opidrv etc.
Okay so, If you take a look at the map, you will see few of the Linux processes spawned first which called Oracle Internal functions next which moved to the query execution (i.e SELECT FETCH -> GROUP BY SORT -> TABLE ACCESS) and then to other functions. For example you can see the execution time during the sampling was spent in kdstf0100101000km functions which was called by kdsttgr which is Kernel Data Scan Table Get Row.

So, now question ‘Which function is on CPU the most ??‘
The top edge shows who is on CPU directly, copy_user_generic_unrolled function which is used when there is no optimization on CPU level.
Remember Flame Graphs follows ancestry, means copy_user_generic_unrolled was called by __pread_nocancel and __pread_nocancel was called by ksfd_skgfqio and which called … and then it points to …

Next look little above from the bottom of the graph with three functions __libc_start_main, __pread_nocancel and kaf4reasrp1km. If we visually compare them it is clear that __pread_nocancel function was running more often than other two. We see one of the linux function __pread_nocancel() was sampled a lot during the query execution and found repeated 2 times in the graph. This is used by libc when a cancellation point (a POSIX threads concept) is not permitted.

In Short: So, the Flame Graphs might help you when you have a problem in hand ‘My Production Database had poor performance”.
It was a heavy CPU consumer, so i used to CPU profiler to see WHY ? – But do you thing reading those lengthy, messy, long, raw traces that easy. So, here at this point there is a need of some data representation tool to capture the problems and sources and to visualize them. Here comes Flame Graphs to picture, The hero! It also gives you flexibility to move your cursor to any of the object to understand how much percent the object was present during the sample time.

Well, in later editions for this subject i will try to cover more scenarios. The sole purpose of this post is to give audience an idea, a gist of the topic.

Hope It Helps
Prashant Dixit

Posted in Advanced | Tagged: , | Leave a Comment »

How to troubleshoot, understand HW events & measure performance using PERF (Linux Profiler)! – Part 1

Posted by FatDBA on January 8, 2018

Hi Pals,

Some time back i was part of a performance task-force for one of our customers in Czech where i did a complete full 360 degrees auditing and troubleshooting of their full database stack. But you know most of the times it not just the software or the database but the real problem lies underneath the OS layer.

Okay, so back to the discussion – So during the assignment at one point where i had to collect some statistics for the OS to prove my point that ‘Problem is not with the Database‘ but with something else, i had used few of the tools like DTrace, Linux Trace Toolkit or LTT, Systemtap and one of the profiler which i hardly used before ‘Perf‘.

Yes, so today’s topic of discussion is to give you some basic idea (Will cover the deep analysis using the tool in future) about the ‘Perf’ profiler which helps to dissect and understand hardware events and measures performance.

Okay, let me take a real time scenario where a sudden surge in CPU in a production environment raised questions.
During the initial probing of issues we have found that that about 25% of the CPU is consumed by SQLPLUS command with PID 6720, lets see what and why its doing that ….

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 6720 oracle    20   0  113m  14m 9.8m R 25.8  0.5   0:20.32 sqlplus   as sysdba

Lets first Sample on-CPU functions for the process ID on 99 Hertz using below command for PID 6720.

perf record -F 99 -p 6720

This will produce a raw file named ‘perf.data’ under the same directory from where you called it.
Let it run for some time. You can cancel the execution by pressing CTRL+C.

[root@dixitlab perf]# ls -ltrh
total 12K
-rw-------. 1 root root 12K Jan  8 01:14 perf.data

Now lets see what’s there inside the raw file —- lets Dump raw contents from perf.data as hex for debugging.
For this we will use perf script with flag D, it will automatically considers the perf.data file present under the directory to show charts.


[root@dixitlab perf]# perf script -D

0x110 [0x48]: event: 1
.
. ... raw event: size 72 bytes
.  0000:  01 00 00 00 01 00 48 00 ff ff ff ff 00 00 00 00  ......H.........
.  0010:  00 00 00 00 00 00 00 00 ff ff ff 9f ff ff ff ff  ................
.  0020:  00 00 00 81 ff ff ff ff 5b 6b 65 72 6e 65 6c 2e  ........[kernel.
.  0030:  6b 61 6c 6c 73 79 6d 73 5d 5f 74 65 78 74 00 00  kallsyms]_text..
.  0040:  00 00 00 00 00 00 00 00                          ........
.
0x110 [0x48]: PERF_RECORD_MMAP -1/0: [0(0xffffffff9fffffff) @ 0xffffffff81000000]: [kernel.kallsyms]_text

0x158 [0x78]: event: 1
.
. ... raw event: size 120 bytes
.  0000:  01 00 00 00 01 00 78 00 ff ff ff ff 00 00 00 00  ......x.........
.  0010:  00 00 00 a0 ff ff ff ff ff bf 01 00 00 00 00 00  ................
.  0020:  00 00 00 00 00 00 00 00 2f 6c 69 62 2f 6d 6f 64  ......../lib/mod
.  0030:  75 6c 65 73 2f 32 2e 36 2e 33 39 2d 34 30 30 2e  ules/2.6.39-400.
.  0040:  32 39 37 2e 33 2e 65 6c 36 75 65 6b 2e 78 38 36  297.3.el6uek.x86
.  0050:  5f 36 34 2f 6b 65 72 6e 65 6c 2f 64 72 69 76 65  _64/kernel/drive
.  0060:  72 73 2f 6d 64 2f 64 6d 2d 6d 6f 64 2e 6b 6f 00  rs/md/dm-mod.ko.
.  0070:  00 00 00 00 00 00 00 00                          ........
.


0x2bf8 [0x68]: PERF_RECORD_MMAP 4387/4387: [0x7f6062943000(0x4c000) @ 0]: /u01/app/oracle/product/12.2.0/dbhome_1/lib/libons.so

0x2c60 [0x68]: event: 1
.
. ... raw event: size 104 bytes
.  0000:  01 00 00 00 02 00 68 00 23 11 00 00 23 11 00 00  ......h.#...#...
.  0010:  00 80 ba 62 60 7f 00 00 00 90 4d 00 00 00 00 00  ...b`.....M.....
.  0020:  00 00 00 00 00 00 00 00 2f 75 30 31 2f 61 70 70  ......../u01/app
.  0030:  2f 6f 72 61 63 6c 65 2f 70 72 6f 64 75 63 74 2f  /oracle/product/
.  0040:  31 32 2e 32 2e 30 2f 64 62 68 6f 6d 65 5f 31 2f  12.2.0/dbhome_1/
.  0050:  6c 69 62 2f 6c 69 62 6e 6e 7a 31 32 2e 73 6f 00  lib/libnnz12.so.
.  0060:  00 00 00 00 00 00 00 00                          ........
.
0x2c60 [0x68]: PERF_RECORD_MMAP 4387/4387: [0x7f6062ba8000(0x4d9000) @ 0]: /u01/app/oracle/product/12.2.0/dbhome_1/lib/libnnz12.so

0x2cc8 [0x68]: event: 1
.
. ... raw event: size 104 bytes
.  0000:  01 00 00 00 02 00 68 00 23 11 00 00 23 11 00 00  ......h.#...#...
.  0010:  00 10 2f 63 60 7f 00 00 00 c0 22 00 00 00 00 00  ../c`.....".....
.  0020:  00 00 00 00 00 00 00 00 2f 75 30 31 2f 61 70 70  ......../u01/app
.  0030:  2f 6f 72 61 63 6c 65 2f 70 72 6f 64 75 63 74 2f  /oracle/product/
.  0040:  31 32 2e 32 2e 30 2f 64 62 68 6f 6d 65 5f 31 2f  12.2.0/dbhome_1/
.  0050:  6c 69 62 2f 6c 69 62 69 70 63 31 2e 73 6f 00 00  lib/libipc1.so..
.  0060:  00 00 00 00 00 00 00 00                          ........
.
0x2cc8 [0x68]: PERF_RECORD_MMAP 4387/4387: [0x7f60632f1000(0x22c000) @ 0]: /u01/app/oracle/product/12.2.0/dbhome_1/lib/libipc1.so

0x2d30 [0x68]: event: 1
.
. ... raw event: size 104 bytes
.  0000:  01 00 00 00 02 00 68 00 23 11 00 00 23 11 00 00  ......h.#...#...
.  0010:  00 50 72 63 60 7f 00 00 00 20 07 00 00 00 00 00  .Prc`.... ......
.  0020:  00 00 00 00 00 00 00 00 2f 75 30 31 2f 61 70 70  ......../u01/app
.  0030:  2f 6f 72 61 63 6c 65 2f 70 72 6f 64 75 63 74 2f  /oracle/product/
.  0040:  31 32 2e 32 2e 30 2f 64 62 68 6f 6d 65 5f 31 2f  12.2.0/dbhome_1/
.  0050:  6c 69 62 2f 6c 69 62 6d 71 6c 31 2e 73 6f 00 00  lib/libmql1.so..
.  0060:  00 00 00 00 00 00 00 00                          ........
.
0x2d30 [0x68]: PERF_RECORD_MMAP 4387/4387: [0x7f6063725000(0x72000) @ 0]: /u01/app/oracle/product/12.2.0/dbhome_1/lib/libmql1.so

0x2d98 [0x78]: event: 1
.
. ... raw event: size 120 bytes
.  0000:  01 00 00 00 02 00 78 00 23 11 00 00 23 11 00 00  ......x.#...#...
.  0010:  00 c0 99 63 60 7f 00 00 00 90 3a 00 00 00 00 00  ...c`.....:.....
.  0020:  00 00 00 00 00 00 00 00 2f 75 30 31 2f 61 70 70  ......../u01/app
.  0030:  2f 6f 72 61 63 6c 65 2f 70 72 6f 64 75 63 74 2f  /oracle/product/
.  0040:  31 32 2e 32 2e 30 2f 64 62 68 6f 6d 65 5f 31 2f  12.2.0/dbhome_1/
.  0050:  6c 69 62 2f 6c 69 62 63 6c 6e 74 73 68 63 6f 72  lib/libclntshcor
.  0060:  65 2e 73 6f 2e 31 32 2e 31 00 00 00 00 00 00 00  e.so.12.1.......
.  0070:  00 00 00 00 00 00 00 00                          ........
.
0x2d98 [0x78]: PERF_RECORD_MMAP 4387/4387: [0x7f606399c000(0x3a9000) @ 0]: /u01/app/oracle/product/12.2.0/dbhome_1/lib/libclntshcore.so.12.1

0x2e10 [0x70]: event: 1
.
. ... raw event: size 112 bytes
.  0000:  01 00 00 00 02 00 70 00 23 11 00 00 23 11 00 00  ......p.#...#...
.  0010:  00 a0 f6 63 60 7f 00 00 00 30 65 03 00 00 00 00  ...c`....0e.....
.  0020:  00 00 00 00 00 00 00 00 2f 75 30 31 2f 61 70 70  ......../u01/app
.  0030:  2f 6f 72 61 63 6c 65 2f 70 72 6f 64 75 63 74 2f  /oracle/product/
.  0040:  31 32 2e 32 2e 30 2f 64 62 68 6f 6d 65 5f 31 2f  12.2.0/dbhome_1/
.  0050:  6c 69 62 2f 6c 69 62 63 6c 6e 74 73 68 2e 73 6f  lib/libclntsh.so
.  0060:  2e 31 32 2e 31 00 32 2e 00 00 00 00 00 00 00 00  .12.1.2.........
.
0x2e10 [0x70]: PERF_RECORD_MMAP 4387/4387: [0x7f6063f6a000(0x3653000) @ 0]: /u01/app/oracle/product/12.2.0/dbhome_1/lib/libclntsh.so.12.1

0x2e80 [0x70]: event: 1
.
. ... raw event: size 112 bytes
.  0000:  01 00 00 00 02 00 70 00 23 11 00 00 23 11 00 00  ......p.#...#...
.  0010:  00 20 a1 67 60 7f 00 00 00 40 0e 00 00 00 00 00  . .g`....@......
.  0020:  00 00 00 00 00 00 00 00 2f 75 30 31 2f 61 70 70  ......../u01/app
.  0030:  2f 6f 72 61 63 6c 65 2f 70 72 6f 64 75 63 74 2f  /oracle/product/
.  0040:  31 32 2e 32 2e 30 2f 64 62 68 6f 6d 65 5f 31 2f  12.2.0/dbhome_1/
.  0050:  6c 69 62 2f 6c 69 62 73 71 6c 70 6c 75 73 2e 73  lib/libsqlplus.s
.  0060:  6f 00 32 2e 31 00 32 2e 00 00 00 00 00 00 00 00  o.2.1.2.........
.
0x2e80 [0x70]: PERF_RECORD_MMAP 4387/4387: [0x7f6067a12000(0xe4000) @ 0]: /u01/app/oracle/product/12.2.0/dbhome_1/lib/libsqlplus.so

0x2ef0 [0x38]: event: 1
.
. ... raw event: size 56 bytes
.  0000:  01 00 00 00 02 00 38 00 23 11 00 00 23 11 00 00  ......8.#...#...
.  0010:  00 90 8f 09 ff 7f 00 00 00 10 00 00 00 00 00 00  ................
.  0020:  00 00 00 00 00 00 00 00 5b 76 64 73 6f 5d 00 70  ........[vdso].p
.  0030:  00 00 00 00 00 00 00 00                          ........
.

Now lets get back to the point where we had the process (PID: 6720) consuming highest CPU resources.

now profiling CPU’s with sample rate of 99 Hertz/second, -a for all CPU’s, -g for stack traces, sleep of 60 seconds for duration of run.

[root@dixitlab perf]# perf record -F 99 -a -g -- sleep 60 
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.388 MB perf.data (~16948 samples) ]

Now as i discussed above, it will generate a new perf.data under the directory.

[root@dixitlab perf]# ls -ltrh
total 768K
-rw-------. 1 root root 379K Jan  8 01:30 perf.data.old
-rw-------. 1 root root 386K Jan  8 01:31 perf.data

Next, we will interpret results from the newly created ‘perf.data’ file.
I have used ‘perf report -n –stdio’, you can simply use ‘perf report’ too.


[root@dixitlab perf]#
[root@dixitlab perf]# perf report -n --stdio 
# Events: 2K cpu-clock
#
# Overhead  Samples            Command          Shared Object                                    Symbol
# ........ ..........  ...............  .....................  ........................................
#
    50.05%       1001             sshd  [e1000]                [k] e1000_phy_read_status
                       |
                       --- e1000_phy_read_status
                          |
                          |--99.60%-- dev_hard_start_xmit
                          |          sch_direct_xmit
                          |          dev_queue_xmit
                          |          ip_finish_output
                          |          ip_output
                          |          ip_local_out
                          |          ip_queue_xmit
                          |          tcp_transmit_skb
                          |          tcp_write_xmit
                          |          __tcp_push_pending_frames
                          |          tcp_push
                          |          tcp_sendmsg
                          |          inet_sendmsg
                          |          sock_aio_write
                          |          do_sync_write
                          |          vfs_write
                          |          sys_write
                          |          system_call_fastpath
                          |          __write_nocancel
                           --0.40%-- [...]

     8.35%        167          swapper  [kernel.kallsyms]      [k] native_safe_halt
                    |
                    --- native_safe_halt
                        default_idle
                        cpu_idle
                        rest_init
                        start_kernel
                        x86_64_start_reservations
                        x86_64_start_kernel

     4.45%         89          sqlplus  [kernel.kallsyms]      [k] finish_task_switch
                    |
                    --- finish_task_switch
                        __schedule
                       |
                       |--87.64%-- __cond_resched
                       |          _cond_resched
                       |          mutex_lock
                       |          |
                       |          |--91.03%-- process_output_block
                       |          |          n_tty_write
                       |          |          tty_write
                       |          |          vfs_write
                       |          |          sys_write
                       |          |          system_call_fastpath
                       |          |          __write_nocancel
                       |          |
                       |           --8.97%-- process_output
                       |                     n_tty_write
                       |                     tty_write
                       |                     vfs_write
                       |                     sys_write
                       |                     system_call_fastpath
                       |                     __write_nocancel
                       |
                        --12.36%-- schedule
                                  sysret_careful
                                  |
                                  |--81.82%-- __write_nocancel
                                  |
                                   --18.18%-- __write_nocancel


.................
...........................
..................................

Note:
You can further sort results based on cpu, shared objects etc.
perf report –sort=cpu
perf report –sort=dso

Above results reflects the division. While taking a look at the tree, it shows a combination or a club of few OS Layer components (Starting from dev_hard_start_xmit) was sampled of time 50.05*99.60 = 49.84%
with parent sshd command and shared object [e1000] and is a physical read operation.

Rest of all might seem gibberish but needs time and patience to understand. I will try to cover the profiler in more detail with more use case scenarios in future. Till the time you can check the WIKI page of the tool.

Note: In case if you are using perf on VMWare system, you might get error “perf.data file has no samples“.
In order to fix it try using the ‘perf record’ option with -e cpu-clock arguments to collect the sample data and then use ‘perf report’ to interpret the file.

example:
perf record -F 99 -a -g -e cpu-clock — sleep 20


Hope It Helps
Prashant Dixit

Posted in Advanced | Tagged: , | 3 Comments »

Script to measure the Source & Contribution of any Wait Event within AWR snapshots.

Posted by FatDBA on January 1, 2018

🤘 Happy New Year Everyone! 🤘

So, here i am with the maiden post of this year.
Today i will discuss about a script that might be handy and useful while you are investigating any performance problem with the database.
In fact this is one of my favorite script which i always considers to run at a point when we have identified the leading wait event and the time-frame of the spike or for any transient variation in system behavior. This script helps you to measure the source and contribution or the impact of any specific wait event.

The script is a join between ASH views and historical views (dba_hist_active_sess_history AND dba_hist_snapshot) and provides you stats based on Average Number of Active Sessions (AAS) were waiting on this event during the period and with what total contribution in terms of percentage.

Lets work on a use case.
Suppose you have a system with huge User IOs happening, as conformed by any report or tool (AWR, ASH, Via any Script, OEM etc.)

For example using AWR we found one of the User IO wait class event ‘db file scattered read’ was leading the ‘Top Foreground Wait Events’ charts on a specific time when we have the high User IO load on system.

                                           Total Wait       Avg   % DB Wait
Event                                Waits Time (sec)      Wait   time Class
------------------------------ ----------- ---------- --------- ------ --------
db file scattered read              10,282       82.8    41.50ms  43.9 User I/O

So, now you have identified the leading wait event and the timeframe with maximum intensity or frequency, Lets call the script.

Script Inputs:
Example:

Enter value for event_class: User I/O
Enter value for event_name: db file scattered read
Enter value for begin_snap: 193
Enter value for end_snap: 200
Enter value for dbid: 2896132084

WITH
events AS (
SELECT /*+ MATERIALIZE NO_MERGE */
       SUBSTR(TRIM(h.sql_id||' '||h.program||' '||
       CASE h.module WHEN h.program THEN NULL ELSE h.module END), 1, 128) source,
       h.dbid,
       COUNT(*) samples
  FROM dba_hist_active_sess_history h,
       dba_hist_snapshot s
 WHERE h.wait_class = TRIM('&Event_Class') AND h.event = TRIM('&Event_Name')
   AND h.snap_id BETWEEN &Begin_Snap AND &End_Snap
   AND h.dbid = &dbid
   AND s.snap_id = h.snap_id
   AND s.dbid = h.dbid
   AND s.instance_number = h.instance_number
 GROUP BY
       h.sql_id,
       h.program,
       h.module,
       h.dbid
 ORDER BY
       3 DESC
),
total AS (
SELECT SUM(samples) samples,
       SUM(CASE WHEN ROWNUM > 15 THEN samples ELSE 0 END) others
  FROM events
)
SELECT e.source,
       e.samples,
       ROUND(100 * e.samples / t.samples, 1) percent,
       (SELECT DBMS_LOB.SUBSTR(s.sql_text, 1000, 1) FROM dba_hist_sqltext s WHERE s.sql_id = SUBSTR(e.source, 1, 13) AND s.dbid = e.dbid AND ROWNUM = 1) sql_text
  FROM events e,
       total t
 WHERE ROWNUM  0.1
 UNION ALL
SELECT 'Others',
       others samples,
       ROUND(100 * others / samples, 1) percent,
       NULL sql_text
  FROM total
 WHERE others > 0
   AND ROUND(100 * others / samples, 1) > 0.1;

Below is the output of the query.

SOURCE                                                                      SAMPLES      PERCENT       SQL_TEXT
------------------------------------------------------------------------  ---------- ---------------  -------------------------------
5av23g8w7f3ka sqlplus@dixitlab.localdomain (TNS V1-V3)                          1          45.23       select * from dixit.bigtab
9a7gbkahasj1a sqlplus@dixitlab.localdomain (TNS V1-V4)                          1           5          select * from dixit.gianttable
OTHERS                                                                          14         49.77

Okay so result shows one of the SQL statement with SQL id 5av23g8w7f3ka was responsible for more than 45% of these waits.
So, now you have the evidence and can start troubleshooting this specific statement to reduce the USER IOs.

Hope It Helps
Prashant Dixit

Posted in Advanced, troubleshooting | Tagged: , , | 3 Comments »

 
%d bloggers like this: