Skip to main content

SQL Tuning Techniques

Tuning has always been a hard nut to crack for Oracle DBA's. Every DBA who had spent time dealing with databases would have faced situations when he got complaints about slowness in the database or most of the time a query which was running fine in the past but has been degraded in performance. It is not always advisable to extract the AWR or statspack report and go through each section for digging into such issues. Oracle has provided some very useful database views which can help the administrators in such situations. Below are some steps and queries making use of these views which can help administrators to find the cause of the slowness in the database.

Rather than checking the full AWR report always identify the bottle necks and try fixing those.

Check for Active Wait Events in the Database.

select event,count(*) from  gv$session where event not like 'Streams%' and wait_class <> 'Idle' group by event order by 2;

Find the most executed sql’s which are waiting on active wait events.


select sql_id,sql_hash_value,count(*) from gv$session where event='&ev' group by sql_id,sql_hash_value order by 3;

Check the Gets/Execution for all the plan hash values of the problematic sql_id.

select PARSING_SCHEMA_NAME,BUFFER_GETS , EXECUTIONS , BUFFER_GETS/EXECUTIONS, ROWS_PROCESSED ,ROWS_PROCESSED/EXECUTIONS,SQL_PROFILE,CHILD_NUMBER,PLAN_HASH_VALUE from v$sql where sql_id='&sql_id' and EXECUTIONS<>0;

 If the result of above query shows more than one plan hash value for the same sql_id, find for which plan hash value the gets/execution is low. Get the history of the SQL plans generated for the SQL_ID. Find if the SQL plan is recently changed for the sql_id and if the changed plan is poor than the earlier plan.

Find different plan's for a SQL

select distinct plan_hash_value,timestamp from v$sql_plan where sql_id='&sqlid';

In case current plan is not giving the desired performance, check the output of the above query and also check with user about when the SQL was performing well. SQL profile using the good plan_hash_value can be created for sql_id so that it uses that good plan.

Steps to implement SQL Profile

  •  If sql is in AWR.

declare
ar_profile_hints sys.sqlprof_attr;
cl_sql_text      clob;
begin
select
extractvalue(value(d), '/hint') as outline_hints bulk collect into ar_profile_hints from xmltable('/*/outline_data/hint'  passing ( select xmltype(other_xml) as xmlval from dba_hist_sql_plan
where sql_id = '&&1' and     plan_hash_value = &&2 and  other_xml is not null)) d;

select sql_text into cl_sql_text from dba_hist_sqltext where sql_id = '&&1';

dbms_sqltune.import_sql_profile(sql_text    => cl_sql_text, profile     => ar_profile_hints
, category    => '&&3', name  => 'PROFILE_&&1' 
-- use force_match => true
-- to use CURSOR_SHARING=SIMILAR
-- behaviour, i.e. match even with
-- differing literals
, force_match => &&4
);
end;

============================================
input
==========

Enter value for 1: 6k4j4r9k777h6 (sql_id) 
Enter value for 2: 2421233948 (plan_hash_value) 
Enter value for 3:(default Category) 
Enter value for 4: true (force match)
================

  • If the sql is in cursor(recently executed).

declare
ar_profile_hints sys.sqlprof_attr;
cl_sql_text      clob;
begin
select extractvalue(value(d), '/hint') as outline_hints bulk collect into ar_profile_hints from
xmltable('/*/outline_data/hint' passing (select xmltype(other_xml) as xmlval from v$sql_plan
where sql_id = '&&1' and     child_number = &&2 and     other_xml is not null)) d;

select sql_fulltext into cl_sql_text from v$sql where sql_id = '&&1' and      child_number = &&2;

dbms_sqltune.import_sql_profile(
sql_text    => cl_sql_text
, profile     => ar_profile_hints
, category    => '&&3'
, name        => 'PROFILE_&&1'
-- use force_match => true
-- to use CURSOR_SHARING=SIMILAR
-- behaviour, i.e. match even with
-- differing literals
, force_match => &&4
);
end;
/
/*============================================
--input
==========

--Enter value for 1: 6k4j4r9k777h6 (sql_id) 
Enter value for 2: 2421233948 (plan_hash_value) 
Enter value for 3: (default Category) 
Enter value for 4: true (force match) ================*/
  •  Measure the performance after profile implementation and check if the profile is being used.
select PARSING_SCHEMA_NAME,BUFFER_GETS , EXECUTIONS , BUFFER_GETS/EXECUTIONS, ROWS_PROCESSED ,ROWS_PROCESSED/EXECUTIONS,SQL_PROFILE,CHILD_NUMBER,PLAN_HASH_VALUE from v$sql where sql_id='&sql_id' and EXECUTIONS<>0

  • If performance has improved, keep the new profile, else profile can be dropped.
select name , sql_text from dba_sql_profiles where name like '&sqlid';

exec DBMS_SQLTUNE.DROP_SQL_PROFILE (name => '&1'); 


Comments

Popular posts from this blog

Configure Oracle ASM Disks on AIX

Configure Oracle ASM Disks on AIX You can use below steps to configure the new disks for ASM after the raw disks are added to your AIX server by your System/Infrastructure team experts: # /usr/sbin/lsdev -Cc disk The output from this command is similar to the following: hdisk9 Available 02-T1-01 PURE MPIO Drive (Fibre) hdisk10 Available 02-T1-01 PURE MPIO Drive (Fibre) If the new disks are not listed as available, then use the below command to configure the new disks. # /usr/sbin/cfgmgr Enter the following command to identify the device names for the physical disks that you want to use: # /usr/sbin/lspv | grep -i none This command displays information similar to the following for each disk that is not configured in a volume group: hdisk9     0000014652369872   None In the above example hdisk9 is the device name and  0000014652369872  is the physical volume ID (PVID). The disks that you want to use may have a PVID, but they must not belong to a volu...

Gitlab installation steps on Redhat Linux

In this blog we will see the steps to install Gitlab on Redhat Enterprise Linux 6. I will be using the virtual machine "gitserver" that I have created on Google Cloud. You can use any server or VM running RHEL 6 and follow these steps. Follow the below steps to install gitlab. Run these steps as root user. # yum install -y curl policycoreutils-python openssh-server cronie # lokkit -s http -s ssh  # yum install postfix  # service postfix start  # chkconfig postfix on  # curl https://packages.gitlab.com/install/repositories/gitlab/gitlab-ee/script.rpm.sh | sudo bash  # EXTERNAL_URL="http://34.69.44.142" yum -y install gitlab-ee  You will see a screen similar to below, once your gitlab installation is successful. You can now access the gitlab console using the http or https url that you provided during the installation, i.e., http://<ip/server_name> http://gitserver.localdomain.com or  http://34.69.44.142 When you open the c...

Load records from csv file in S3 file to RDS MySQL database using AWS Data Pipeline

 In this post we will see how to create a data pipeline in AWS which picks data from S3 csv file and inserts records in RDS MySQL table.  I am using below csv file which contains a list of passengers. CSV Data stored in the file Passenger.csv Upload Passenger.csv file to S3 bucket using AWS ClI In below screenshot I am connecting the RDS MySQL instance I have created in AWS and the definition of the table that I have created in the database testdb. Once we have uploaded the csv file we will create the data pipeline. There are 2 ways to create the pipeline.  Using "Import Definition" option under AWS console.                    We can use import definition option while creating the new pipeline. This would need a json file which contains the definition of the pipeline in the json format. You can use my Github link below to download the JSON definition: JSON Definition to create the Data Pipeline Using "Edit Architect" ...