Paul
As if the organization is big with hundred of the database, it will be
difficult for DBA team who are also not Application DBA to keep track of all
critical statement and plan. So how do you keep track while working with
multiple environments and with Less of Application knowledge but aware of list
of the critical database to business?
What I am meant to say is1. Is there any way to automate the process that can
be done to get a list of Sqlid/Hashplan/execution plan details based on the
number of executions or so?2. Is there a way to get this details from ASH/AWR
history or Oracle Auto job which also analyze the environment to help SQL plan
management to make changes in the query execution plan or so?
Any suggestion or approach that can help in collecting the data on regular
basis and consult when such issue arises beside using available repository when
issue arises
Thanks for your comment
Sanjay
On Friday, August 25, 2017 1:28 PM, Paul Drake <bdbafh@xxxxxxxxx> wrote:
Having an inventory of typically used and business critical statements as well
as their execution plans and resource usage might be the kind of thing to have
around when response time issues occur.
Knowing the before part makes it a whole lot easier to diff.
Paul
On Aug 25, 2017 1:18 PM, "Sanjay Mishra" <dmarc-noreply@xxxxxxxxxxxxx> wrote:
Thanks Cary. After making the required parameter change, application query is
back to normal. Will keep the suggested command for future use.
RgdSanjay
On Thursday, August 24, 2017 2:57 PM, Cary Millsap
<cary.millsap@xxxxxxxxxxxx> wrote:
exec dbms_monitor.session_trace_ enable(:sid,:serial)
...where :sid and :serial are the session_id and serial# of the Oracle session
you're curious about. The resulting trace file will show you all the database
calls and system calls that the session has executed. If you're curious about
your trace file, feel free to send it to me, and I'll be glad to have a look.
Cary Millsap
Method R CorporationAuthor of Optimizing Oracle Performance and The Method R
Guide to Mastering Oracle Trace Data, 2nd edition
On Fri, Aug 18, 2017 at 2:49 PM, Sanjay Mishra <dmarc-noreply@xxxxxxxxxxxxx>
wrote:
After going thru patch details with Oracle it was found patch was doing the
following and so until got downtime to test patch thru all dev/test, the
following changes at database level resolve the issue_optimizer_dsdir_usage_
control=0 -- disable use of directives
_sql_plan_directive_mgmt_ control=0 -- disable creation of directives
Tx for all suggestion and updates
Sanjay
On Friday, August 18, 2017 11:41 AM, Sanjay Mishra
<dmarc-noreply@xxxxxxxxxxxxx> wrote:
Jonathan
Thanks for your input. After having long session with Oracle Support yesterday
where SQL plan management and other options tried to stick with one of the good
execution but issue remains and finally Oracle provided that it is due to bug
and need to apply Patch 16470650
Bug 16470650 - Plans missing after loading from AWR to a SQL Tuning Set ( Doc
ID 16470650.8 )
Working to get downtime to apply patch and will try the SQL again.
TxSanjay
On Friday, August 18, 2017 2:37 AM, Jonathan Lewis
<jonathan@xxxxxxxxxxxxxxxxxx> wrote:
If it's taking a lot of CPU it's not hanging.
The most likely explanation - in the absence of any detailed information - is
that the query has changed it's execution path.
If the select is taking seconds while the insert is taking hours this may mean
the insert path is not allowed to take the access path of the select statement
(e.g. it's a distributed query which is allowed to use a "driving_site
(remote)" strategy while the insert has to drive off the local site).
Alternatively the query started at a point in time when it had to do a huge
amount of read-consistency work, but the query doesn't have to do any because
it started at a much later point in time. (This one is a little unlikely given
the difference in scale, but a technical possibility).
Are your licensed to use the AWR, or have you installed Statspack. If the
insert has taken hours then its execution plan will have been captured in AWR
and you can check the plan and compare it with the "seconds" query. You could
query the dba_hist_active_sess_history to see where the insert spent most of
it's time (Randolf Geist has some excellent "XPLAN_ASH" material to do this for
you, but essentially it means pulling ASH rows for the SQL_ID and picking out
the plan operation details.
Regards
Jonathan Lewis
______________________________ __________
From: oracle-l-bounce@xxxxxxxxxxxxx <oracle-l-bounce@xxxxxxxxxxxxx > on behalf
of Sanjay Mishra <dmarc-noreply@xxxxxxxxxxxxx>
Sent: 18 August 2017 04:20
To: Oracle-L Freelists
Subject: Insert statement hanging
Hi
I had insert statement which is selecting database from 3-4 table and hanging
for several hours. Some main points are
- The Target Table where insert is going is empty table
- Select statement as itself is working geting 10K records in few sec
- Tried to create new table to insert but still not worked
- Insert session is showing very high wait on CPU and taking big CPU time
- Bouncing database and running insert worked first few min but as App is
started , it is again hanged and never complete
Opened Oracle SR now but they ask for Trace analyzer and so want to check
experts as what can be other thing to check. This was working fine and suddenly
started in last few days where no patching on Oracle/OS or major changes to
involved table are done.
Environment is Linux with Non-RAC using ASM as storage and Oracle 12c(12.1.0.2)
TIA
Sanjay
--
//www.freelists.org/ webpage/oracle-l
