RE: Can this sql restarts by itself
Date: Thu, 14 Oct 2021 11:27:12 -0400
Message-ID: <34ff01d7c10f$f67f7f60$e37e7e20$_at_rsiz.com>
Now you need to look for everything that calls that procedure, and the odds on favorite is that when it catches a failure status return, it (silently?) retries it.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Pap
Sent: Thursday, October 14, 2021 5:17 AM
To: Andy Sayer
Cc: Lok P; Oracle L
Subject: Re: Can this sql restarts by itself
Thank You Andy. Unfortunately that session has been killed now and I was monitoring during run time but was not aware that the sql monitor can be fetched based on sql_exec_id and considering its 11.2 that is no sql monitor history also available. But i got your point , that in case this must be failing and retriggering ten the execution count in the v$sql/dba_hist_sqlstat won't increase.
I am not sure if this will make any difference but this INSERT was getting executed by the "execute immediate" statement from the plsql proc. But i was trying to see, just in case this insert is failing after a few times (say 10+minutes) it should throw the error to the 'when others' block which is outside the loop. No exception handling is done inside the loop. And also I was able to see the sql_monitor report of the main_proc which was showing a single execution running over 6hrs+ span duration, which means the main_proc was definitely just called once only. And the PROC2 call is not inside any loop.
Below is the basic block, i have removed detail logic from it. But from this exception handling , it/PROC2 should not rerun automatically in case of failure of the Insert query which is called within it.
procedure PROC2(.......)
is
rsql Varchar2(4000) := 'select .........';
myrec sys_refcursor;
insertSqlprefix varchar2(2000) := 'insert into TAB1 ......';
BEGIN
open myrec for rsql using IN_DATE;
LOOP
FETCH myrec into .......;
EXIT WHEN myrec%NOTFOUND;
rerunFlag := null;
raiseExcpFlag := 'N';
.........
...........
addsql := ' select .......' ;
execute immediate insertSqlprefix || addsql using id, cby;
COMMIT;
end if;
end loop;
EXCEPTION
when OTHERS then
PKG1.updatelog(logid, 'FAIL');
raise_application_error (-20002,'An error has occurred running PROC2: '||SQLCODE||' -ERROR- '||SQLERRM);
END PROC2;
/
procedure MAIN_PROC(...... )
is
....................
BEGIN
...............
IF lrtyp !='XXX' THEN
PROC2(IN_DATE, logid, IN_DATE, lctyp, lrtyp);
END IF;
commit;
Exception
when others then
dbms_output.put_line ('Oracle error code : ' || to_char(sqlcode));
dbms_output.put_line ('Oracle error message : ' || substr(sqlerrm,1,200));
ROLLBACK;
updateRunLogEntry(logid, FAIL);
raise_application_error (-20002,'An error has occurred runing MAIN_PROC: '||SQLCODE||' -ERROR- '||SQLERRM);
END MAIN_PROC;
/
On Thu, Oct 14, 2021 at 1:15 PM Andy Sayer <andysayer_at_gmail.com> wrote:
Hi,
Different sql_exec_id means that it’s being called multiple times, this isn’t traditional SQL restarts but most likely being called again by the user code.
You should be able to pull out the previous executions of the SQL by providing their sql_exec_id to your SQL monitor query. Are they completing or are they failing?
Your SQL is being executed from plsql, would be worth looking at that code. My guess is something in the exception clause causing it to retry itself.
Thanks,
Andy
On Thu, 14 Oct 2021 at 06:09, Lok P <loknath.73_at_gmail.com> wrote:
I have experienced a similar situation and also it's well noted in Jonathan's blog below regarding how a UPDATE having a full scan in its execution path can restart itself. But if i remember correctly the sql monitor used to be the same just that the Execs column used to increase pointing to increase in number of tablescans. But here in your case , you have noted a whole new set of sql_exec_ids(which mean new sql monitors altogether), so I am not sure if an INSERT can show symptoms when its historical partitions are being dropped . In an ideal scenario a partition with all local indexes should not cause any issue for other DMLS which are happening on other partitions even at the same time. But yes the executions count should have been increased to more than 1, in gv$sql and dba_hst_sqlstat considering your query ran for 10+ minutes each time/iteration before it changes the sql_exec_id. From your plan it seems INSERT as a SELECT query, and that to operate in parallel, so not sure if that can cause such a thing. Others can comment.
https://jonathanlewis.wordpress.com/2019/09/10/update-restarts/
Regards
Lok
On Thu, Oct 14, 2021 at 1:26 AM Pap <oracle.developer35_at_gmail.com> wrote:
Hello Listers, We have a 11.2.0.4 customer database executing plsql procedure. And below INSERT seems to keep coming with new sql_exec_id and sql_exec_start whereas it's not inside any looping kind of logic. So we were wondering why the same sqls execute so many times. As per dba_hist_active_sess_history it has already completed ~25+ executions as we see ~25different different sql_exec_id and sql_exec_start for all of those executions, each running for ~10minutes+ sequentially one after other, without any gap. But then I noticed the v$sql showing executions as 0 and even dba_hist_sqlstat also showing as 1 for this same sql_id. We do have a session running in parallel, dropping historical partitions from the same table to which this below query inserting data into, but this insert is happening into the live partition. So are we hitting any related bugs here? And if it's the same, sql is restarting and not really moving , is there any other way to confirm?
The sql monitor looks something like below
Global Information
Status : EXECUTING
Instance ID : 1
SQL Execution ID : 16777240
Execution Started : 10/13/2021 11:55:14
First Refresh Time : 10/13/2021 11:55:17
Last Refresh Time : 10/13/2021 11:59:13
Duration : 241s
PLSQL Entry Ids (Object/Subprogram) : 161298,9
PLSQL Current Ids (Object/Subprogram) : 161298,15
Global Stats
| Elapsed | Cpu | IO | Application | Concurrency | PL/SQL | Buffer | Read | Read | Write | Write | Cell | | Time(s) | Time(s) | Waits(s) | Waits(s) | Waits(s) | Time(s) | Gets | Reqs | Bytes | Reqs | Bytes | Offload |
| 243 | 53 | 190 | 0.08 | 0.00 | 0.36 | 5M | 42591 | 39GB | 163 | 19MB | |
SQL Plan Monitoring Details (Plan Hash Value=3677945390)
| Id | Operation | Name | Rows | Cost | Time | Start | Execs | Rows | Read | Read | Write | Write | Mem | Temp | Activity | Activity Detail | Progress | | | | | (Estim) | | Active(s) | Active | | (Actual) | Reqs | Bytes | Reqs | Bytes | | | (%) | (# samples) | |
==============================================================================================================================================================================================================================================
| 0 | INSERT STATEMENT | | | | | | 1 | | | | | | | | | | | | 1 | LOAD TABLE CONVENTIONAL | | | | | | 1 | | | | | | | | | | | | 2 | SEQUENCE | TAB1_SEQ | | | | | 1 | | | | | | | | | | | | 3 | PX COORDINATOR FORCED SERIAL | | | | | | 1 | | | | | | | | | | | | 4 | PX SEND QC (RANDOM) | :TQ10001 | 3314 | 110K | | | 1 | | | | | | | | | | | | 5 | HASH JOIN | | 3314 | 110K | 231 | +7 | 1 | 0 | | | 154 | 18MB | 2M | 21M | | | | | 6 | PX BLOCK ITERATOR | | 8262 | 110K | 231 | +7 | 1 | 158K | | | | | | | | | | | -> 7 | TABLE ACCESS STORAGE FULL | TAB2 | 8262 | 110K | 239 | +2 | 13 | 158K | 39391 | 38GB | | | | | 99.58 | Cpu (42) | 100% | | | | | | | | | | | | | | | | | | cell multiblock physical read (196) | | | 8 | BUFFER SORT | | | | | | | | | | | | | | | | | | 9 | PX RECEIVE | | 60 | 1 | | | | | | | | | | | | | | | 10 | PX SEND BROADCAST | :TQ10000 | 60 | 1 | | | | | | | | | | | | | | | 11 | INDEX SKIP SCAN | IDX_TAB3 | 60 | 1 | | | | | | | | | | | | | |
==============================================================================================================================================================================================================================================
-- http://www.freelists.org/webpage/oracle-lReceived on Thu Oct 14 2021 - 17:27:12 CEST