Hi Stefan,
can you set in your main block (before you call set_start) something like
DBMS_SESSION.SET_IDENTIFIER(something-unique);
and in every procedure
DBMS_APPLICATION_INFO.SET_ACTION ('set_begin');
and then trace the calls?
I am curious if the negative time is visible there also.
Martin
2018-01-13 9:01 GMT+01:00 Stefan Knecht <knecht.stefan@xxxxxxxxx>:
Hi all
If you have a code structure similar to this:
begin
procedure set_start
is
pragma autonomous_transaction;
begin
update t set start_time=systimestamp;
commit;
end;
procedure set_end
is
pragma autonomous_transaction;
begin
update t set end_time=systimestamp;
commit;
end;
begin
set_start;
-- run some code here that takes anywhere between 0.1 and 2 seconds to
complete
commit;
set_end;
end;
/
I have some executions of this code, 40 out of 6000 runs, where I end up
with the end_time being a timestamp BEFORE the begin_time and not just a
microsecond or two. 0.4 seconds, 0.6 seconds. A significant amount of time.
This leads me to believe that Oracle somehow fires autonomous transactions
asynchronously or out of order of the code structure?
Before I go off on a digging spree - has anyone seen this before or has an
explanation?
Oracle 12.1 on *nix.
Cheers
Stefan
--
//
zztat - The Next-Gen Oracle Performance Monitoring and Reaction Framework!
Visit us at zztat.net | Support our Indiegogo campaign at igg.me/at/zztat
| @zztat_oracle
