Is this a bug?

[seal]

Hi fellows,

I'm trying to find a solution to the following query almost 2 days, and feel found something but do not know the rationale behind it;((

DB is Oracle Version 9.2.0.1.0 and my table is as follows

TABLE
(
MID NUMBER(1),
SID NUMBER(4),
TYPE NUMBER(1),
NO NUMBER(18),
SEG NUMBER(2),
P_TYPE NUMBER(1),
Z NUMBER,
C NUMBER,
D NUMBER
)

and contains 9.200.000 records.On this table the following query is run

create table test2 as
select /*+PARALLEL(test,10)*/decode(mod(mid,2),1,'O','I') as mid,seg,p_type,
count(distinct no) as scnt,sum(c) as gc,sum(d) as gd
from TEST
group by decode(mod(mid,2),1,'O','I'),seg,p_type

At each run this SQL generates different scnt, gc and gd. After numeruos tries I guessed that PARALLEL clause causes this fault (may be because of threads? or processes? for each distinct parallel server) and removed it. After removal, it always gave the same value;(( Is this a misusage of PARALLEL (The group by expression is always made by different threads and so different groupings are made each time??)? Or is this a bug? or What is the reason, please help)

Best Regs..
seal

[Alien]

Hi Seal,

There used to be a bug about unpredictable results when the parallel degree was above a certain value. Can you try with lesser degrees of parallelism ( Where is the spell-check for that one?)

I'll try to find the bug number for you for extra info.

Regards,

Arian

[Alien]

Ok. Bad luck. I was thinking about
Bug: 1271545. Voor 8i, solved in 9i.

But I also found bug: 4188467.

So I'd suggest you contact support about it. Try to make a small test-case for them. And mention if the degree of parallel makes a difference or not.

Regards,

Arian

[seal]

Hi Arian

I changed the parallel to NOPARALLEL and then everything is ok;(( And also I tried the query on smaller data by extracting the results that were errenous (Good luck, I have another table to make sure the results are true or not, if i have not, then it would be a real tragedy;(( And on smaller data or with NOPARALLEL the results are as I expect. Thank you for your answer, I will seek that bug and also ask to Oracle

Thanks again with my best regards

seal