Re: Weird! select * _AND_ large recordsize 'hangs' connection???

John,

Your though is not that far of, it's just that I wouldn't know which buffer you hinting at nor where to set it's size...any tips would be appreciated since this is the first time I've experienced this...

As far as monitor programs go, care to suggest a good one? I've tried throwing EM's Performance Pack at it but I could not see anything ordinary (probably because it's suffering from the same problem)

TIA
Raj Gabrielse

TurkBear wrote:

> As a first thought, the size of even one row in your large table may
> be bigger than the buffer allocated for returned values - Also, you
> could use a monitor program to watch the RDBMS and see what's
> happening when the query is executing... This may help narrow down the
> 'location' of the problem.
>
> Just a thought ( or a reasonable facsimilie of one)
>
> Reply to : john.greco_at_dot.state.mn.us
>
> Raj Gabrielse <Raj_at_Botany_Bay.Con> wrote:
>
> >Hi all,
> >
> >I have a weird problem and am hopelesly lost in where to look to find
> >the cause. The problem occurs on Oracle 7.3.2.3.0 on Novell
> >3.12 (patched up to date) and 16bit clientsoftware on *all* W95 WSs,
> >talking over a SPX/IPX connection. BTW, it's not only our program that's
> >affected, it also happens with SQLPlus, and
> >all the ODBC connection testtools I have.
> >
> >A select * from any size table (tested up to 10000 recs) with:
> >
> > _small_ records (i.e. around 100 bytes), everything is ok.
> >
> > large records (i.e. over 1000 bytes), there's no response and
> > our clientsoftware, including SQLPlus, 'hangs' (doesn't
> > timeout)
> >
> > ad criteria, to drastically limit the the number of records
> > returned, and we get immediate response
> >
> > change the * into a single, large column, char(255), and remove
> > the criteria, again no repsonse
> >
> > pick a single field or a combination of fields so the returned
> > 'recordsize' is around 100 bytes again and it works ok again
> >
> >It gets weirder, entering the exact same queries at the
> >serverconsole produces lightning results for every possible
> >rowlength I could put together.
> >
> >Sadly, I'm not Oracle proficient enough to know _where_ I have
> >to look. But because it happens on every WS, when there's a
> >protocol involved, this has to be caused by something related to the
> >network layer, but where?
> >At least that's what _I_ think, but I may be wrong.
> >
> >Now I got three questions since I'm at a loss here...
> >
> >What's happening? IOW, why is there 'no response' after a
> >select, that would result in a high number of large rows. While the same
> >statement, selecting only a part of each row, producing
> >the same number, but smaller rows , comes back with results
> >instantly...(like <wink>)
> >
> >What can I do to *see* where the communication breaks down, (tried
> >listener trace, but that seems only to track
> >connectionattempts). I mean, the chain obviously is breaking
> >somewhere when a select * from table of large rows ( > 150-200
> >bytes) does not return data.
> >
> >Why me?
> >
> >Raj Gabrielse
> >
> >PS. Please reply to group, email address is bogus
> >PPS. if you feel the urge to reply via email anyhow,
> > remove not_here from Reply_To
> >
Received on Mon Jan 12 1998 - 00:00:00 CST