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

Make sure that the SQL*Net packet size is less than the MTU on your network, is there any router in the way that might be fragmenting packets ? Make sure that the frame size in the client protocol configurations is suitable for your network.

Just a guess...

Raj Gabrielse wrote in message <34B3F474.AAFA593B_at_Botany_Bay.Con>...
>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 Fri Jan 09 1998 - 00:00:00 CST