Xref: alice comp.databases.oracle:76765 comp.databases.oracle.misc:49097 comp.databases.oracle.server:77495
Path: alice!news-feed.fnsi.net!netnews.com!howland.erols.net!newspeer.monmouth.com!nntp2.deja.com!nnrp1.deja.com!not-for-mail
From: lempert@my-deja.com
Newsgroups: comp.databases.oracle,comp.database.oracle,comp.databases.oracle.server,comp.databases.oracle.misc
Subject: Re: Oracle, multi thread and heavy load leads to access violation in ORA73.dll
Date: Sun, 12 Dec 1999 13:12:54 GMT
Organization: Deja.com - Before you buy.
Lines: 50
Message-ID: <83070k$b4s$1@nnrp1.deja.com>
References: <384FBE0F.36AF6C57@ej.insta.fi>
X-Article-Creation-Date: Sun Dec 12 13:12:54 1999 GMT
X-Http-User-Agent: Mozilla/4.7 [en] (WinNT; I)
X-Http-Proxy: 1.0 x25.deja.com:80 (Squid/1.1.22) for client 192.117.141.162
X-MyDeja-Info: XMYDJUIDlempert

In article <384FBE0F.36AF6C57@ej.insta.fi>,
  Suuronen Henri <henri.suuronen@ej.insta.fi> wrote:
> We have the following environment:
> - Oracle Server 7.3.4.5 on HP-UX 10.20
> - Oracle Client 7.3.4.5 on Windows NT 4.0 WS + SP5
> - Direct Oracle Access 3.3.1
> - Delphi 4 + update 3
>
> The problem is that when the load rises (50 transactions/sec) the
> client-side ORA73.dll starts to throw access violations. We have
tracked
> this one with Oracle support to no avail (upgraded, patched etc.). The
> problem does not occur if the load is lighter. The same queries are
> repeated
> over and over again so the problem shouldn't be on our application.
>
> Problem persists on every machine we have tried and is repeatable.
What
> makes it nasty is the fact that ORA73.dll does not recover until
> application is restarted (and thus dll is initialized). Also we
haven't
> been able to track down the responsible party, it could Oracle server
or
> client, DOA or our application.
>
> We have traced SQL*Net and found out that when a select query is
> executed atleast four relevant packets are sent/received:
> - Query itself (to transport)
> - Received column names (from transport)
> - Request for the data (to transport)
> - Receive the data (from transport)
>
> When the error occurs the last two packets are missing. Don't know if
> this means anything because these could just be symptoms not the
cause.
>
> Any help or hint will be greatly appreciated!
>
> --
> Henri Suuronen
>

Hi
for a load of  50 transactions/sec I would expect at least 128MB of RAM
and IDE disk of UDMA/66 disk+controller. Heavy load yields heavy
hardware.


Sent via Deja.com http://www.deja.com/
Before you buy.
