FreeTDS Development Group

[Matthias Apitz <guru AT Sisis.de>]

El día Thursday, April 20, 2006 a las 11:32:09AM +0200, guru AT Sisis.de 
escribió:

> El día Thursday, April 20, 2006 a las 10:37:26AM +0200, 
> michael.peppler AT bnpparibas.com escribió:
> 
> > I see.
> > 
> > The next step might be to enable Sybase auditing, which can be made to log
> > all SQL statements to the sybsecurity database.
> > 
> > Michael
> 
> Thanks for the hint. We'll give it a try to log on 'User level' 'cmdtext',
> even with the risk, that ASE in case of such a performance dropout
> has no time for this, we'll see. I somehow, from a lot of monitoring,
> think that we just hunting some (unwanted) full table scans or other
> operations where our application instructs (unwanted again) ASE to copy
> over tables to temp for or'ing and others.
> 
>       matthias

This is only to close this threat. After monthes of digging and logging
it turned out that setting in Sybase ASE

        allow sql server async i/o = 0

avoids that problem where from time to time normal SQL statements
caused a 3-4 second high CPU-load from the dataserver without responding
to the client on TCP/IP.

The beast is dead and the customers are happy.

        matthias
-- 
Matthias Apitz
Manager Technical Support - OCLC PICA GmbH
Gruenwalder Weg 28g - 82041 Oberhaching - Germany
t +49-89-61308 351 - f +49-89-61308 399 - m +49-170-4527211
e <m.apitz AT oclcpica.org> - w http://www.oclcpica.org/ http://guru.UnixLand.de/