> > Server: NT Server 4.0 (SP2), SQL Server 6.5 (SP2)
> > Client: NT Workstation 4.0 (No SP), Client app in PB5 & iSQL/w
> > Clients can run 1-3 queries fine, then client apps die.
> > iSQL/w generates Dr. Watson.
> > PB5 app returns error message:
> > Bad token from SQL Server: Datastream processing out of
synchronization.
> Steve,
> 1. Have all the clients had their db-lib/obdc drivers updated to 6.5 SP2
> level?
> 2. This problem happens far more often with the tcp-ip protocol,
especially
> with sockets. Can you use named-pipes instead? In any case a network
trace
> will probably be needed to see where the error is.
> Protech Computing Ltd (MS Solution Provider)
> Using Virtual Access 3.52 build 159c (32-bit) on NT 4.0 SP2
We are using named pipes and I have not had the opportunity to check the
dblib/odbc versions. A check of the event log shows no stop errors or
strange happenings.
In the interest of functionality (not my decision, mind you) SQL was moved
to an NT Server 3.51 (no SP) machine. Things are running fine, but the new
machine only has 32 MB RAM and things are going slow.
I noticed this new feature in SP2 for NTS 4.0:
"Microsoft also added two new API calls to NT 4.0, using SP2 as a delivery
vehicle. These two calls, ReadFileScatter() and WriteFileGather(), let
programs pull in data from an uncached file and store it in a discontiguous
buffer, and vice versa. They also give disk I/O-intensive applications such
as SQL Server and Internet Information Server (IIS) a serious performance
boost because you can now move data from memory to disk in one pass."
I am beginning to wonder if this is the culprit, or at least a contributing
factor.
Still trying to figure out why.
Steve Kastl