M
mfreeman
I am trying to debug an intermittent error that my client is seeing in
their web logs from an internal web service. I am using Web Service
Studio 2.0 to eliminate the actual calling application as a potential
cause of the problem. There are no proxy servers involved. We are
running the latest SP of Windows 2003 and .NET 1.1 on ther servers. The
services being called do simple queries in an Oracle database and
should return very quickly.
When I call one service the first response is quick and successful, as
are each additional call to that same service. If I change to another
service, the first call to that second service triggers the "underlying
connection" error that we are seeing in production. Additional calls
to the second service then work, even with different inputs.
We had someone from the infrastructure group use AppVantage to perform
a trace and we were seeing that my localhost would change ports only
after each failure. The failures occurred when I would first invoke a
new service because it was using the old port from the previous service
call. As soon as Web Service Studio got the failure, it would generate
a new port and work just fine. It appears to us that something doesn't
like to receive requests from more than one service over the same port,
from the same client.
The next test was calling multiple servers and multiple services from
the same instance of Web Service Studio. We were seeing the same
problem from the first test, but it didn't seem to have a problem with
ports when I would select a new server. So it would automatically
generate a new port when I would select a new server, but not when I
would select a new service. So say I called Service A on Server 1, it
would work just fine on Port 100. Then I call Service A on Server 2,
it would work fine on Port 200. Then I call Service B on Server 2 on
Port 200, I get the error. Call again, Port 201 is used and it works
fine. Then I call Service B on Server 1 on Port 100, and get the error
again.
Many of the posts I have seen regarding this type of error want the
client code modified to use HTTP 1.0 instead of 1.0 and/or to set
keep-alives to false in an overload of the GetWebRequest method. I am
unaware of a way to modify Web Service Studio in this way. Isn't there
a solution that does not involve changing the client code?
I would appreciate any constructive input on this issue.
Thanks,
- Mark
their web logs from an internal web service. I am using Web Service
Studio 2.0 to eliminate the actual calling application as a potential
cause of the problem. There are no proxy servers involved. We are
running the latest SP of Windows 2003 and .NET 1.1 on ther servers. The
services being called do simple queries in an Oracle database and
should return very quickly.
When I call one service the first response is quick and successful, as
are each additional call to that same service. If I change to another
service, the first call to that second service triggers the "underlying
connection" error that we are seeing in production. Additional calls
to the second service then work, even with different inputs.
We had someone from the infrastructure group use AppVantage to perform
a trace and we were seeing that my localhost would change ports only
after each failure. The failures occurred when I would first invoke a
new service because it was using the old port from the previous service
call. As soon as Web Service Studio got the failure, it would generate
a new port and work just fine. It appears to us that something doesn't
like to receive requests from more than one service over the same port,
from the same client.
The next test was calling multiple servers and multiple services from
the same instance of Web Service Studio. We were seeing the same
problem from the first test, but it didn't seem to have a problem with
ports when I would select a new server. So it would automatically
generate a new port when I would select a new server, but not when I
would select a new service. So say I called Service A on Server 1, it
would work just fine on Port 100. Then I call Service A on Server 2,
it would work fine on Port 200. Then I call Service B on Server 2 on
Port 200, I get the error. Call again, Port 201 is used and it works
fine. Then I call Service B on Server 1 on Port 100, and get the error
again.
Many of the posts I have seen regarding this type of error want the
client code modified to use HTTP 1.0 instead of 1.0 and/or to set
keep-alives to false in an overload of the GetWebRequest method. I am
unaware of a way to modify Web Service Studio in this way. Isn't there
a solution that does not involve changing the client code?
I would appreciate any constructive input on this issue.
Thanks,
- Mark