K
Kevin Clements
I have a Win Forms fat client talking to a web servce, both authored
here. They're using WSE 2.0 (Tech Preview) to pass authentication
information.
In my development environent, when the client and web service are
running on the same box, all is well.
When I've pushed the web service out to it's intended home, I'm
getting random delays in execution of the service. In an attempt to
troubleshoot this, I've added a test call to my service that recieves
in the parameters of the real method, and simply creates one of my
response structures, populating a string, and int, and a boolean with
hardcoded values and returns.
An invocation of this test method from the fat client normally takes
less than a second, but sporadically takes 15-20 seconds. This full
15-20 second interval WILL appear in the web server logs as
"time-taken" for the request.
A packet capture of traffic to the web server will show complete
receipt of the 2 or 3 packets required to hold the request document,
then 15 seconds later will show the response. The string in the test
response packet is a timestamp, which makes it appear that the web
method is getting control after something has taken the majority of
this lag time doing something prior to my web method getting control.
The only suspicious element I find is that when I recieve the last
inbound packet for one of these requests at the web server, the web
server seems to be sending a broadcast of a NETLOGON packet.
The pattern of successful (1 second) to slow (15-20 second) requests
seems non-deterministic. There may be a pattern, but I haven't found
it yet - I can send 15 requests in a row and have #16 be slow, or get
to #50 and then get a slow one.
The machine used for the testing is otherwise generally quiet as this
is going on - CPU usage < 10%, virtually no traffic on the box.
Any suggestions as to where to look for the cause of the time between
receipt of the last packet of the request and my web method getting
control?
here. They're using WSE 2.0 (Tech Preview) to pass authentication
information.
In my development environent, when the client and web service are
running on the same box, all is well.
When I've pushed the web service out to it's intended home, I'm
getting random delays in execution of the service. In an attempt to
troubleshoot this, I've added a test call to my service that recieves
in the parameters of the real method, and simply creates one of my
response structures, populating a string, and int, and a boolean with
hardcoded values and returns.
An invocation of this test method from the fat client normally takes
less than a second, but sporadically takes 15-20 seconds. This full
15-20 second interval WILL appear in the web server logs as
"time-taken" for the request.
A packet capture of traffic to the web server will show complete
receipt of the 2 or 3 packets required to hold the request document,
then 15 seconds later will show the response. The string in the test
response packet is a timestamp, which makes it appear that the web
method is getting control after something has taken the majority of
this lag time doing something prior to my web method getting control.
The only suspicious element I find is that when I recieve the last
inbound packet for one of these requests at the web server, the web
server seems to be sending a broadcast of a NETLOGON packet.
The pattern of successful (1 second) to slow (15-20 second) requests
seems non-deterministic. There may be a pattern, but I haven't found
it yet - I can send 15 requests in a row and have #16 be slow, or get
to #50 and then get a slow one.
The machine used for the testing is otherwise generally quiet as this
is going on - CPU usage < 10%, virtually no traffic on the box.
Any suggestions as to where to look for the cause of the time between
receipt of the last packet of the request and my web method getting
control?