If Windows Communication Foundation (WCF) implementation has its own idiosyncrasies, named pipes provider is the champion. First, let’s start with the name of the provider. Named Pipes in Windows can be used to communicate between process on the same machine or between different machines across a network. WCF only implements the on-machine part of it. Named Pipes name has the format \\ServeName\pipe\PipeName and for on-machine pipes you can use \\.\pipe\PipeName. WCF, if you remember, use the uri format net.pipe://host/path but unlike it may seem, the actual pipe name (Windows Object) will not be anything near \\host\pipe\path but rather a randomly generated GUID that will be different every time you start your host. When troubleshooting a TCP host/client in WCF, you can simple use netstat /ano to verify if the port is there and which process is listening to it. If it is not listening you know that the host is either not running or running in a different port. For Named Pipes, you will not be able to identify whether the host is listening or not because you don’t have the pipe name.
I will explain how WCF client can figure out the pipe name. Instead of writing a very large paragraph no one will read let’s put it all in perspective with a scenario: I have created a host with a named pipe biding. The endpoint is net.pipe://localhost/TradeService/Service1. I deploy my solution in my server as Windows Service and one day I cannot connect to my host. I check the ABC (Address-Biding-Contract) of my client and it is as expected. The same happens in the host. My objective is identify if the named pipe is available.
First I download SysInternals tools (http://technet.microsoft.com/en-us/sysinternals/default.aspx) and install everything on c:\systenternals. SysInternals include a tool to list all pipes available in the machine. I start a new cmd prompt with Administrator Privileges, move to my SysInternals folder a run PipeList.exe.
Read more: Rodney Viana's (MSFT) Blog
Posts
0 comments:
Post a Comment