Running a program on a remote machine as part of continuous integration

Question

We use TeamCity, nant and psexec to run a command on a remote machine as part of the release packaging. Everything works fine when I run the nant from the console but when running from teamcity psexec hangs (freezes) 50% of the times.
I looked through many forums and there seems to be workarounds that increase complexity of the call and involve loosing the output and the errorcode of the command.
Does anyone know an easier way to run a command on a remote machine?
I don't mind setting up some application on the remote machine, like a telnet server, any advices on what to do?
Thanks

Answer 1

I have solved this issue with a combination of RemCom and a custom MSBuild task called ExecParse.

RemCom, because it doesn't do odd things with STDOUT (thus hanging the build). We used, and ExecParse to capture the output of the remote task, and parse the Exit Code from the output, because the standard MSBuild Exec task does not capture output. Some NAnt equivalent that captures the output would work.

I've detailed this in a blog post: "Continuous Integration: Executing Remote Tasks with TeamCity, MSBuild, RemCom, and ExecParse"

Answer 2

PsExec does some funky things with the standard input/output, and invoking this from Java (which TeamCity is built on) raises all kinds of problems and stability issues. psexec -d did not work wither.

I solved it by using Powershell in Team City.

The script below stops an IIS 7 ApplicationPool on a remote server:

[string]$HostName = "myWebServer"
[string]$Cmd = "C:\Windows\System32\inetsrv\appcmd.exe stop apppool MyMainAppPool” 
Invoke-WmiMethod -class Win32_process -name Create -ArgumentList ($Cmd) -ComputerName $HostName

More about it on my blog: http://blog.degree.no/2012/03/executing-commands-and-programs-on-a-remote-machine-using-powershell/

Answer 3

How about putting a (nant) time-out on the psexec and repeat the call until no time-out happens?

Answer 4

I use PSExec with the -d option (don't wait for it to finish) and capture the return code. The return code when you used -d is the process ID of the process running on the remote system. then I use PSList to poll the remote system for the process ID until I don't find it on the remote system any longer.

Answer 5

What happens if you setup TeamCity build agent on remote machine and let it perform the operation locally, passing it the binaries with "Artifact Dependencies"?