How to configure/track the HPC cluster debugger? RRS feed

  • Question

  • I am having trouble with using the cluster debugger. When I try to debug a SOA service, Visual Studio starts not responding. I investigated this by looking at the SoaDebugger log found in "C:\Users\<username>\AppData\Local\SoaDebugger" and the only line in the logfile is Wait until cleanup completes or timeout.How can I change the timeout settings of the debugger or cleanup manually? Or better yet, what does this cleanup mean as there is nothing ni the ServiceRegistration folder and deployment paths on the cluster.

    I am using HPC Pack 2008 R2 SP1, Visual Studio 2010 and SOADebugger_VS2010. I should also say that this problem only happens with a custom service, the sample services work fine.

    Any idea on how to track and fix this problem will be greatly appreciated.

    EDIT: At times I also get the following error when the debugger tries to deploy: The device is not ready. This also happens only for my custom build service, not for the samples provided by Microsoft. How can I get more info or fix this problem?

    EDIT: The device is not ready problem was related to some configuration files I wanted to deploy together with my service that were actually not available.

    • Edited by frunzaa Monday, July 25, 2011 8:51 AM
    Wednesday, July 20, 2011 9:46 AM

All replies

  • Hi,

    Did VS hang before or after you run your client? Did your client close the service before it exits? Is the job finished/cancelled when your application exits?

    Also, 1 simple work around is to uncheck the "clean up service" option in the project property page/HPC tab and run it again. Let me know if VS will hang or not in this case.


    Friday, July 22, 2011 1:37 PM
  • Hi,

    VS actually hangs before the deployment is done, so the client is not run yet. This actually happens after a run that was stopped in the middle of execution because of some errors, so session.Close() was not called (if that is what you mean by closing the service). The job stopped in the middle of execution is in the cancelled state.

    If I uncheck the "clean up service" option in the project property page/HPC tab the bahavior doesn't change (VS still hangs waiting for clean up to complete).




    Monday, July 25, 2011 8:49 AM
  • Hi

    According to your words "This actually happens after a run that was stopped in the middle of execution because of some error", I think the vs debugger is waiting for the previous job to terminate. Could you check the "All Jobs" and "Clusrun Commands" view in the HpcClusterManager to see if the previous soa debugger related jobs are running. You can cancel them.




    Tuesday, July 26, 2011 1:55 AM
  • Thanks for the replys. I checked the "All Jobs" and "Clusrun Commands" views in HpcClusterManager and all the jobs and clusrun commands are either in the Finished or Canceled state.

    Wednesday, July 27, 2011 8:49 AM
  • If all the jobs are terminated, I would suggest you to restart vs to have a try.


    Thursday, July 28, 2011 1:30 AM
  • Hi, I work with frunzaa and have been trying to solve this problem for a few days now. I have found a solution and just want to write it here in case anyone has the same problem in the future.

    The problem: When you start debugging, Visual Studio hangs. The only line written to the log file is Wait until cleanup completes or timeout.

    Solution: If we get the above problem, the only solution to getting a successful deployment in the future has been to restart the head node. Also, when Visual Studio hangs it is not possible to close Visual Studio, or even to kill the process, so we have had to restart our dev machines. When we can successfully deploy once, after each deployment we restart the service "HPC Job Scheduler Service" on the head node. Then we can deploy repeatedly, though I have no idea why this works.

    I have also noticed that when Visual Studio hangs before deployment, a single file does get deployed (but not any of the supporting files such as referenced dll's). The file that is deployed is the DLL of our service, and is written to //<headnode>/CcpServiceRegistration/<username>/<projectname>.

    Friday, July 29, 2011 6:44 AM