none
joining compute node to domain during deployment

    Question

  • When provisioning compute node the log reports the following:
    Exit code:1355 the specified domain cannot be contacted or does not exist

    occurs after initializing deployment operations

    the deployment does create the domain account for the compute node with a account supplied during HPC2008 setup on head node
    When I connect to compute node console I can log on as a local admin , can ping head node , resolve names and get to the internet.
    The head node in on local network with 2nd nic connected to compute node network using 192.168.x.x  private addresses.
    Any ideas?  must admit new to HPC 
    Friday, January 22, 2010 3:29 PM

Answers

  • I'm sorry to hear that you are having problems.

    It really shouldn't be that difficult.  However, you have modified the default 'wizard' settings for your private network which takes into consideration that we enable routing on your head node for topology 1.  If you are doing this on a corporate network, then it might be failing due to one or more domain group policy objects or IPsec policies.  If you could provide "ipconfig /all" for both HN and CN then I might be able to assist. 

    Thanks,
    --Brian

    PS...  you could uniformily modify IP numbers to preserve your identity or contact me directly (at 'bkbroker at hotmail dot com').
    Wednesday, February 3, 2010 2:13 PM

All replies

  • Hello, a few quick questions.
    Does the deployment account have permissions to carry out domain joins?
    Are you able to resolve the fqdn of the domain you're attempting to join from a compute node?
    Are you able to manually join a compute node to the domain after deployment has completed?
    Have you specifically defined the Domain in the 'Join Domain' step of the node template, or have you left it blank?
    Cheers
    Dan


    Monday, January 25, 2010 2:46 PM
  • Dan,

    Yes
    Yes
    No
    Yes

    The compute node is left in a runnable state and I am able to logon with local admin account I setup for the build.
    Unable to manually join domain but can resolve dsn names , intenal or external internet names.

    Would you know how the compute finds the domain controller?
    THe computer account is created in the domain during the deployment process but this happens before the node joins the domain
    so I would have to say the head node is creating the compute node account in the domain.

    confused?

    Friday, January 29, 2010 6:04 PM
  • The compute node 'create' is done by the head node but the CN join must be done by the node itself.  If the compute node cannot join manually then you must examine your environment to determine why. Which topology is your cluster using?  "Ipconfig /all" will probably help most if you cannot ping your DC.   Can you share IP config info?

    --Brian
     

    Saturday, January 30, 2010 11:13 AM
  • Brian ,

    the topology is type 1   head node connects to enterprise with another nic to private network (192.168.5.x), compute nodes only have a private network connection

    Think I have solved the compute node joining domain, I specified an enterprise network dns server in the dhcp server config on head node
    so that compute nodes get the enterprise dns server as first dns entry with head node as second dns entry.
    The HPC network setup wizard had defaulted to the head node for compute node dns server and when various names would resolve
    correctly I got off on the wrong track.

    so ... compute nodes join domain but fail to provision on the cluster

    I will post the provisioning log next if you could take a look at it
    Monday, February 1, 2010 2:05 PM
  • so now one node actullay built and joined domain but i do not know why

    building second node and after the HPC pack is installed recieve this message

    Time Message
    2/1/2010 10:17:20 AM The Management service encountered an error while performing a change on this node. Access is denied to user 'NT AUTHORITY\ANONYMOUS LOGON'. Check the operation log in the Administration Console for more information. 
     
    ... 3 times before rollback log entries appear

    I cannot ping this compute node but can ping the first one

    any ideas? 

    Monday, February 1, 2010 4:05 PM
  • If I go to the node that failed to loin the cluster and logon locally ..... I can ping the head node.

    but I cannot ping the failed node from the head node...?

    If I create a share on the failed node ......

    some thing happens and I can then ping the fialed node from the head node .....

    is this really that difficult?   .... argh
    Monday, February 1, 2010 4:44 PM
  • I'm sorry to hear that you are having problems.

    It really shouldn't be that difficult.  However, you have modified the default 'wizard' settings for your private network which takes into consideration that we enable routing on your head node for topology 1.  If you are doing this on a corporate network, then it might be failing due to one or more domain group policy objects or IPsec policies.  If you could provide "ipconfig /all" for both HN and CN then I might be able to assist. 

    Thanks,
    --Brian

    PS...  you could uniformily modify IP numbers to preserve your identity or contact me directly (at 'bkbroker at hotmail dot com').
    Wednesday, February 3, 2010 2:13 PM