Using MS MPI in Client Server Model

All replies

• 

  

  0

  Sign in to vote

  Hi,

  One way to resolve your problem is:

  1) Build 2 applications. Both uses MSMPI and share the same COMM WORLD in the API.

  2) Run the 2 application at the same time like below:  

  job submited /numodes:3 /askednodes:server1,node1,node2 mpiexec -hosts 1 server1 1 win-form-app.exe : -hosts 2 node1 1 node2 1 console-app.exe

  where:

  - totally 3 nodes are used. each node runs  1 process. If you want to run more than 1 process on certain node. you can do -hosts 2 node1 M -node2 N ...

  - server1 will run your win form application; node1 and node2 will run the console application.

  Does this answer your question?

  Liwei

   

  Tuesday, May 18, 2010 2:18 PM
• 

  

  0

  Sign in to vote

  Thanks Liwei,

  I will try this...

  Friday, May 21, 2010 1:17 PM
• 

  

  0

  Sign in to vote

  Hi Liwei,

  Let define the problem in a more clear way:

  I have two executable files "client.exe" and "server.exe". We have a cluster setup with 1 Head node and two compute nodes(The head node is also configured as a compute node). Now I want to run one instance of "client.exe" on the head node and as many instances of "server.exe" as the no.of compute nodes available. In short I want to execute one instance of "server.exe" on each compute node.

  Now I want to achieve communication between the client on head node and all the server's running on compute nodes.

  I have tested client.exe and server.exe on a single machine and they are running absolutely fine.

  As you have suggested I have tried to achive this by submitting a job:

  The command i have used looks like this

  "mpiexec -n 1 -host <headnode hostname> : -n 1 -host <computenode1 host name> : -n 1 -host <compute node 2 host name>"

  I have assigned the head node and all the compute nodes exclusively for this job and tried to execute it but it is giving some error which looks as shown below:

  job aborted:
  [ranks] message

  [0] terminated

  [1] fatal error
  Fatal error in MPI_Comm_dup: Other MPI error, error stack:
  MPI_Comm_dup(136)..............: MPI_Comm_dup(MPI_COMM_WORLD, new_comm=0x00F8AEE0) failed
  MPIR_Comm_copy(500)............:
  MPIR_Get_contextid(248)........:
  MPI_Allreduce(609).............: MPI_Allreduce(sbuf=MPI_IN_PLACE, rbuf=0x0023F058, count=32, MPI_INT, MPI_BAND, MPI_COMM_WORLD) failed
  MPIR_Allreduce(211)............:
  MPIC_Sendrecv(170).............:
  MPID_Send(66)..................:
  MPIDI_CH3_SendEager(85)........:
  MPIDI_CH3I_VC_post_connect(376):
  MPIDI_CH3I_Sock_connect(304)...: [ch3:sock] rank 1 unable to connect to rank 2 using business card <port=54836 description="10.66.194.131 computenode1 " shm_host=computenode1 shm_queue=1048:592 >
  MPIDU_Sock_post_connect(1118)..:
  save_valid_endpoints(1047).....: unable to connect to 10.66.194.131 computenode1  on port 54836, no endpoint matches the netmask 10.66.193.0/255.255.255.0

  [2] terminated

  ---- error analysis -----

  [1] on computenode1
  mpi has detected a fatal error and aborted C:\MPITEST\server.exe

  ---- error analysis -----

   

  Please help me on this.

  Thursday, June 3, 2010 1:05 PM
• 

  

  0

  Sign in to vote

  Hi Humayun,

  It looks like there is problem in your network communications. If you have installed the Microsoft HPC pack, please run:

  cluscfg setenvs MPICH_NETMASK=10.66.0.0/255.255.0.0

  and then try your test again?

  Thanks,

  James

  • Proposed as answer by Don Pattee Wednesday, October 6, 2010 11:44 PM
  • Marked as answer by Don Pattee Wednesday, January 12, 2011 2:47 AM

  Wednesday, August 4, 2010 9:31 PM