Fluids

Fluids

Unable to Run Parallel Simulation in Fluent 19.2

    • rj33320
      Subscriber

      Hi, I'm trying to simulate a flow analysis in fluent 19.2 (Standalone). It's a polyhedral mesh with about 17M cells. I,m using 128 GB RAM, 40 processes. The problem is, I could read the case in parallel mode, but whenever I try to initialize it, I get a series or errors as:


       


      Node 21: Process 13748: Received signal SIGSEGV.??


      then this:


      999999: mpt_accept: error: accept failed: No error


      then this:


      error: accept failed: No errorMPI Application rank 0 exited before MPI_Finalize() with status 2?****************************************************************************?* hwloc has encountered what looks like an error from the operating system.?*?* Socket (cpuset 0xfff000ff) intersects with Socket (cpuset 0x000fffff) without inclusion!?* Error occurred in topology.c line 940?*?* Please report this error message to the hwloc user's mailing list,?* along with any relevant topology information from your platform.?


       


      The case is running perfectly fine in 19.2 serial. 


       


      I tried running the case in Fluent 17 in another machine. It ran fine on serial as well as parallel. I also tried to run the case in different machine running fluent 19.2 parallel, got the same error. Are there some additional settings for 19.2 parallel? Why is it not running on 19.2 parallel.

    • DrAmine
      Ansys Employee

      The hwloc message is only a warning and you disable writing that message ( HWLOC_HIDE_ERRORS env. variable)


      Now regarding the mpi error itself: that might be due to manifold reasons. Might that one node crashed or ran diverged and node-0 or host lost connection to the nodes.


      You said you can launch the case in 19.2 in Serial. If you choose the option "serial" under Fluent Launcher that is not real serial process but a parallel process with single node.

    • rj33320
      Subscriber

      After the hwloc warning, it shows fl process could'nt be started. 


       


      Regarding the mpi, it's not just one or two nodes, for every node it gives error.


      What could be the possible solution?

Viewing 2 reply threads
  • You must be logged in to reply to this topic.