July 2016 - Nektar-users - imperial.ac.uk

Problem in getting a parallel case to run
by Amitvikram Dutta 28 Jul '16

28 Jul '16

Hi all, I'm trying to run a simple case on a NACA airfoil on a cluster. The problem is that it keeps failing to run in parallel. The job runs perfectly in serial, but that's hardly any consolation. I know, thr parallel implementation of nektar++ has correctly been installed don the cluster, since I managed to run a test file (tet_channel_m8_par.xml) successfully. I'd be greatful if anyone could point out any obvious mistakes that I'm making in the setup of my .xml file. For further reference I have attached the .xml file for my problem (naca0018_new_par.xml), the error message that I get from the cluster(0018_pw.pbs.e.....) and the test file that I ran successfully in parallel. Sincerely, -- *Amitvikram Dutta* *MASc Candidate* *Graduate Research Assistant * *Okanagan CFD Laboratory* *University of British Columbia | Okanagan Campus*

1 0

Traction-Free boundary condition
by mohammadali alidoost 24 Jul '16

24 Jul '16

Dear Nektar++ Developers, We are going to use a Traction-Free boundary condition as an outflow in our simulation. In addition, we intend to use “Velocity Correction Scheme” of “Incompressible Navier-Stokes Solver”. As far as we know, the Traction-Free boundary condition contains two parts which should be satisfied simultaneously: zero velocity gradient and zero pressure. Based on the scheme we are going to use, we are wondering if you could tell us if it is necessary to implement both conditions in Nektar++ or we just need to set zero pressure condition alone. Kind Regards Mohammadali Alidoost

2 1

Gitlab update
by David Moxey 23 Jul '16

23 Jul '16

Dear all, Just a quick note that I've updated gitlab this morning. As usual there are a few interface tweaks. Let me know if you have any issues. Cheers, Dave -- David Moxey (Research and Teaching Fellow) d.moxey(a)imperial.ac.uk | www.imperial.ac.uk/people/d.moxey Room 364, Department of Aeronautics, Imperial College London, London, SW7 2AZ, UK.

1 0

Recent updates
by David Moxey 21 Jul '16

21 Jul '16

Dear all, A couple of recent updates you should be aware of: - as of this morning, FieldConvert modules have now been moved to a separate library called FieldUtils. This allows for the use of some of the post-processing modules to be run whilst the simulation is running through the use of a filter. See the MR for more details: https://gitlab.nektar.info/nektar/nektar/merge_requests/589 - in MR !654, shared filesystems (i.e. those where all processes share a common filesystem) are now automatically detected and the --shared-filesystem command line option has been removed. Cheers, Dave -- David Moxey (Research and Teaching Fellow) d.moxey(a)imperial.ac.uk | www.imperial.ac.uk/people/d.moxey Room 364, Department of Aeronautics, Imperial College London, London, SW7 2AZ, UK.

1 0

Coupling CompressibleFlowSolver and APESolver to simulate the acoustic of a NACA-profile
by Selbach, Fabian 20 Jul '16

20 Jul '16

Hello, i have a question about coupling two solvers. First of all i make a simulation with a NACA-profile by using the NavierStokesCFE of the CompressibleFlowSolver. The solution of this i want to use to simulate the acoustic of its by solving the APE-equation. Is there an easy way to couple both solver and to simulate the acoustic of the flow around a NACA-profile? I need the acoustic source terms (c²*qc) for the APE-Solver to simulate the sound. I should get those by the solution of the NACA flow (by the CompressibleFlowSolver), but i actually i do not see a possible to get them. Is there a way or do i have to program it's on my own into the code of the CompressibleFlowSolver to handle and couple them into the APESolver? The difference between the solver-variables p, u and v ( in the CompressibleFlowSolver the absolute values and in the APESolver the fluctuations) i solved by calculating them in form of p(APE) = p(CFS) - pInf and so on. I hope somebody can give me an answer. Thank you already! Best regards Fabian

2 2

Simulation of flow over a NACA 0012 Airfoil
by Gipperich, Manuel 20 Jul '16

20 Jul '16

Hi all, I need help for my simulation of the flow over a NACA 0012 Airfoil. I try to simulate the flow with a mach Number of M=0.4 and a reynoldsnumber of Re=408000. In every test my simulation is terminated with the message: NaN found in variable "rhou", terminating NaN found in variable "rhov", terminating NaN found in variable "E", terminating. I started the simulation with the same mach number and a lower reynoldsnumber of 4080 and it worked. So my question is, is there any solution for that problem to fix the instability of the simulation? Here is my SolverInfo data which i entered in the .xml File. Is it possible that an other UpwindType will fix the problem? <SOLVERINFO> </SOLVERINFO> My simulation runs with a CFL-number of 0.7. Thanks for any hint. If you need more information, please let me know . Best regards Manuel Gipperich

2 1

Nektar++ v4.3.3 Released
by Chris Cantwell 16 Jul '16

16 Jul '16

Dear all, We are pleased to announce the release of Nektar++ v4.3.3. This release includes a number of further bug fixes and minor enhancements to the 4.3 release: **Library**: - Auto-detect a shared filesystem and removed --shared-filesystem option (!654) - Fix filters when using adaptive driver to avoid output being overwritten after each adaptive update (!588) - Minor fix to suppress Xxt output unless `--verbose` is specified (!642) - Fix of DirectFull solver in case where only Neumann boundary conditions are imposed. (!655) **FieldConvert**: - Fix to avoid repeated import of field file (!649) - Fix issue with C^0 projection (!644) - Fix verbose output when using --procid (!648) **NekMesh:** - Fix namespace issue in Star-CCM+ input header in NekMesh (!661) **CompressibleFlowSolver**: - Fix issue with residual output (!647) - Issues with 1D Euler solver fixed (!565) - Fix deadlocking issue with boundary conditions (!657) **Packaging**: - Fix NekMesh dependencies for DEB package (!650) - Fix PETSc build on newer linux distributions (!646) As always, if you encounter any problems please feel free to contact us on the mailing list nektar-users(a)imperial.ac.uk. Cheers, Chris -- Chris Cantwell Imperial College London South Kensington Campus London SW7 2AZ Email: c.cantwell(a)imperial.ac.uk www.imperial.ac.uk/people/c.cantwell

1 0

Problem 3D CompressibleFlowSolver
by Selbach, Fabian 13 Jul '16

13 Jul '16

Hello, i need your feedback about the simulation of a 3 dimensional flow around a cylinder or a NACA-profile with the CompressibleFlowSolver of the NavierStokesCFE or the EulerCFE. In every case the simulation starts, but always ends at the point of "Writing: "file_0.chk" " due to time-limit of the Cluster (limited to 1 day). I tried several times. The parameters, Solver Info and so on were shown (see below). In 2D there is not such a problem. So my question is, if it depends on the calculation capacity (CPU etc.) or what the reason for this? Actually iam calculating on a Cluster with 10 nodes and 120 CPUs. I think, that the calculation normally should be faster and another time-step should be done after one day or is it normal to a 3D-simulation with the compressible solver? So please can you give me an advice about the Problem? For example the 3D-problem of the cylinder with the CompressibleFlowSolver with the NavierStokesCFE produces this output after one day of calculation: (The example with the cylinder is exact the same you published at this page: http://www.nektar.info/compressible-flow/ ) Parameters: ALPHA = 0 CINF = 294.827 CP = 1004.7 FINTIME = 2 GAMMA = 1.4 GASCONSTANT = 287.058 IO_CHECKSTEPS = 50000 IO_INFOSTEPS = 1 KAPPA = 0.1 L = 1 MACH = 0.1 MU = 0.00275172 MU0 = 0.5 NUMSTEPS = 2e+06 PINF = 22600 PR = 0.72 RE = 3900 RHOINF = 0.364 SKAPPA = -2.3 THERMALCONDUCTIVITY = 3.83981 TIMESTEP = 1e-06 TINF = 216.29 TWALL = 216.29 UINF = 29.4827 VINF = 0 WINF = 0 Solver Info: ADVECTIONTYPE = WeakDG DIFFUSIONTYPE = LDGNS DRIVER = Standard EQTYPE = NavierStokesCFE EVOLUTIONOPERATOR = Nonlinear GLOBALMATRIXSTORAGETYPE = SmvBSR GLOBALSYSSOLN = IterativeStaticCond LINEARPRECONSOLVER = Xxt LOCALMATRIXSTORAGESTRATEGY = Sparse PRECONDITIONER = Diagonal PROBLEMTYPE = General PROJECTION = DisContinuous TIMEINTEGRATIONMETHOD = RungeKutta2_ImprovedEuler UPWINDTYPE = ExactToro VISCOSITYTYPE = Variable gs_setup: 240784 unique labels shared pairwise times (avg, min, max): 0.000148432 0.000120687 0.000157881 crystal router : 0.000947791 0.000929117 0.000967693 used all_to_all method: pairwise handle bytes (avg, min, max): 241216 152740 313476 buffer bytes (avg, min, max): 160523 101632 208640 gs_setup: 1187072 unique labels shared pairwise times (avg, min, max): 0.000483757 0.000410008 0.000513387 crystal router : 0.0109781 0.0108864 0.01108 used all_to_all method: pairwise handle bytes (avg, min, max): 1.18751e+06 656164 1481220 buffer bytes (avg, min, max): 791381 437248 987136 gs_setup: 240784 unique labels shared pairwise times (avg, min, max): 0.000113803 9.51052e-05 0.000123191 crystal router : 0.000868803 0.000849199 0.000890398 used all_to_all method: pairwise handle bytes (avg, min, max): 241216 152740 313476 buffer bytes (avg, min, max): 160523 101632 208640 gs_setup: 1187072 unique labels shared pairwise times (avg, min, max): 0.000463862 0.000412512 0.000496101 crystal router : 0.00953256 0.00941629 0.0096302 used all_to_all method: pairwise handle bytes (avg, min, max): 1.18751e+06 656164 1481220 buffer bytes (avg, min, max): 791381 437248 987136 gs_setup: 240784 unique labels shared pairwise times (avg, min, max): 0.000115333 9.5892e-05 0.000125098 crystal router : 0.00325584 0.0032357 0.0032764 used all_to_all method: pairwise handle bytes (avg, min, max): 241216 152740 313476 buffer bytes (avg, min, max): 160523 101632 208640 gs_setup: 1187072 unique labels shared pairwise times (avg, min, max): 0.00103487 0.00097959 0.00105419 crystal router : 0.00855757 0.00837069 0.0108672 used all_to_all method: pairwise handle bytes (avg, min, max): 1.18751e+06 656164 1481220 buffer bytes (avg, min, max): 791381 437248 987136 gs_setup: 240784 unique labels shared pairwise times (avg, min, max): 0.000114435 8.49009e-05 0.000136399 crystal router : 0.00204957 0.0020185 0.00207429 used all_to_all method: pairwise handle bytes (avg, min, max): 241216 152740 313476 buffer bytes (avg, min, max): 160523 101632 208640 gs_setup: 1187072 unique labels shared pairwise times (avg, min, max): 0.000471148 0.0004215 0.000498414 crystal router : 0.00977638 0.0085747 0.0109815 used all_to_all method: pairwise handle bytes (avg, min, max): 1.18751e+06 656164 1481220 buffer bytes (avg, min, max): 791381 437248 987136 gs_setup: 240784 unique labels shared pairwise times (avg, min, max): 0.000113707 9.59873e-05 0.000121403 crystal router : 0.000872626 0.000852704 0.00089221 used all_to_all method: pairwise handle bytes (avg, min, max): 241216 152740 313476 buffer bytes (avg, min, max): 160523 101632 208640 gs_setup: 1187072 unique labels shared pairwise times (avg, min, max): 0.000467315 0.000390005 0.000493503 crystal router : 0.010846 0.0107322 0.0109444 used all_to_all method: pairwise handle bytes (avg, min, max): 1.18751e+06 656164 1481220 buffer bytes (avg, min, max): 791381 437248 987136 ======================================================================= EquationType: NavierStokesCFE Session Name: Cylinder_NS_3D Spatial Dim.: 3 Max SEM Exp. Order: 4 Expansion Dim.: 3 Riemann Solver: ExactToro Advection Type: Projection Type: Discontinuous Galerkin Diffusion Type: Advection: explicit AdvectionType: WeakDG Diffusion: explicit Time Step: 1e-06 No. of Steps: 2000000 Checkpoints (steps): 50000 Integration Type: RungeKutta2_ImprovedEuler Problem Type: General ======================================================================= Initial Conditions: - Field rho: rhoInf - Field rhou: rhoInf * uInf - Field rhov: rhoInf * vInf - Field rhow: rhoInf * wInf - Field E: pInf / (Gamma - 1) + 0.5 * rhoInf * (uInf * uInf + vInf * vInf + wInf * wInf) Writing: "Cylinder_NS_3D_0.chk" slurmstepd: *** JOB 930941 CANCELLED AT 2016-07-06T03:27:36 DUE TO TIME LIMIT *** I run the solver by: "mpirun -np 120 CompressibleFlowSolver -v 3D_cylinder_mesh.xml 3D_cylinder_conditions.xml" Thanks for any hint. If you need more information, please let me know . Best regards Fabian Selbach University of Siegen

5 14

Problem after parallelization
by Amitvikram Dutta 12 Jul '16

12 Jul '16

Hi all, Recently I was made aware that there could be a possible problem with how nektar++ had been compiled on the cluster that I was trying to run it on. Apparently nektar++ had not been compiled with the MPI turned on with the result being that it was not parallelizing properly. After I pointed this out to the system administrator he went on to fix this issue. After fixing this however, whenever I try to submit a job I get the following error *mpirun has exited due to process rank 18 with PID 8273 onnode pod16a5.ibb exiting improperly. There are two reasons this could occur:1. this process did not call "init" before exiting, but others inthe job did. This can cause a job to hang indefinitely while it waitsfor all processes to call "init". By rule, if one process calls "init",then ALL processes must call "init" prior to termination.2. this process called "init", but exited without calling "finalize".By rule, all processes that call "init" MUST call "finalize" prior toexiting or it will be considered an "abnormal termination"This may have caused other processes in the application to beterminated by signals sent by mpirun (as reported here).* Note however, that when I try to run the same problem in serial mode there are no problems. I was hoping someone could shed light on any well-known issues regarding this. Sincerely, -- *Amitvikram Dutta* *MASc Candidate* *Graduate Research Assistant * *Okanagan CFD Laboratory* *University of British Columbia | Okanagan Campus*

2 1

VTK/Tecplot to FLD
by Lander, Marcus David 11 Jul '16

11 Jul '16

So, here is my situation: I am trying to introduce noise into my initial conditions for a Turbulent Channel DNS using awgn(). Currently, it is set up along the lines of u = 1.0 + 0.1*awgn(1.0). However, for what we are doing, we need this initial noise to be symmetric over the XY-plane at the middle of the channel (i.e. the bottom half is identical to the top half). I have been trying various combinations of modules in FieldConvert, but I feel as though I'm missing something. (If you could offer and insight on this, that would be great.) My other idea was to bring the initial field into a program such as paraview, and then make the necessary edits there. However, after reading and tinkering, I don't believe that FieldConvert has the ability to read in a VTK file and convert it back to a FLD file. Is there any way to actually read in data from a source such as VTK file or some other non-Nektar++ source? I tried this with a tecplot file, but it does not seem to want to read it.

2 1