Closed donnaaboise closed 1 year ago
What kind of problem is this? Any error messages/ways to reproduce?
Hello, this issue seems to persist. I am using OpenMPI version 4 and I get an error message. With openmpi-4* loaded it should be possible to reproduce the problem running following command in the forestclaw/applications/clawpack/advection/2d/swirl directory
$FCLAW_BIN/swirl --user:claw-version=4 --clawpack46:vtk-out=T
Any ideas?
Thanks in advance!
Thanks for following up on this.
I generally use MPICH when it is available, in part to avoid problems with openMPI. That said, it would be good to know what problem you are running into.
Can you post the following ?
The error message are you getting exactly?
the first few lines in the command
$ git log
Your version of openMPI
The version that runs on my machine is actually v3.1.3 from October 2018. Sorry for the confusion.
I was able to more or less reproduce your error using openmpi3/gcc/3.1.3
. Other versions of OpenMPI also gave strange errors.
(r2) ~/.../advection/2d/swirl (develop) % mpirun -n 1 swirl --user:claw-version=4 --clawpack46:vtk-out=T
[r2:12364] mca_base_component_repository_open: unable to open mca_pml_ucx: libibcm.so.1: cannot open shared object file: No such file or directory (ignored)
[r2:12364] mca_base_component_repository_open: unable to open mca_osc_ucx: libibcm.so.1: cannot open shared object file: No such file or directory (ignored)
[libsc] This is libsc 2.1.47-eaa88
[p4est] This is p4est 2.0.94-00da
[fclaw] This is ForestClaw 0.1.4928-8180
[fclaw] CPP mpicc -E
[fclaw] CPPFLAGS
[fclaw] F77 gfortran
[fclaw] FFLAGS -O2 -cpp
[fclaw] CC mpicc
[fclaw] CFLAGS -O2 -Wall -std=c99 -pedantic
[fclaw] CXX mpicxx
[fclaw] CXXFLAGS -O2 -Wall
[fclaw] LDFLAGS
[fclaw] FLIBS -L/cm/local/apps/cuda/libs/current/lib64/../lib64 -L/cm/shared/apps/slurm/17.11.12/lib64/../lib64 -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5 -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../lib64 -L/lib/../lib64 -L/usr/lib/../lib64 -L/cm/local/apps/cuda/libs/current/lib64 -L/cm/shared/apps/cuda10.0/toolkit/10.0.130/targets/x86_64-linux/lib -L/cm/shared/apps/slurm/17.11.12/lib64/slurm -L/cm/shared/apps/slurm/17.11.12/lib64 -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../.. -lgfortran -lm -lquadmath
[fclaw] LIBS -lz -lm
[fclaw] Options:
[fclaw] help false
[fclaw] version false
[fclaw] print-options false
[fclaw] verbosity essential
[fclaw] lib-verbosity essential
[fclaw] initial_dt 0.005
[fclaw] max_cfl 1
[fclaw] desired_cfl 0.9
[fclaw] reduce-cfl true
[fclaw] use_fixed_dt false
[fclaw] outstyle 1
[fclaw] tfinal 4
[fclaw] nout 16
[fclaw] nstep 1
[fclaw] advance-one-step true
[fclaw] outstyle-uses-maxlevel true
[fclaw] subcycle true
[fclaw] weighted_partition false
[fclaw] time-sync false
[fclaw] flux-correction true
[fclaw] fluctuation-correction true
[fclaw] output true
[fclaw] output-gauges false
[fclaw] gauge-buffer-length 1
[fclaw] tikz-out false
[fclaw] tikz-figsize 4 4
[fclaw] tikz-plot-prefix plot
[fclaw] tikz-plot-suffix png
[fclaw] tikz-mesh-only false
[fclaw] tikz-plot-fig true
[fclaw] prefix fort
[fclaw] vtkspace 0
[fclaw] init_ghostcell false
[fclaw] minlevel 2
[fclaw] maxlevel 6
[fclaw] regrid_interval 1
[fclaw] refratio 2
[fclaw] smooth-refine true
[fclaw] smooth-level 0
[fclaw] coarsen-delay 0
[fclaw] refine_threshold 0.25
[fclaw] coarsen_threshold 0.05
[fclaw] run-user-diagnostics false
[fclaw] compute-error false
[fclaw] conservation-check false
[fclaw] report-timing true
[fclaw] report-timing-verbosity summary
[fclaw] ghost_patch_pack_area true
[fclaw] ghost_patch_pack_extra false
[fclaw] ghost_patch_pack_numextrafields 0
[fclaw] trapfpe true
[fclaw] mpi_debug false
[fclaw] ax 0
[fclaw] bx 1
[fclaw] ay 0
[fclaw] by 1
[fclaw] manifold false
[fclaw] mi 1
[fclaw] mj 1
[fclaw] periodic_x false
[fclaw] periodic_y false
[fclaw] scale 1 1 1
[fclaw] shift 0 0 0
[fclaw] phi 0
[fclaw] theta 0
[fclaw] clawpatch:mx 8
[fclaw] clawpatch:my 8
[fclaw] clawpatch:maux 3
[fclaw] clawpatch:mbc 2
[fclaw] clawpatch:meqn 1
[fclaw] clawpatch:interp_stencil_width 3
[fclaw] clawpatch:ghost_patch_pack_aux true
[fclaw] clawpack46:order 2 2
[fclaw] clawpack46:mcapa 0
[fclaw] clawpack46:src_term false
[fclaw] clawpack46:use-fwaves false
[fclaw] clawpack46:mwaves 1
[fclaw] clawpack46:mthlim 3
[fclaw] clawpack46:mthbc 1 1 1 1
[fclaw] clawpack46:ascii-out true
[fclaw] clawpack46:vtk-out true
[fclaw] clawpack5:order 2 2
[fclaw] clawpack5:mcapa 0
[fclaw] clawpack5:src_term false
[fclaw] clawpack5:use_fwaves false
[fclaw] clawpack5:mwaves 1
[fclaw] clawpack5:mthlim 3
[fclaw] clawpack5:mthbc 1 1 1 1
[fclaw] clawpack5:ascii-out true
[fclaw] clawpack5:vtk-out false
[fclaw] user:period 4
[fclaw] user:claw-version 4
[fclaw] Arguments: none
[p4est 0] Local minimum/maximum levels: 2 2
[p4est] Global minimum/maximum levels: 2 2
[p4est 0] Patches on level 2: 16
[fclaw] Max threads set to 0
[fclaw] Output Frame 0 at time 0.00000000e+00
Improper use of FILE Mode, Using WRONLY for Read!
[libsc 0] Abort: MPI error
[libsc 0] Abort: ../forestclaw/src/patches/clawpatch/fclaw2d_clawpatch_output_vtk.c:424
[libsc 0] Abort: Obtained 10 stack frames
[libsc 0] Stack 0: swirl() [0x46497c]
[libsc 0] Stack 1: swirl() [0x463bfa]
[libsc 0] Stack 2: swirl() [0x4640dc]
[libsc 0] Stack 3: swirl() [0x41833f]
[libsc 0] Stack 4: swirl() [0x41843c]
[libsc 0] Stack 5: swirl() [0x4354fd]
[libsc 0] Stack 6: swirl() [0x42e5d4]
[libsc 0] Stack 7: swirl() [0x403b85]
[libsc 0] Stack 8: libc.so.6(__libc_start_main+0xf5) [0x7ffff55b4505]
[libsc 0] Stack 9: swirl() [0x403bd2]
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
with errorcode 1.
NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
As a possible work-around, do you have access to MPICH? I just recompiled and ran using MPICH mpich/ge/gcc/64/3.2.1
and everything worked fine. That could of course just mean that MPICH is more forgiving of subtle errors, so it would still be good to figure out what the problem with OpenMPI is.
Hello,
I was able to more or less reproduce your error using
openmpi3/gcc/3.1.3
. Other versions of OpenMPI also gave strange errors.
my personal experience over the years has lead me to discourage use of OpenMPI altogether. Ymmv. Does make check in the p4est installation work ok? It tests the MPI I/O load/save mechanism we're using.
(r2) ~/.../advection/2d/swirl (develop) % mpirun -n 1 swirl --user:claw-version=4 --clawpack46:vtk-out=T [r2:12364] mca_base_component_repository_open: unable to open mca_pml_ucx: libibcm.so.1: cannot open shared object file: No such file or directory (ignored) [r2:12364] mca_base_component_repository_open: unable to open mca_osc_ucx: libibcm.so.1: cannot open shared object file: No such file or directory (ignored) [libsc] This is libsc 2.1.47-eaa88 [p4est] This is p4est 2.0.94-00da [fclaw] This is ForestClaw 0.1.4928-8180 [fclaw] CPP mpicc -E [fclaw] CPPFLAGS [fclaw] F77 gfortran [fclaw] FFLAGS -O2 -cpp [fclaw] CC mpicc [fclaw] CFLAGS -O2 -Wall -std=c99 -pedantic [fclaw] CXX mpicxx [fclaw] CXXFLAGS -O2 -Wall [fclaw] LDFLAGS [fclaw] FLIBS -L/cm/local/apps/cuda/libs/current/lib64/../lib64 -L/cm/shared/apps/slurm/17.11.12/lib64/../lib64 -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5 -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../../../lib64 -L/lib/../lib64 -L/usr/lib/../lib64 -L/cm/local/apps/cuda/libs/current/lib64 -L/cm/shared/apps/cuda10.0/toolkit/10.0.130/targets/x86_64-linux/lib -L/cm/shared/apps/slurm/17.11.12/lib64/slurm -L/cm/shared/apps/slurm/17.11.12/lib64 -L/usr/lib/gcc/x86_64-redhat-linux/4.8.5/../../.. -lgfortran -lm -lquadmath [fclaw] LIBS -lz -lm [fclaw] Options: [fclaw] help false [fclaw] version false [fclaw] print-options false [fclaw] verbosity essential [fclaw] lib-verbosity essential [fclaw] initial_dt 0.005 [fclaw] max_cfl 1 [fclaw] desired_cfl 0.9 [fclaw] reduce-cfl true [fclaw] use_fixed_dt false [fclaw] outstyle 1 [fclaw] tfinal 4 [fclaw] nout 16 [fclaw] nstep 1 [fclaw] advance-one-step true [fclaw] outstyle-uses-maxlevel true [fclaw] subcycle true [fclaw] weighted_partition false [fclaw] time-sync false [fclaw] flux-correction true [fclaw] fluctuation-correction true [fclaw] output true [fclaw] output-gauges false [fclaw] gauge-buffer-length 1 [fclaw] tikz-out false [fclaw] tikz-figsize 4 4 [fclaw] tikz-plot-prefix plot [fclaw] tikz-plot-suffix png [fclaw] tikz-mesh-only false [fclaw] tikz-plot-fig true [fclaw] prefix fort [fclaw] vtkspace 0 [fclaw] init_ghostcell false [fclaw] minlevel 2 [fclaw] maxlevel 6 [fclaw] regrid_interval 1 [fclaw] refratio 2 [fclaw] smooth-refine true [fclaw] smooth-level 0 [fclaw] coarsen-delay 0 [fclaw] refine_threshold 0.25 [fclaw] coarsen_threshold 0.05 [fclaw] run-user-diagnostics false [fclaw] compute-error false [fclaw] conservation-check false [fclaw] report-timing true [fclaw] report-timing-verbosity summary [fclaw] ghost_patch_pack_area true [fclaw] ghost_patch_pack_extra false [fclaw] ghost_patch_pack_numextrafields 0 [fclaw] trapfpe true [fclaw] mpi_debug false [fclaw] ax 0 [fclaw] bx 1 [fclaw] ay 0 [fclaw] by 1 [fclaw] manifold false [fclaw] mi 1 [fclaw] mj 1 [fclaw] periodic_x false [fclaw] periodic_y false [fclaw] scale 1 1 1 [fclaw] shift 0 0 0 [fclaw] phi 0 [fclaw] theta 0 [fclaw] clawpatch:mx 8 [fclaw] clawpatch:my 8 [fclaw] clawpatch:maux 3 [fclaw] clawpatch:mbc 2 [fclaw] clawpatch:meqn 1 [fclaw] clawpatch:interp_stencil_width 3 [fclaw] clawpatch:ghost_patch_pack_aux true [fclaw] clawpack46:order 2 2 [fclaw] clawpack46:mcapa 0 [fclaw] clawpack46:src_term false [fclaw] clawpack46:use-fwaves false [fclaw] clawpack46:mwaves 1 [fclaw] clawpack46:mthlim 3 [fclaw] clawpack46:mthbc 1 1 1 1 [fclaw] clawpack46:ascii-out true [fclaw] clawpack46:vtk-out true [fclaw] clawpack5:order 2 2 [fclaw] clawpack5:mcapa 0 [fclaw] clawpack5:src_term false [fclaw] clawpack5:use_fwaves false [fclaw] clawpack5:mwaves 1 [fclaw] clawpack5:mthlim 3 [fclaw] clawpack5:mthbc 1 1 1 1 [fclaw] clawpack5:ascii-out true [fclaw] clawpack5:vtk-out false [fclaw] user:period 4 [fclaw] user:claw-version 4 [fclaw] Arguments: none [p4est 0] Local minimum/maximum levels: 2 2 [p4est] Global minimum/maximum levels: 2 2 [p4est 0] Patches on level 2: 16 [fclaw] Max threads set to 0 [fclaw] Output Frame 0 at time 0.00000000e+00 Improper use of FILE Mode, Using WRONLY for Read! [libsc 0] Abort: MPI error [libsc 0] Abort: ../forestclaw/src/patches/clawpatch/fclaw2d_clawpatch_output_vtk.c:424 [libsc 0] Abort: Obtained 10 stack frames [libsc 0] Stack 0: swirl() [0x46497c] [libsc 0] Stack 1: swirl() [0x463bfa] [libsc 0] Stack 2: swirl() [0x4640dc] [libsc 0] Stack 3: swirl() [0x41833f] [libsc 0] Stack 4: swirl() [0x41843c] [libsc 0] Stack 5: swirl() [0x4354fd] [libsc 0] Stack 6: swirl() [0x42e5d4] [libsc 0] Stack 7: swirl() [0x403b85] [libsc 0] Stack 8: libc.so.6(__libc_start_main+0xf5) [0x7ffff55b4505] [libsc 0] Stack 9: swirl() [0x403bd2] -------------------------------------------------------------------------- MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD with errorcode 1. NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes. You may or may not see output from other processes, depending on exactly when Open MPI kills them. --------------------------------------------------------------------------
As a possible work-around, do you have access to MPICH? I just recompiled and ran using MPICH
mpich/ge/gcc/64/3.2.1
and everything worked fine. That could of course just mean that MPICH is more forgiving of subtle errors, so it would still be good to figure out what the problem with OpenMPI is.-- You are receiving this because you were assigned. Reply to this email directly or view it on GitHub: https://github.com/ForestClaw/forestclaw/issues/144#issuecomment-680049645
Thanks for looking into this. I compiled with mpich and it works just fine. Make check with OpenMPI does not indicate any problems though.
I doubt there is a problem with p4est. Rather the problem is more likely with how forestclass is handling the VTK. We should keep this on the list of things to look at at some point, but may be not the highest priority.
@hommedesbois is MPICH a good work-around for you?
@donnaaboise yes, definitely.
I doubt there is a problem with p4est. Rather the problem is more likely with how forestclass is handling the VTK. We should keep this on the list of things to look at at some point, but may be not the highest priority.
The forestclaw VTK output and p4est_save use very similar MPI I/O code. I'd suspect one to be an indicator for the other.
@hommedesbois is MPICH a good work-around for you?
For me, it usually is.
Solution seems to be to avoid using OpenMPI.
It seems that there is a problem with the VTK output and the latest version of OpenMPI. Version 2.1 works, but version 3.1 (from Nov. 2018) does not.