search

open-mpi / ompi

Open MPI main development repository
https://www.open-mpi.org
Other
2.16k stars 859 forks source link

btl_openib_component.c:3556:handle_wc #4529

Closed lcebaman closed 3 years ago

lcebaman commented 6 years ago

Runnning OpenMPI 3.0.0 and RMA (MPI_THREAD_MULTIPLE) I get 

[[27217,0],12][btl_openib_component.c:3556:handle_wc] from nvb27 to: nvb27 error polling L
P CQ with status REMOTE ACCESS ERROR status number 10 for wr_id 19059c8 opcode 4  vendor e
rror 136 qp_idx 3

I've noticed that this happens when the number of MPI processes per node >=4 . Here is some more info that could be (or not) related to this issue:

-------------------------------------------------------------------------
A system call failed during shared memory initialization that should
not have.  It is likely that your MPI job will now either abort or
experience performance degradation.

  Local host:  nvb26
  System call: unlink(2) /scratch/tmp/1076409/ompi.nvb26.3009/jf.27833/0/window_11.nvb26
  Error:       No such file or directory (errno 2)
--------------------------------------------------------------------------
slurmstepd: *** STEP 1076409.0 ON nvb10 CANCELLED AT 2017-11-23T21:03:19 ***
srun: Job step aborted: Waiting up to 32 seconds for job step to finish.
[[27833,0],2][btl_openib_component.c:3556:handle_wc] from nvb10 to: nvb10 error
polling LP CQ with status REMOTE ACCESS ERROR status number 10 for wr_id 13efd68
 opcode 0  vendor error 136 qp_idx 3
jsquyres commented 6 years ago

@lcebaman Can you provide a simple reproducer?

@hjelmn Is this a race condition in some one-sided code?

lcebaman commented 6 years ago

@jsquyres I don't have simple reproducer, but I can try to produce one. It seems this could well be a race condition.

jsquyres commented 6 years ago

@lcebaman Nathan (@hjelmn) just recently committed several one-sided fixes to master. Could you try a master nightly snapshot tarball, and see if the issue is resolved for you?

https://www.open-mpi.org/nightly/master/

hjelmn commented 6 years ago

I vaguely remember fixing this. I don't know if I upstreamed the fix though.... Will take a look when I get back to the office on Friday.

jsquyres commented 6 years ago

@lcebaman Any news?

@hjelmn Did you get to look and see if you put the fix in release branches?

hjelmn commented 6 years ago

We just got a relevant system installed. Will try to look at it today.

hppritcha commented 6 years ago

@lcebaman could you try testing against one of 3.0.1 release candidates? There were fixes for multi-threaded RMA in this release.

ghost commented 5 years ago

It seems issue also can be reproduced with Intel/OPA hardware. openmpi3 segment fault. Latest openmpi4 does not segment fault, but hang.

[root@rdma05 ~]# /usr/lib64/openmpi3/bin/mpirun --allow-run-as-root -mca pml ob1 -mca btl openib,self -mca btl_openib_if_include hfi1_0:1 --bind-to none -np 2 -H rdma05,rdma06  /usr/lib64/openmpi3/bin/mpitests-IMB-EXT
#------------------------------------------------------------
#    Intel (R) MPI Benchmarks 2018 Update 1, MPI-2 part    
#------------------------------------------------------------
# Date                  : Tue Jun  4 21:23:26 2019
# Machine               : x86_64
# System                : Linux
# Release               : 3.10.0-1053.el7.x86_64
# Version               : #1 SMP Sat Jun 1 09:11:33 EDT 2019
# MPI Version           : 3.1
# MPI Thread Environment: 

# Calling sequence was: 

# /usr/lib64/openmpi3/bin/mpitests-IMB-EXT

# Minimum message length in bytes:   0
# Maximum message length in bytes:   4194304
#
# MPI_Datatype                   :   MPI_BYTE 
# MPI_Datatype for reductions    :   MPI_FLOAT
# MPI_Op                         :   MPI_SUM  
#
#

# List of Benchmarks to run:

# Window
# Unidir_Get
# Unidir_Put
# Bidir_Get
# Bidir_Put
# Accumulate

#----------------------------------------------------------------
# Benchmarking Window 
# #processes = 2 
#----------------------------------------------------------------
       #bytes #repetitions  t_min[usec]  t_max[usec]  t_avg[usec]
            0          100       309.22       309.25       309.24
            4          100       312.05       312.06       312.05
            8          100       320.37       320.38       320.37
           16          100       310.31       310.33       310.32
           32          100       313.46       313.48       313.47
           64          100       311.57       311.58       311.57
          128          100       313.01       313.03       313.02
          256          100       317.72       317.73       317.72
          512          100       318.02       318.03       318.02
         1024          100       323.31       323.33       323.32
         2048          100       316.87       316.88       316.87
         4096          100       318.84       318.85       318.84
         8192          100       316.71       316.72       316.72
        16384          100       319.34       319.35       319.34
        32768          100       354.19       354.26       354.23
        65536          100       319.15       319.18       319.16
       131072          100       316.22       316.23       316.22
       262144          100       320.43       320.45       320.44
       524288           80       319.72       319.74       319.73
      1048576           40       320.70       320.74       320.72
      2097152           20       319.10       319.19       319.15
      4194304           10       329.20       329.72       329.46

#---------------------------------------------------
# Benchmarking Unidir_Get 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0         1000         0.04         0.00
            4         1000         3.77         1.06
            8         1000         3.70         2.16
           16         1000         3.72         4.31
           32         1000         3.87         8.26
           64         1000         3.67        17.43
          128         1000         3.72        34.40
          256         1000         3.79        67.55
          512         1000         4.04       126.66
         1024         1000         4.13       247.97
         2048         1000         4.19       488.82
         4096         1000         4.30       951.81
         8192         1000         4.48      1828.61
        16384         1000         5.15      3181.56
        32768         1000         6.77      4838.10
        65536          640        10.79      6072.77
       131072          320        18.79      6977.02
       262144          160        35.18      7451.71
       524288           80        68.51      7653.20
      1048576           40       139.07      7540.05
      2097152           20       277.37      7560.95
      4194304           10       551.66      7603.10

#---------------------------------------------------
# Benchmarking Unidir_Get 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: NON-AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0          100         6.01         0.00
            4          100        13.92         0.29
            8          100        13.75         0.58
           16          100        13.86         1.15
           32          100        14.13         2.26
           64          100        13.97         4.58
          128          100        13.75         9.31
          256          100        14.07        18.20
          512          100        14.47        35.38
         1024          100        14.81        69.16
         2048          100        14.86       137.79
         4096          100        15.73       260.33
         8192          100        20.17       406.23
        16384          100        21.90       748.18
        32768          100        21.68      1511.18
        65536          100        30.58      2142.92
       131072          100        38.67      3389.73
       262144          100        54.69      4793.52
       524288           80        89.49      5858.36
      1048576           40       158.37      6621.18
      2097152           20       354.18      5921.20
      4194304           10       607.61      6902.92

#---------------------------------------------------
# Benchmarking Unidir_Put 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0         1000         0.03         0.00
            4         1000         3.26         1.23
            8         1000         3.21         2.49
           16         1000         3.18         5.03
           32         1000         3.33         9.62
           64         1000         3.32        19.26
          128         1000         3.50        36.59
          256         1000         3.75        68.33
          512         1000         2.37       215.71
         1024         1000         2.42       422.86
         2048         1000         2.39       856.31
         4096         1000         2.40      1708.07
         8192         1000         2.71      3018.75
        16384         1000         4.02      4074.74
        32768         1000         6.30      5201.24
        65536          640        10.71      6117.21
       131072          320        20.96      6254.20
       262144          160        41.63      6297.14
       524288           80        82.95      6320.63
      1048576           40       161.72      6483.93
      2097152           20       315.88      6639.08
      4194304           10       661.03      6345.08

#---------------------------------------------------
# Benchmarking Unidir_Put 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: NON-AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0          100         5.92         0.00
            4          100        10.98         0.36
            8          100        10.84         0.74
           16          100        10.93         1.46
           32          100        10.94         2.92
           64          100        11.11         5.76
          128          100        11.05        11.58
          256          100        11.14        22.97
          512          100        13.81        37.07
         1024          100        13.76        74.41
         2048          100        14.16       144.63
         4096          100        14.62       280.09
         8192          100        19.16       427.62
        16384          100        20.22       810.14
        32768          100        21.92      1494.61
        65536          100        27.74      2362.59
       131072          100        37.86      3461.88
       262144          100        58.78      4460.09
       524288           80       100.46      5218.69
      1048576           40       177.70      5900.81
      2097152           20       362.20      5790.01
      4194304           10       683.71      6134.61

#---------------------------------------------------
# Benchmarking Bidir_Get 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0         1000         0.03         0.00
            4         1000         6.98         0.57
            8         1000         6.89         1.16
           16         1000         6.84         2.34
           32         1000         7.03         4.55
           64         1000         7.02         9.12
          128         1000         7.02        18.22
          256         1000         6.97        36.74
          512         1000         7.77        65.86
         1024         1000         7.70       132.92
         2048         1000         8.04       254.83
         4096         1000         8.51       481.14
         8192         1000         8.49       964.62
        16384         1000        10.03      1633.00
        32768         1000        15.24      2149.52
        65536          640        24.01      2729.18
       131072          320        40.28      3253.94
       262144          160        74.42      3522.60
       524288           80       146.86      3570.05
      1048576           40       289.99      3615.92
      2097152           20       585.49      3581.86
      4194304           10      1153.52      3636.10

#---------------------------------------------------
# Benchmarking Bidir_Get 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: NON-AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0          100         6.00         0.00
            4          100        25.99         0.15
            8          100        26.02         0.31
           16          100        26.00         0.62
           32          100        25.50         1.25
           64          100        25.59         2.50
          128          100        26.85         4.77
          256          100        26.57         9.63
          512          100        26.71        19.17
         1024          100        26.62        38.47
         2048          100        27.08        75.64
         4096          100        28.79       142.29
         8192          100        36.64       223.56
        16384          100        39.80       411.62
        32768          100        38.35       854.36
        65536          100        49.20      1332.06
       131072          100        70.34      1863.29
       262144          100       109.35      2397.37
       524288           80       191.23      2741.65
      1048576           40       329.95      3177.95
      2097152           20       669.58      3132.05
      4194304           10      1238.61      3386.31

#---------------------------------------------------
# Benchmarking Bidir_Put 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0         1000         0.03         0.00
            4         1000         3.77         1.06
            8         1000         3.50         2.28
           16         1000         3.53         4.53
           32         1000         3.84         8.33
           64         1000         3.75        17.05
          128         1000         3.76        34.08
          256         1000         4.00        64.06
          512         1000         2.86       178.94
         1024         1000         2.89       353.76
         2048         1000         2.96       691.32
         4096         1000         2.87      1426.03
         8192         1000         3.77      2173.03
        16384         1000         5.39      3038.07
        32768         1000         9.34      3506.89
        65536          640        17.72      3698.51
       131072          320        35.71      3670.56
       262144          160        70.18      3735.28
       524288           80       142.26      3685.46
      1048576           40       272.11      3853.45
      2097152           20       554.40      3782.75
      4194304           10      1099.62      3814.33

#---------------------------------------------------
# Benchmarking Bidir_Put 
# #processes = 2 
#---------------------------------------------------
#
#    MODE: NON-AGGREGATE 
#
       #bytes #repetitions      t[usec]   Mbytes/sec
            0          100         6.00         0.00
            4          100        11.55         0.35
            8          100        11.50         0.70
           16          100        29.16         0.55
           32          100        11.49         2.79
           64          100        11.76         5.44
          128          100        11.55        11.08
          256          100        11.85        21.60
          512          100        15.14        33.82
         1024          100        15.34        66.77
         2048          100        15.72       130.25
         4096          100        16.31       251.14
         8192          100        20.72       395.40
        16384          100        23.89       685.86
        32768          100        28.71      1141.22
        65536          100        37.67      1739.61
       131072          100        53.10      2468.24
       262144          100        90.86      2885.01
       524288           80       168.10      3118.85
      1048576           40       319.97      3277.07
      2097152           20       601.83      3484.64
      4194304           10      1121.62      3739.52
[rdma05:20630:0:20630] Caught signal 11 (Segmentation fault: address not mapped to object at address (nil))
==== backtrace ====
    0  /lib64/libucs.so.0(+0x1a82c) [0x7fc53a78e82c]
    1  /lib64/libucs.so.0(+0x1a9e2) [0x7fc53a78e9e2]
    2  /usr/lib64/openmpi3/lib/openmpi/mca_btl_openib.so(mca_btl_openib_get+0x131) [0x7fc53d549bf1]
    3  /usr/lib64/openmpi3/lib/openmpi/mca_osc_rdma.so(ompi_osc_get_data_blocking+0x1bc) [0x7fc53b44ec8c]
    4  /usr/lib64/openmpi3/lib/openmpi/mca_osc_rdma.so(+0x11783) [0x7fc53b458783]
    5  /usr/lib64/openmpi3/lib/openmpi/mca_osc_rdma.so(ompi_osc_rdma_accumulate+0x187) [0x7fc53b45d607]
    6  /usr/lib64/openmpi3/lib/libmpi.so.40(PMPI_Accumulate+0x2d3) [0x7fc550de0c53]
    7  /usr/lib64/openmpi3/bin/mpitests-IMB-EXT() [0x4088bd]
    8  /usr/lib64/openmpi3/bin/mpitests-IMB-EXT() [0x4063c8]
    9  /usr/lib64/openmpi3/bin/mpitests-IMB-EXT() [0x402016]
   10  /lib64/libc.so.6(__libc_start_main+0xf5) [0x7fc550786545]
   11  /usr/lib64/openmpi3/bin/mpitests-IMB-EXT() [0x402286]
===================
[rdma05:20630] *** Process received signal ***
[rdma05:20630] Signal: Segmentation fault (11)
[rdma05:20630] Signal code:  (-6)
[rdma05:20630] Failing at address: 0x5096
[rdma05:20630] [ 0] /lib64/libpthread.so.0(+0xf630)[0x7fc550b41630]
[rdma05:20630] [ 1] /usr/lib64/openmpi3/lib/openmpi/mca_btl_openib.so(mca_btl_openib_get+0x131)[0x7fc53d549bf1]
[rdma05:20630] [ 2] /usr/lib64/openmpi3/lib/openmpi/mca_osc_rdma.so(ompi_osc_get_data_blocking+0x1bc)[0x7fc53b44ec8c]
[rdma05:20630] [ 3] /usr/lib64/openmpi3/lib/openmpi/mca_osc_rdma.so(+0x11783)[0x7fc53b458783]
[rdma05:20630] [ 4] /usr/lib64/openmpi3/lib/openmpi/mca_osc_rdma.so(ompi_osc_rdma_accumulate+0x187)[0x7fc53b45d607]
[rdma05:20630] [ 5] /usr/lib64/openmpi3/lib/libmpi.so.40(PMPI_Accumulate+0x2d3)[0x7fc550de0c53]
[rdma05:20630] [ 6] /usr/lib64/openmpi3/bin/mpitests-IMB-EXT[0x4088bd]
[rdma05:20630] [ 7] /usr/lib64/openmpi3/bin/mpitests-IMB-EXT[0x4063c8]
[rdma05:20630] [ 8] /usr/lib64/openmpi3/bin/mpitests-IMB-EXT[0x402016]
[rdma05:20630] [ 9] /lib64/libc.so.6(__libc_start_main+0xf5)[0x7fc550786545]
[rdma05:20630] [10] /usr/lib64/openmpi3/bin/mpitests-IMB-EXT[0x402286]
[rdma05:20630] *** End of error message ***
[[53556,1],1][btl_openib_component.c:3644:handle_wc] from rdma06 to: rdma05 error polling LP CQ with status REMOTE ACCESS ERROR status number 10 for wr_id 1ab6e58 opcode 3  vendor error 0 qp_idx 0
[root@rdma05 ~]#
awlauria commented 3 years ago

Openib btl is removed. Closing.