Bellerophon MPI bug: Seg fault

Hi All,

I am having a problem with implementing Bellerophon on a Linux Ubuntu Workstation using OpenMPI version 1.5. I get a seq fault after ~ 4 hours of running… Here is the beginning and end of the logfile… Any tips would be really appreciated!
I am using mpirun -np 15 ./mothur "#chimera.bellerophon(fasta=bacall.align, processors=15)


Linux version Using ReadLine Running 64Bit Version mothur v.1.17.3 Last updated: 3/07/2011 ... Type 'quit()' to exit program Using MPI version 2.1 Script Mode
mothur > chimera.bellerophon(fasta=bacall.align, processors=15) Checking sequences from bacall.align ... Reading sequences from bacall.align...Done.

Processing sliding window: 1850
Processing sliding window: 530
Processing sliding window: 1190
Processing sliding window: 1720

Processing sliding window: 264
[macmanes:25440] *** Process received signal ***
[macmanes:25440] Signal: Segmentation fault (11)
[macmanes:25440] Signal code: (128)
[macmanes:25440] Failing at address: (nil)
[macmanes:25440] [ 0] /lib/libpthread.so.0(+0xfb40) [0x7f7f7c11ab40]
[macmanes:25440] [ 1] /lib/libc.so.6(+0x1275cf) [0x7f7f7beaf5cf]
[macmanes:25440] [ 2] /usr/lib/libstdc++.so.6(_ZNSsC2EPKcRKSaIcE+0x30) [0x7f7f7c85fe20]
[macmanes:25440] [ 3] /home/matthew/Mothur.1.17.3/Mothur.source/mothur(_ZN11Bellerophon11getChimerasEv+0x361) [0x4aaab1]
[macmanes:25440] [ 4] /home/matthew/Mothur.1.17.3/Mothur.source/mothur(_ZN25ChimeraBellerophonCommand7executeEv+0xa6b) [0x4e9efb]
[macmanes:25440] [ 5] /home/matthew/Mothur.1.17.3/Mothur.source/mothur(_ZN12ScriptEngine8getInputEv+0x5f8) [0x6587f8]
[macmanes:25440] [ 6] /home/matthew/Mothur.1.17.3/Mothur.source/mothur(main+0x101c) [0x7c424c]
[macmanes:25440] [ 7] /lib/libc.so.6(__libc_start_main+0xfe) [0x7f7f7bda6d8e]
[macmanes:25440] [ 8] /home/matthew/Mothur.1.17.3/Mothur.source/mothur() [0x4678a9]
[macmanes:25440] *** End of error message ***

mpirun noticed that process rank 0 with PID 25440 on node macmanes exited on signal 11 (Segmentation fault).

Hi,

I am having a very similar problem/error. I compiled Mothur 1.19.0 for a 32-bit cluster running Centos 5.4 and to run it using mpi. Once I run the following command:

/usr/local/bin/mpirun -np 8 ~/Mothur.source/mothurmpi32 < ~/step2-trim.batch >> debug.log

with step2-trim.batch being a batch file containing the following command,

trim.seqs(fasta=F0NCP3D01.fasta, oligos=primers_F0NCP3D01.fas, qfile=F0NCP3D01.qual, maxambig=0, maxhomop=8, flip=T, bdiffs=1, pdiffs=2, qwindowaverage=35, qwindowsize=50, processors=8)

I get the following error:

[a31:04672] *** Process received signal ***
[a31:04672] Signal: Segmentation fault (11)
[a31:04672] Signal code: Address not mapped (1)
[a31:04672] Failing at address: (nil)
[a31:04672] [ 0] [0x4001d440]
[a31:04672] [ 1] /home/vbucci/Mothur.source/mothurmpi32(_ZN14InteractEngine8getInputEv+0xb5) [0x831c7f5]
[a31:04672] [ 2] /home/vbucci/Mothur.source/mothurmpi32(main+0xd10) [0x850a880]
[a31:04672] [ 3] /lib/libc.so.6(__libc_start_main+0xdc) [0x40343e9c]
[a31:04672] [ 4] /home/vbucci/Mothur.source/mothurmpi32 [0x80a18a1]
[a31:04672] *** End of error message ***

mpirun noticed that process rank 0 with PID 4672 on node a31.cluster.private exited on signal 11 (Segmentation fault).

Any suggestion about what I may be doing wrong?

I suspect it may be a problem with your input files. Only some of mothur’s commands take advantage of multiple processors. Trim.seqs does, but not with mpi, so only one process would be running the trim.seqs code. Can you download the mothur executable for linux and run trim.seqs(fasta=F0NCP3D01.fasta, oligos=primers_F0NCP3D01.fas, qfile=F0NCP3D01.qual, maxambig=0, maxhomop=8, flip=T, bdiffs=1, pdiffs=2, qwindowaverage=35, qwindowsize=50, processors=8) to see if you still get an error?

Hi macmanes,
Thanks for bringing the bug with chimera.bellerophon to our attention. It will be fixed in the 1.23.0 release.
-Sarah