Segmentation fault using align.seqs

mothur bugs
1

Hi!

I am currently analyzing a V4 Miseq 2x150 dataset using the latest MOTHUR release. I am using a cluster with 30 cores and should have about 16 TB memory space. At this stage I have 6416527 sequences of which 419706 are unique left and I tried to align them to the v4 segment of the silva reference, created by following the MiseqSOP.

However, after running this command:

align.seqs(fasta=ABFI_stab_ms2.contigs.good.unique.fasta, reference=silva.v4.fasta, processors=30)

I first got this:

Reading in the silva.v4.fasta template sequences...DONE.
It took 4 to read  14956 sequences.
Aligning sequences from ABFI_stab_ms2.trim.contigs.good.unique.fasta ...
*** glibc detected *** ./mothur: munmap_chunk(): invalid pointer: 0x0000000040382dc0 ***
*** glibc detected *** ./mothur: munmap_chunk(): invalid pointer: 0x0000000040382dc0 ***
(......)
Segmentation fault (core dumped)

And mothur crashed. The second time I tried I got:

Reading in the silva.v4.fasta template sequences... DONE.
It took 4 to read  14956 sequences.
Aligning sequences from ABFI_stab_ms2.trim.contigs.good.unique.fasta ...
Segmentation fault (core dumped)

Is this a problem that you have seen before? Hopefully you can point me in the right direction!

thanks, cheers, Aniek

2

MPI? Did you build the version of mothur or download an executable?

3

Hi!

Thanks for your reply. Sorry, I am new here. What do you mean by MPI? I compiled Mothur in linux from the mothur-1.35.1.tar.gz source code download.

hope this helps,
thanks
Aniek

4

Welcome to mothur! I was asking if you built the executable with the USEMPI=yes flag set in the makefile? It sounds like you did not. Have you tried running our prebuilt executable, https://github.com/mothur/mothur/releases? The Mothur.cen_64.zip is our linux version.

5

Hi!

Thanks for your reply. I have now downloaded your executable and used it. Same problem:

mothur > align.seqs(fasta=ABFI_stab_ms2.trim.contigs.good.unique.fasta, reference=silva.v4.fasta, processors=30)

Using 30 processors.

Reading in the silva.v4.fasta template sequences... DONE.
It took 8 to read  14956 sequences.
Aligning sequences from ABFI_stab_ms2.trim.contigs.good.unique.fasta ...
Segmentation fault (core dumped)

regards
Aniek

6

Hello Aniek,
I have exactly the same problem with 1.35. Did you solve it finally?

Alexandra

7

Hi Alexandra

no, I have not been able to solve it. The problem stayed when I ran it on the computer cluster, but on my mac it was fine (less cores). I ran it there instead and proceeded on the cluster,

good luck
Aniek

8

The problem is likely that you were using 30 cores and so the reference data gets copied 30 times sucking up tons of RAM and crashing out your computer.

9

Ok that explains. Thanks Pat!