Metavelvet segmentation fault

Question

[makisthegreat]

Hello all,

I've been using the latest version of Velvet (1.2.10) and the latest version of Metavelvet (1.2.02) to assemble metagenomic datasets of pair-end reads of about 120 bp length.

I use Velvetoptimiser to choose the appropriate k-mer then run vevleth again (as for some reason velvetoptimiser doesn't produce the Roadmaps file) and then try to run metavelvet. But when I do so I get a segmentation fault error like this:

[126.515876] Starting pebble resolution...
Segmentation fault (core dumped)

It's not a memory problem as I am running into a 512GB RAM machine (64 cpus) and the process is quite quick. (I've also checked with free -m)

I've compiled both velvet and metavelvet at MAKMERLENGTH=101, OPENMP=1.

My server is running on CentOS 7.1.1503 with g++ version 4.8.5.

Any ideas anyone?

Oh and Merry Christmas!

[makisthegreat]

Bumping for any answers... I could use any suggestions...

[rwinand]

Did you get any further with this problem? I'm also running into segmentation faults and tried everything I could imagine but nothing seems to work.

[karishma.guru]

Hello all,

I am facing the same issue of Segmentation fault (core dumped) when running meta-velvetg.

My data is size R1 and R2 being 5.2 GB each. What could be a possible solutions?

1. Can I subsample the data and run it?
2. Can I select contigs greater than 500bp from contig.fa file generated from velvetg and then run meta-velvetg > I will have to make changes to the Graph files also I think.

Please let me know how to go about this.

Regards
KKumar

Answer 1

[cclaing]

I'm not particularly familiar with metavelvet, but a SegFault is caused when a program tries to read and write memory outside of the bounds it has allocated for itself.

Have you compiled it with MAKMERLENGTH=101 or MAXKMERLENGTH=101? Maybe missing the X could cause it not to be flagged by the program causing undocumented behavior?

[makisthegreat]

No it was just a typo in my post. I compiled it with MAXKMERLENGTH=101. Is there a way to check what memory bounds the executables have allocated for itself? I've never heard that before.

Thank you

Answer 2

[yanhuihui]

I have come across the same problem, I run the metavelvet1.2.02 in Centos7 with memory of 256G，gcc-4.8.5, and I have got the same error [126.515876] Starting pebble resolution... Segmentation fault (core dumped)

However, I have googled a lot, there's somebody advice to downgrade the gcc-4.8.5 to gcc-4.6.2 to compile the Metavelvet. Then copy the whole folder to the new environment with gcc-4.8.5, then it may worked. But I am sticking at the stage of install gcc-4.6.2, It's really hard to make. So, I am not sure if this method really worked.