comprna/RATTLE

Segmentation Fault in Correcting Step

qhauck16 opened this issue · 20 comments

qhauck16 commented

Hi,
I'm attempting to use rattle on multiple fastq files from cDNA (actually, the files come from a bam file back-converted into a fastq using bedtools bamToFastq, as I'm interested in a specific set of reads).
Anyways, I have been able to run the clustering step fine on multiple of these files after filtering using the code commented on another issue (I was originally having the std:bad alloc error), but the ./rattle correct step throws a segfault:

Reading fasta file... Done
Segmentation fault (core dumped)                                                 ] 1154/147477 (0.782495%)

I am not very experienced at reading fastq files, as I have only recently begun working with them, but as far as I can tell, the fastq files have no obvious issues at the read number indicated by the number it fails at (the same number is consistently the issue when run with one thread)

Any help is appreciated!
Best,
Quinn

novikk commented

Hi!

The easiest way for me to debug this is to get the fastq file. Is it possible in this case?

Thanks

qhauck16 commented

Yes, how should I contact you?

novikk commented

You can send it via email to ivan.delarubia at upf.edu

Thanks!

claumer commented

Hey, just thought I'd chip in to mention I've seen a similar problem. Segfault within the first second of running rattle correct:

Reading fasta file... Done
Segmentation fault (core dumped)

Happy to send fastq along to the email address you gave above if useful. It's a subsampled file so only 50,000 ONT reads.

Kirovez commented

Thank you! It helps to perform clustering. But the next step again gives the error:

Reading fasta file... Done
Segmentation fault (core dumped)

My command was:
extract_clusters -i my_linux_modified.fastq -c clusters.out -o clusters

novikk commented

Hi @Kirovez , remember to add the "--fastq" flag.

In the upcoming updates this won't be necessary but it is still required right now.

pfrank76 commented

Hi, I had the same problem mentioned above in the "rattle correct" step:

Reading fasta file... Done
Segmentation fault (core dumped)

Did you find the solution to this issue?

Thanks!

Kirovez commented
bauerlev commented

Hello,
I ran into a similar issue during the correcting step. @novikk do you have any advice/updates on fixes coming to rattle?

eKariuki-sleepy commented

I believe I am facing a similar challenge using cDNA data. Any fixes yet?

EduEyras commented
eileen-xue commented

Hi all,

In case anyone still watching this issue. We have an update to fix this issue now.

Thanks,
Eileen

eileen-xue commented
bauerlev commented

Eileen, here is my script. I've given the job 10 cores and 50G of memory. For clarification the error I'm receiving is an "Illegal instruction (core dumped)"

ml GCCcore/10.1.0
ml GCC/10.1.0

mkdir clusters_extracted clusters_corrected2

RATTLE/rattle cluster -i coff_filt_cdna_genespace.fastq.gz -t 10

RATTLE/rattle extract_clusters -i coff_filt_cdna_genespace.fastq.gz -c clusters.out -o clusters_extracted/ --fastq

RATTLE/./rattle correct -i coff_filt_cdna_genespace.fastq.gz -c clusters.out -o clusters_corrected2/ -t 10

eileen-xue commented

Hi Levi,

Try to use the gcc version 9.3.0. Also, which CPU do you use? Does it support sse4.1? RATTLE sub-module spoa has this error "Illegal instruction (core dumped)" with some old processors.

Thanks,
Eileen

bauerlev commented

Thank you for this tip! It's hard to tell what CPUs I'm using as I run my jobs on a large university hpcc, but I was able to play around and find a dev-node that ran the toy dataset and run the job using those resources. Also learned that "illegal instruction" in general indicates software incompatibility for all programs which was a good lesson to learn.

lpryszcz commented

hi, I'm also experiencing Segmentation fault (core dumped) during with rattle correct (the latest version from github).
I tried on two different computers, both with SSE4.1 support and GCC v9.4.
The same error happens with toy dataset. The commands I'm using

s=test.rna; f=~/src/RATTLE/toyset/rna/input/sample.fastq
mkdir -p rattle/$s/clusters
~/src/RATTLE/rattle cluster -t 4 --rna --iso -i $f -o rattle/$s/clusters
~/src/RATTLE/rattle correct -t 4 -i $f -c rattle/$s/clusters -o rattle/$s
~/src/RATTLE/rattle polish -t 4 -i rattle/$s/consensi.fq -o rattle/$s
eileen-xue commented

Hi,

In your RATTLE correct command, it should be -c rattle/$s/clusters/clusters.out which points to clusters.out file location, not the folder location.

Eileen

  • 😄1
lpryszcz commented

thank you! pointing to clusters.out works.
could rattle report more meaningful I/O error?

eileen-xue commented

RATTLE checks whether the input and output files or folder exits and provides error report. However, it won't check whether it points to a file or a folder.