STAR alignment process killed after sometime

Question

[archisman]

Hi, I am trying to run STAR alignment using the command below:

   STAR \
--runThreadN 8 \
--runMode genomeGenerate \
--genomeDir /scratch/workspace/rnaseq/STAR_genome/ \
--genomeFastaFiles /scratch/workspace/rnaseq/refs/GRCh38.d1.vd1.fa \
--sjdbOverhang 100 \
--sjdbGTFfile /scratch/workspace/rnaseq/refs/gencode.v32.annotation.gtf

And this is happening every single time

Dec 19 00:31:17 ..... started STAR run
Dec 19 00:31:17 ... starting to generate Genome files
Killed

Here is the memory status of the node that I'm using

top - 00:46:35 up 127 days, 13:01,  3 users,  load average: 0.35, 0.49, 0.55
Tasks: 575 total,   1 running, 564 sleeping,   3 stopped,   7 zombie
%Cpu(s):  1.7 us,  1.6 sy,  0.9 ni, 95.6 id,  0.1 wa,  0.0 hi,  0.1 si,  0.0 st
KiB Mem : 13148420+total,   528652 free,  5211976 used, 12574358+buff/cache
KiB Swap: 13421772+total, 12861864+free,  5599076 used. 12442298+avail Mem

I don't understand why. It would be great help if someone can explain. Thanks!

[GenoMax]

killed indicates that process is hitting some kind of process/hardware limit. Was the job killed right away or after some time?

Also include SLURM part of the command. I belive you said you were using that in previous posting?

[archisman]

Thanks, genomax for replaying. The process is failing after some time not immediately. Actually I was asking in the previous post that how to use SLURM ? Because sbatch command is not working and also I recently started working on node or Linux code in that matter.

[GenoMax]

Ah. It is possible that your cluster admins are killing jobs that run outside the SLURM job scheduler. In that case you should definitely use SLURM.

We can't provide you with a tutorial on how to use SLURM. For that you should contact your local IT support folks. But a potential SLURM job submission via command line will wrap your own command above in needed SLURM bits like so

(you will need to use real options that are locally available e.g. in place of partition use job queue you want to submit to, for this to work)

$ sbatch -p partition -n 8 -N 1 -t 1-0 --mem=30g -o file.out -e file.err --wrap="STAR --runThreadN 8 --runMode genomeGenerate and so on"