This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Error when running OMA on PBS with job array

Hi there,

I'm trying to run OMA on PBS with parallel jobs. To do this, I first ran OMA -c for completing the database conversion. Then, I submitted a PBS script specifying to run a job array of 32 tasks with the following line:

#PBS -t 1-32

The jobs fail with the following error message: 

Error, (in ReadDb) Failed to mmap Pat index

I also tried to do the database conversion when running the parallelized jobs and I get errors like this example:

Pre-processing input (st=2, dftype=1, totc=18235392, i=14680064

area of error:
"GMTPGGGHLEGECCDDAVYADRSAEISAWL\000\000\000\000\000\000\000\000\000\000\000\000\..." 
                               ^
error detected in position 14680065
Error, (in ReadDb) invalid character in input SEQ

This does not occur when running OMA -c first. Do you have any idea about what is wrong?

Thanks in advance!

oma orthologs pbs parallelization
• 3,136 views • link
written by gonzalez.pech.ra • gonzalez.pech.ra

2 answers

Hi,

which version of OMA standalone are you using? On version (I think it was 1.0.5) was buggy in that respect that the conversion was done by all running processes if started with a scheduler. The multiple process write to the same file and create garbage. In your case, the ascii character 0x00 was introduced into the sequence, which is not allowed on most systems.

With the most recent version, this bug should be fixed. If you observe the problem also with this version, please let us know.

Best

• 0 views • link
written by Adrian Altenhoff Adrian Altenhoff

Hi,

I just updated OMA to the most recent version 1.1.2 and had the same problem posted here with the array. I was running the 1.0.6 before and did not have this issue. Is there something that can be done to start oma directly with the array or is it the only option to start with -c and then restart with -s? Error was: (on SLURM)

Job 197 [pid 13121]: waiting for too long. abort.

It seems that your parallelisation framework hasn't launched the master job properly, or you are using a very slow filesystem. Please restart a single process of OMA with the command line option "-c" to just convert the databases. Once this job successfully finished, you can restart OMA with many parallel jobs.

Thanks!

• 0 views • link
written by tauanacunha • tauanacunha

Thanks for your reply.

It is actually the most recent version:

[...~]$ OMA -v
OMA version 1.0.6
• 0 views • link
written by gonzalez.pech.ra • gonzalez.pech.ra

Log in to answer this question.

More posts like this