Parallel run using migrate-n-mpi stops running, no clear error message

[Thijs Mattheus Peter Bal]

Hello,

My runs using the parallel version "migrate-n-mpi" just stop after a few hours, with no clear error message appearing in the printed output. The dataset (see attached "migrate-n-mpi_8pop_38loci_infile") can be summarised as:

• 8 populations
• 38 loci (SNPs)

For now I was mostly just using the default settings (see attached "migrate-n-mpi_8pop_38loci_parmfile"), but I made sure the input data type was set to SNPs. In the printed output I see that the burn-in and sampling of each loci seems to be completed, but shortly after this stage the program just stops running (see attached "migrate-n-mpi_8pop_38loci_outputlog"). Since I was running only 1 replicate I initiated the run using (38 x 1) +1 = 39 cores.

What could be the problem, or where could I possibly find out what the problem is? First I thought it could have been a memory issue, but this does not seem the case.

(ps. I have been running the same dataset under the same settings in the single CPU "migrate-n" version and that seems to work fine. In the printed output I see that the program has reached the stage at which the parameters are being printed and the prognosed run time is given. However, this run will just take very long so I would like to successfully run the parallel version.)

I appreciate any suggestions/tips.

Thijs M.P. Bal

[Peter Beerli]

Thijs

In the parmfile I changed 

mig-histogram=YES:0.001000:/home/thijs/projects/Stickleback/WGS2018/3s/MIGRATE/Output-MPI/ThreeSpined_filtered_SNP_DP3_GQ20_Bi_Miss10_MAF005_PopOrder_subset_MIG-N

to

mig-histogram=NO

and it works fine, particularly for first runs I would always try to do very short runs e.g. use long-inc=5 and long-sample=1000 and burnin-in=100

to check whether it works and then increase the values.

I find the migration histograms through time rather unhelpful with data such as a single-snps because an individual locus does not have great information about what happened in the past. 

if you start migrate-n in a shell in a particular directory you will not need to give the absolute path to the files, the filename only will be fine.

Peter

--
You received this message because you are subscribed to the Google Groups "migrate-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to migrate-suppo...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/migrate-support/52e660c4-2ad0-4557-bdbc-d41c8c3357d9n%40googlegroups.com.
<migrate-n-mpi_8pop_38loci_outputlog.log><migrate-n-mpi_8pop_38loci_infile.txt><migrate-n-mpi_8pop_38loci_parmfile>

[Thijs Mattheus Peter Bal]

Hi Peter,

Thank you for the quick response and for having a look at my files. I will start a new run following your recommendations.

Thijs

Op donderdag 7 juli 2022 om 15:53:28 UTC+2 schreef Peter:

[Thijs Mattheus Peter Bal]

Hi Peter,

Additional update: the run has completed and the output file was successfully created. The histograms and ESS look good already, put will carefully experiment now whether longer runs make any difference. Thanks again, looking forward to test some different types of population genetic models now! 

Thijs

Op donderdag 7 juli 2022 om 16:08:30 UTC+2 schreef Thijs Mattheus Peter Bal: