# loop modelling error on Rosetta3.2 or Rosetta3.3

6 posts / 0 new
loop modelling error on Rosetta3.2 or Rosetta3.3
#1

I am trying to remodel the loop giving low-resolution electron density map. I have tried Rosetta3.2 and 3.3 however the job could not complete when I run in mpi mode. Could you have any solution for this?

The log of error is like this. It is due to mpi.

output filename: .//outdir_2/S_0003.pdb
ERROR: Unable to open file: .//outdir_2/S_0003.pdb
ERROR:: Exit from src/protocols/jobdist/JobDistributors.cc line:727

Post Situation:
Thu, 2011-09-29 03:58
exchhattu

That code is still using the old job distributor. It wants you to pre-create (run_directory)/outdir_2 for it to put results in. Rosetta will create the output PDB files but refuses to create the output directory to put it in. I would guess you need outdir_0 through outdir_N, where N is the number of processors in the MPI job.

Thu, 2011-09-29 07:00
smlewis

Thanks a lot for helpful information.

Could you tell me how we can get rid of this?

Thu, 2011-09-29 18:51
exchhattu

Get rid of what? The error? You need to create the output directories it expects manually ahead of time...

Thu, 2011-09-29 18:53
smlewis

By the way, in these sorts of situations, some bash (or whatever shell you're using) knowledge makes life easier.

For example, entering something like:
 for i in {0..31} do mkdir outdir_\${i} done

on the bash shell command line is a lot more pleasant than manually creating each directory by itself.

Fri, 2011-09-30 10:19
rmoretti

Exactly. For a loop modeling mpi job, you would need to create one output directory per processor to avoid that independent processes overwrite each others outfiles. The same is true for running jobs in parallel on a cluster.

Sat, 2011-10-01 10:23
rpache