help
489 Subscribers 
help > Submitting job to SLURM cluster
Showing 1-5 of 5 posts
Apr 13, 2018 02:04 PM | Martijn Teuchies - University of Florida
Submitting job to SLURM cluster
Dear people at CONN,
I am a new conn user and am currently figuring out how to use conn on the University of Florida computing cluster using SLURM. I ran the test to see whether conn is compatible with the cluster, which finished successfully, so I used those settings as the default. I am trying to run a project with 5 participants, each having 6 resting state sessions, using preprocessed data (SPM12). I specified structural, functional (incl set 1), gray matter, white matter and csf images for each pp for each of the 6 sessions, as well as motion covariate files for each session. When I finish the set up and click done, I select the option to submit to the SLURM cluster, using 5 parallel jobs, as suggested by the gui.
Then, however, it does not continue past the 'submitted' message, which remained for over an hour, after which I cancelled the job. I looked at the logfiles, and in the stderr file it reads:
"Error using cd Too many input arguments." and "ERROR: ld.so: object 'librrfaker.so' from LD_PRELOAD cannot be preloaded: ignored."
Is this what's causing the job to freeze on the submitted screen? And would you have an idea what is causing the error and how I might solve it?
Kind regards,
Martijn
I am a new conn user and am currently figuring out how to use conn on the University of Florida computing cluster using SLURM. I ran the test to see whether conn is compatible with the cluster, which finished successfully, so I used those settings as the default. I am trying to run a project with 5 participants, each having 6 resting state sessions, using preprocessed data (SPM12). I specified structural, functional (incl set 1), gray matter, white matter and csf images for each pp for each of the 6 sessions, as well as motion covariate files for each session. When I finish the set up and click done, I select the option to submit to the SLURM cluster, using 5 parallel jobs, as suggested by the gui.
Then, however, it does not continue past the 'submitted' message, which remained for over an hour, after which I cancelled the job. I looked at the logfiles, and in the stderr file it reads:
"Error using cd Too many input arguments." and "ERROR: ld.so: object 'librrfaker.so' from LD_PRELOAD cannot be preloaded: ignored."
Is this what's causing the job to freeze on the submitted screen? And would you have an idea what is causing the error and how I might solve it?
Kind regards,
Martijn
Apr 13, 2018 05:04 PM | Alfonso Nieto-Castanon - Boston University
RE: Submitting job to SLURM cluster
Hi Martijn,
Could you please zip/send me the other contents of your qlog folder (minimally one info.mat file and at least one node.*.mat file)?
Thanks
Alfonso
Originally posted by Martijn Teuchies:
Could you please zip/send me the other contents of your qlog folder (minimally one info.mat file and at least one node.*.mat file)?
Thanks
Alfonso
Originally posted by Martijn Teuchies:
Dear people at CONN,
I am a new conn user and am currently figuring out how to use conn on the University of Florida computing cluster using SLURM. I ran the test to see whether conn is compatible with the cluster, which finished successfully, so I used those settings as the default. I am trying to run a project with 5 participants, each having 6 resting state sessions, using preprocessed data (SPM12). I specified structural, functional (incl set 1), gray matter, white matter and csf images for each pp for each of the 6 sessions, as well as motion covariate files for each session. When I finish the set up and click done, I select the option to submit to the SLURM cluster, using 5 parallel jobs, as suggested by the gui.
Then, however, it does not continue past the 'submitted' message, which remained for over an hour, after which I cancelled the job. I looked at the logfiles, and in the stderr file it reads:
"Error using cd Too many input arguments." and "ERROR: ld.so: object 'librrfaker.so' from LD_PRELOAD cannot be preloaded: ignored."
Is this what's causing the job to freeze on the submitted screen? And would you have an idea what is causing the error and how I might solve it?
Kind regards,
Martijn
I am a new conn user and am currently figuring out how to use conn on the University of Florida computing cluster using SLURM. I ran the test to see whether conn is compatible with the cluster, which finished successfully, so I used those settings as the default. I am trying to run a project with 5 participants, each having 6 resting state sessions, using preprocessed data (SPM12). I specified structural, functional (incl set 1), gray matter, white matter and csf images for each pp for each of the 6 sessions, as well as motion covariate files for each session. When I finish the set up and click done, I select the option to submit to the SLURM cluster, using 5 parallel jobs, as suggested by the gui.
Then, however, it does not continue past the 'submitted' message, which remained for over an hour, after which I cancelled the job. I looked at the logfiles, and in the stderr file it reads:
"Error using cd Too many input arguments." and "ERROR: ld.so: object 'librrfaker.so' from LD_PRELOAD cannot be preloaded: ignored."
Is this what's causing the job to freeze on the submitted screen? And would you have an idea what is causing the error and how I might solve it?
Kind regards,
Martijn
Apr 13, 2018 06:04 PM | Martijn Teuchies - University of Florida
RE: Submitting job to SLURM cluster
Hi Alfonso,
Attached you will find the zipped files.
Thank you,
Martijn
Attached you will find the zipped files.
Thank you,
Martijn
Apr 13, 2018 06:04 PM | Alfonso Nieto-Castanon - Boston University
RE: Submitting job to SLURM cluster
Hi Martijn,
Thanks for the additional info. It seems the problem was a bug caused by the presence of a space in the "Research Projects" folder name which was confusing CONN's job submission scripts. I am attaching a patch that should fix this issue. Please let me know if that works (this patch is for conn 2018a release, to install the patch simply copy the attached file to your conn distribution folder overwriting the file with the same name there)
Best
Alfonso
Originally posted by Martijn Teuchies:
Thanks for the additional info. It seems the problem was a bug caused by the presence of a space in the "Research Projects" folder name which was confusing CONN's job submission scripts. I am attaching a patch that should fix this issue. Please let me know if that works (this patch is for conn 2018a release, to install the patch simply copy the attached file to your conn distribution folder overwriting the file with the same name there)
Best
Alfonso
Originally posted by Martijn Teuchies:
Hi Alfonso,
Attached you will find the zipped files.
Thank you,
Martijn
Attached you will find the zipped files.
Thank you,
Martijn
Apr 15, 2018 12:04 AM | Martijn Teuchies - University of Florida
RE: Submitting job to SLURM cluster
Dear Alfonso,
Thank you very much your swift reply and solution, this seems to take care of the problem.
Kind regards,
Martijn
Thank you very much your swift reply and solution, this seems to take care of the problem.
Kind regards,
Martijn