devel version duplicates jobs

[Marcel Falkiewicz]

Hello,

 I've been running the -devel version of nipype and there seems to be a bug which causes each job to duplicate (in my case, fsl_glm). Each command is run twice, with identical command line:

$ ps ux | grep fsl_glm

marcel   18913  0.3  0.0   4180   584 pts/1    S+   16:13   0:00 /bin/sh -c fsl_glm -i /mnt/alpha/SDT_MRI/new/resting/apmsk2/timeseries_smooth_filt_trans_masked.nii.gz -d /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.mat -o timeseries_smooth_filt_trans_masked_glm.nii.gz -c /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.con --out_cope=beta.nii.gz

marcel   18914 91.5  7.0 2924028 1742820 pts/1 R+   16:13   0:07 fsl_glm -i /mnt/alpha/SDT_MRI/new/resting/apmsk2/timeseries_smooth_filt_trans_masked.nii.gz -d /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.mat -o timeseries_smooth_filt_trans_masked_glm.nii.gz -c /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.con --out_cope=beta.nii.gz

marcel   18915  0.2  0.0   4180   588 pts/1    S+   16:13   0:00 /bin/sh -c fsl_glm -i /mnt/alpha/SDT_MRI/new/resting/apmsk2/timeseries_smooth_filt_trans_masked.nii.gz -d /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.mat -o timeseries_smooth_filt_trans_masked_glm.nii.gz -c /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.con --out_cope=beta.nii.gz

marcel   18916  103  7.0 2924028 1746508 pts/1 R+   16:13   0:07 fsl_glm -i /mnt/alpha/SDT_MRI/new/resting/apmsk2/timeseries_smooth_filt_trans_masked.nii.gz -d /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.mat -o timeseries_smooth_filt_trans_masked_glm.nii.gz -c /mnt/alpha/SDT_MRI/new/resting/_idx_67/feet/run0.con --out_cope=beta.nii.gz

And the above is just with Linear plugin.

 It's ok for linear processing, but drains the machine out of memory quicky when using any parallel method - for both condor and multiproc. Is it a known bug, or maybe I'm doing something wrong? Have you got any ideas how to correct this?

Best regards,

 Marcel

[Satrajit Ghosh]

hi marcel,

this is fixed in the current master. it comes from sending the stdout and stderr file descriptors to multiprocess. we aren't sure exactly why it forks yet, but we have disabled this in the multiprocessing plugin.

it will be great if you can see if the current master solves the problem for you.

cheers,

satra

--
 
---
You received this message because you are subscribed to the Google Groups "NiPy Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to nipy-user+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Marcel Falkiewicz]

Hello Satra,

 it doesn't help - I pulled the master today and the problem persists.

[Satrajit Ghosh]

hi marcel,

could i take a look at your script? 

or can you ensure that the terminal_output input of any commandline interface is set to anything other than "file"

cheers,

satra