help
help > RE: Job does not continue past "Step 5/7: Importing ROI data"
Apr 20, 2018 09:04 AM | Alfonso Nieto-Castanon - Boston University
RE: Job does not continue past "Step 5/7: Importing ROI data"
Dear Martijn,
The 9-minute limit must be the default settings in your own cluster (CONN does not specify any time-limit on its own unless you explicitly include that info in the cluster configuration options). In a slurm cluster you can explicitly specify your requested job time-limit using the flag "--time=HH:MM:SS", so I would suggest simply editing in CONN's 'cluster settings' GUI your slurm profile to add in the field named 'additional submit settings . in-line ' the string
--time=02:00:00
and that should do the trick
Hope this helps
Alfonso
Originally posted by Martijn Teuchies:
The 9-minute limit must be the default settings in your own cluster (CONN does not specify any time-limit on its own unless you explicitly include that info in the cluster configuration options). In a slurm cluster you can explicitly specify your requested job time-limit using the flag "--time=HH:MM:SS", so I would suggest simply editing in CONN's 'cluster settings' GUI your slurm profile to add in the field named 'additional submit settings . in-line ' the string
--time=02:00:00
and that should do the trick
Hope this helps
Alfonso
Originally posted by Martijn Teuchies:
Dear Alfonso,
When trying to run my analysis (5 participants, each with 6 resting state sessions) on a slurm computing cluster, the job does not continue past "Step 5/7: Importing ROI data".
Regarding the submission of the job, I think I have all the right settings: (I also ran the test, which finished correctly).
profile name: Slurm computer cluster
command used to submit a job:
sbatch --job-name=JOBLABEL --error=STDERR --output=STDOUT OPTS SCRIPT
check status: squeue --jobs=JOBID
I have not added any additional commands or submit settings. When I submit the job to the cluster, I start 5 parallel jobs, as suggested by the conn toolbox.
The ROI's I specified are subject and session specific grey matter, white matter and CSF images (generated by the segmentation during the preprocessing which I ran outside of conn using SPM12). This seems to work fine, as the segmentation step is skipped in step 2, so conn seems to find and use these images correctly. Additionally, I also specified 8 ROI's based on previous studies (I created these ROI's using MarsBar, by selecting a sphere of 4mm around peak voxel coordinates and writing the images to the same space / dimensions and voxel size as the normalized functional images, so I did not explicitly coregister them, but I assume that's not needed as they are both in the same MNI space). These 8 ROI's are the same for each participant.
I also selected the option to have the data set 1 (the functional images minus the 's' prefix).
When I submit the job, it starts running correctly through steps 1 - 4. When I check on the job status on the cluster, I see that each of the 5 jobs has about 9 minutes of run time to complete the set-up steps. After these 9 minutes, the job is still at step 5, importing the ROI data and then just runs out of time.
So I guess I have two questions, 1) Is this 9 minute window something that CONN determines and passes onto the cluster, or is this something our slurm cluster decides? And if conn determines this, can I extend this?
2) did I specify something wrong regarding the ROI's causing the job to be unable to get beyond step 5?
I'll attach the output / log files I have.
Kind regards,
Martijn Teuchies
When trying to run my analysis (5 participants, each with 6 resting state sessions) on a slurm computing cluster, the job does not continue past "Step 5/7: Importing ROI data".
Regarding the submission of the job, I think I have all the right settings: (I also ran the test, which finished correctly).
profile name: Slurm computer cluster
command used to submit a job:
sbatch --job-name=JOBLABEL --error=STDERR --output=STDOUT OPTS SCRIPT
check status: squeue --jobs=JOBID
I have not added any additional commands or submit settings. When I submit the job to the cluster, I start 5 parallel jobs, as suggested by the conn toolbox.
The ROI's I specified are subject and session specific grey matter, white matter and CSF images (generated by the segmentation during the preprocessing which I ran outside of conn using SPM12). This seems to work fine, as the segmentation step is skipped in step 2, so conn seems to find and use these images correctly. Additionally, I also specified 8 ROI's based on previous studies (I created these ROI's using MarsBar, by selecting a sphere of 4mm around peak voxel coordinates and writing the images to the same space / dimensions and voxel size as the normalized functional images, so I did not explicitly coregister them, but I assume that's not needed as they are both in the same MNI space). These 8 ROI's are the same for each participant.
I also selected the option to have the data set 1 (the functional images minus the 's' prefix).
When I submit the job, it starts running correctly through steps 1 - 4. When I check on the job status on the cluster, I see that each of the 5 jobs has about 9 minutes of run time to complete the set-up steps. After these 9 minutes, the job is still at step 5, importing the ROI data and then just runs out of time.
So I guess I have two questions, 1) Is this 9 minute window something that CONN determines and passes onto the cluster, or is this something our slurm cluster decides? And if conn determines this, can I extend this?
2) did I specify something wrong regarding the ROI's causing the job to be unable to get beyond step 5?
I'll attach the output / log files I have.
Kind regards,
Martijn Teuchies
Threaded View
| Title | Author | Date |
|---|---|---|
| Martijn Teuchies | Apr 19, 2018 | |
| Alfonso Nieto-Castanon | Apr 20, 2018 | |
| Yang Yang | Apr 21, 2018 | |
| Martijn Teuchies | Apr 20, 2018 | |