general-discussion > changing the tmp folder doesnt not work fully
Showing 1-5 of 5 posts
Display:
Results per page:
Dec 5, 2011  05:12 PM | Alexis Machado
changing the tmp folder doesnt not work fully
Hi Pierre, christian

I work on the BIC with the last version of NIAK, and i have issue regarding the space allowed in the /tmp folder which is limited.
I changed the tmp folder destination editing the gb_niak_tmp variable in commands/misc/niak_gb_vars.m ( /mfip/mfip1/alexis/TMP_NIAK/).

I have also edited the psom_gb_vars temporay folder to be sure to copy anything on /tmp. This file is however in my current folder ( there is another copy of psom_gb_vars) in the NIAK toolbox folder


Unfortunately i realized that NIAK continue to place some files ( not all) in the /tmp folder and at some point my pipeline crash because the /tmp is full.

 i.g: /tmp/MATLAB_Files_05-Dec-2011_11_14_52/niak_tmp_slice_timing_subject1_session1_run1_22119996CA1_MNS_BOLD_run01.dat .....
      /tmp/MATLAB_Files_05-Dec-2011_11_19_19/motion_target_subject1_session1_run2.eqsub

I have to delete these files manually, delete some
files whcih are in running and failed state in the log folder  and resume my
pipeline.


It seem that the problem occur in session and qsub mode:
Please fing my psom_gb_vars config for the qsub mode in case the source mode is involved

% how to invoke matlab  
gb_psom_command_matlab = '/export01/local/matlab10a/bin/matlab';

% how to invoke octave
gb_psom_command_octave = 'octave';

% Options for the sge qsub system, example : '-q all.q@yeatman,all.q@zeus'
% will force qsub to only use the yeatman and zeus workstations through the
% queue called all.q
gb_psom_qsub_options = '-r y -q grova.q';

% Options for the shell in batch or qsub modes
gb_psom_shell_options = 'source /data/aces/aces1/pbellec/public/niak-dev/extensions/psom/example_init_qsub.sh';

% Options for the execution mode of the pipeline
gb_psom_mode = 'qsub';

% Options for the execution mode of the pipeline manager
gb_psom_mode_pm = 'batch';

% Options for the maximal number of jobs
gb_psom_max_queued = 10;

% Initialization of matlab
gb_psom_init_matlab = '';

% Matlab search path. An empty value will correspond to the search path of
% the session used to invoke PSOM_RUN_PIPELINE. A value 'gb_psom_omitted'
% will result in no search path initiated (the default Octave path is
% used).
gb_psom_path_search = '';

% where to store temporary files
gb_psom_tmp =['/mfip/mfip1/alexis/TMP_NIAK' filesep];


Best regards
Alexis

     
Dec 6, 2011  06:12 PM | Christian Dansereau
RE: changing the tmp folder doesnt not work fully
Hi Alexis,

you should rename your local version of psom_gb_vars.m to psom_gb_vars_local.m

by default psom_gb_vars.m will look if there is a file named psom_gb_vars_local.m if there is one he will execute that file instead of the default one


Regards
Christian
Dec 7, 2011  05:12 PM | Alexis Machado
RE: changing the tmp folder doesnt not work fully
Hi
Thanks i didn't know this feature.

The problem is still here. The problem is that niak both use the /tmp and the specified tmp folder.Becuase the /tmp is limited in size , niak crash.

please find my pipeline file in attachement: you will see i have creted an if statement if i want to use session or qsub mode.


Plese find my configuration:
1-I downloaded the last verison of NIAK and installed it on mfip/mfip1/alexis/...

2-I changed the niak_gb_var.m located in the /mfip/mfip1/alexis/Sofwares/fMRI/NIAK/NIAK6.4/niak-0.6.4.1/commands/misc/niak_gb_vars.m



Please find the changes:
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% The following variables are needed for very fast initialization %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% where to store temporary files
% gb_niak_tmp = [filesep 'tmp' filesep];
gb_niak_tmp = ['/mfip/mfip1/alexis/TMP_NIAK' filesep];


3- I created a psom_gn_var_local.m in my current directory ( where i lauch the pipeline) that i call (psom_gn_var) when i want to execute a qsub task

Please find the changes:
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% The following variables need to be changed to configure the pipeline %%
%% system                                                               %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% how to invoke matlab  
gb_psom_command_matlab = '/export01/local/matlab10a/bin/matlab';

% how to invoke octave
gb_psom_command_octave = 'octave';

% Options for the sge qsub system, example : '-q all.q@yeatman,all.q@zeus'
% will force qsub to only use the yeatman and zeus workstations through the
% queue called all.q
gb_psom_qsub_options = '-r y -q grova.q';

% Options for the shell in batch or qsub modes
gb_psom_shell_options ='source /mfip/mfip1/alexis/Sofwares/fMRI/NIAK/NIAK6.4/niak-0.6.4.1/extensions/psom-0.8.8/example_init_qsub.sh';
% gb_psom_shell_options ='';

% Options for the execution mode of the pipeline
gb_psom_mode = 'qsub';

% Options for the execution mode of the pipeline manager
gb_psom_mode_pm = 'batch';

% Options for the maximal number of jobs
gb_psom_max_queued = 10;

% Initialization of matlab
gb_psom_init_matlab = '';

% Matlab search path. An empty value will correspond to the search path of
% the session used to invoke PSOM_RUN_PIPELINE. A value 'gb_psom_omitted'
% will result in no search path initiated (the default Octave path is
% used).
gb_psom_path_search = '';

% where to store temporary files
gb_psom_tmp = ['/mfip/mfip1/alexis/TMP_NIAK' filesep];




Best regards
Alexis
Dec 9, 2011  03:12 PM | Alexis Machado
RE: changing the tmp folder doesnt not work fully
PROBLEM SOLVED:


In matlab configuration, under file>preference >general the deleting files

tooble button was set to move to a temporary folder and not delete permanently.



As a results all the files that were cleaned went to the temporay folder





Thank you help

best regards

Alexis
Dec 9, 2011  07:12 PM | Pierre Bellec
RE: changing the tmp folder doesnt not work fully
Thanks a lot Alexis for self-solving your problem and sharing the solution. That will be very helpful if we ever run into the same issue again in the future.

Best,

Pierre