You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You can use abaqus as an example. Create a new directory, cd into it, do module load abaqus and then abaqus fetch job=knee_bolster && abaqus fetch job=knee_bolster_ef1 && abaqus fetch job=knee_bolster_ef2 && abaqus fetch job=knee_bolster_ef3. Then do curl https://raw.githubusercontent.com/OSC/osc-ood-config/master/ondemand.osc.edu/apps/myjobs/templates/Basic_ABAQUS_Job_Owens/basic_abaqus.sh | grep -v fetch > script.sh.
The result is a job directory, with no manifest, that looks like this:
efranz@owens-login02:~/tmp/abaqus_script_z$ ls -1
.
..
knee_bolster_ef1.inp
knee_bolster_ef2.inp
knee_bolster_ef3.inp
knee_bolster.inp
shellscript.sh
In the Job Composer do "New Job => From Specified Path" and put the full path in the Source path box. Leave everything else blank and Save.
The result is the job details claims "Script is not valid! Edit Job Options to select a valid script".
Click Job Options. Notice that "knee_bolster.inp" is selected as the job script! This is a bug. We should not auto-select the first file in the directory as the job script. Rather, we should
make a best guess as to which the job script is
not let you choose a job script that is a file of size > 65k
It is reasonable to expect that a user would open the Job Options, change Cluster from Ruby to Owens, since this is an Owens job script, click Save, and be unaware that they had set the Job Script to an input file. Then they submit and run into problems.
Note that the "validation" to avoid 65k size input files would in the above example only eliminate from the dropdown knee_bolster_ef1.inp and knee_bolster_ef2.inp.
Out of the remaining we would still have shellscript.sh.
This is where @AriettaZ's idea of a recommended options along with other options is good. For recommended options we can look for clues like:
first 1000 bytes has string #PBS or #SBATCH or other
first line starts with a shebang
file name ends with .sh, .job, .slurm, .batch, .qsub, .sbatch, .srun, .bsub
if there is a reliable way to find the mimetype, that could be used to ignore binary files (though the file command is not available on all systems)
Basically I think there would be a number of tests and if one passes we might include that in our recommendations.
You can use abaqus as an example. Create a new directory, cd into it, do
module load abaqus
and thenabaqus fetch job=knee_bolster && abaqus fetch job=knee_bolster_ef1 && abaqus fetch job=knee_bolster_ef2 && abaqus fetch job=knee_bolster_ef3
. Then docurl https://raw.githubusercontent.com/OSC/osc-ood-config/master/ondemand.osc.edu/apps/myjobs/templates/Basic_ABAQUS_Job_Owens/basic_abaqus.sh | grep -v fetch > script.sh
.The result is a job directory, with no manifest, that looks like this:
In the Job Composer do "New Job => From Specified Path" and put the full path in the Source path box. Leave everything else blank and Save.
The result is the job details claims "Script is not valid! Edit Job Options to select a valid script".
Click Job Options. Notice that "knee_bolster.inp" is selected as the job script! This is a bug. We should not auto-select the first file in the directory as the job script. Rather, we should
It is reasonable to expect that a user would open the Job Options, change Cluster from Ruby to Owens, since this is an Owens job script, click Save, and be unaware that they had set the Job Script to an input file. Then they submit and run into problems.
This is related to #104
The text was updated successfully, but these errors were encountered: