Have you ever felt that it is painful to run job scripts in a cluster? A lot of repetitive works can be done automatically. Jobkit contains a bunch of shell codes to help you run jobs smoothly. You just need to add the commands in .myshrc
to your .bashrc or .bash_profile
Remember to read through the .myshrc
file and make necessary changes! You may need to source .bashrc
to make the functions take effect
Here is a list of useful commands.
Are you sick of typing squeue -u username
? Define some commands abbreviations. For example, the q
defined in .myshrc can print your running jobs in queue.
The usage of every function is written as comments. Please read my comments carefully.
Define a template for submitting jobs. You can use the same template for every job submission. I offered one job template in the submit.job
file
When you submit jobs, don't just use sbatch
! Use the function run
to store the job number and submission time.
Have you ever been confused by what job is running? Use printjob_direc job_number
! It will bring you to the directory you submitted the job with job_number.
The jobs you submitted are saved in $HOME/jobs.number (should use the run
function to have jobs.number)
The jobs you finished are saved in $HOME/finished.jobs.number (should use the submit.job template to have finished.jobs.number)
How about syncing files? Are you tired of writing scp
? These three functions will help you! remote2local
, local2remote
, remotemulti
remote2local
will sync files from a remote computer to a local directory
local2remote
will sync local files to a remote computer
remotemulti
will generate commands for using remote2local
. It also checks whether the file exists.
Example usage:
local2remote pod /directory/on/pod/ local_files_1 local_files_2
will sync local files 1,2,... to the remote cluster saved as pod.
remote2local pod ./ a_file_on_pod
will sync a file on pod to the local folder './', which is equivalent to scp -r $username@pod-login1.cnsi.ucsb.edu:a_file_on_pod ./
.
You may know that scp
only sync one file at a time. However, you can also sync multiple files simultaneously by one command. As indicated by this website (https://stackoverflow.com/questions/16886179/scp-or-sftp-copy-multiple-files-with-single-command), it is done by scp -r $username@pod-login1.cnsi.ucsb.edu:/pod_path/{file1,file2,file3,file4} ./
. With my code, you can type something like remote2local pod ./ /pod_path/\{file1,file2,file3\}
In addition, you might also get bored by specifying the detailed path while using remote2local
. Thus, here is another function to generate remote2local
script for you automatically. Assume you want to transfer multiple files from stampede to pod. You can run remotemulti file1 file2 file3
on stampede, and it will output a job script looking like remote2local stampede ./ /stampede_PWD/\{file1,file2,file3\}
. Copy this command, and paste it on a pod terminal window, and the files will sync.
One more comment, copying and pasting remote2local stampede ./ /current_path/\{file1,file2,file3\}
seems to work only for bash
. If you have another interactive login shell like zsh
, this command may not work, and you can uncomment the related part in remotemulti
to output another command which directly uses scp scp -r username@pod-login1.cnsi.ucsb.edu:{file1,file2,file3} ./
Does your cluster clean files every several months? Are you sick of your important files being removed? Use touchall
!
The file sshconfig has two tricks to save typing passwords frequently
Use python draw.datafile_template.py yourdata.txt
to plot a data file