Skip to content

Latest commit

 

History

History
85 lines (64 loc) · 2.38 KB

README.md

File metadata and controls

85 lines (64 loc) · 2.38 KB
Raw

HPC Workflow Recipes

transfer-files

The purpose of this script is to enable transfer of directories/files to/from shared areas whilst on a worker node, within an interactive session or batch job.

[te1st@node001 [stanage] ~]$ transfer-files SharedArea WorkerDirectory

sequenceDiagram
    participant WorkerNode
    participant LoginNode
    participant SharedArea
    participant WorkerDirectory
    WorkerNode->>LoginNode: 1.SSH Connection
    activate LoginNode
    LoginNode->>SharedArea: 2.Initiate rsync
    SharedArea-->>WorkerDirectory: 3.Sync files
    LoginNode-->>WorkerNode: 4.Close SSH Connection
    deactivate LoginNode
Loading

Setup

Move file-transfers to ~/bin, make executable and add to .bashrc

mkdir -p ${HOME}/bin/
mv transfer-files ${HOME}/bin/transfer-files 
chmod +x ${HOME}/bin/transfer-files
echo 'export PATH="$PATH:$HOME/bin"' >> ~/.bashrc
source ~/.bashrc

NOTE: This script does not remove files once they are synced.

Recommend assigning directories to variables. For example, create a setup.sh script:

#!/bin/bash
shared='/shared/path/to/directory/'
working_dir='/mnt/parscratch/users/${USER}/'

These variables are used in the examples below.

Interactive usage

Usage:

transfer-files SOURCE DESTINATION

Example:

transfer-files ${shared}/some/path/ ${working_dir}/some/path/

Batch jobs

Can also be submitted to the SLURM scheduler as a job submission (ex. useful for sbatch dependencies):

Default resource requests are 4G of memory and 10 minutes.

Usage:

sbatch transfer-files SOURCE DESTINATION

Examples:

sbatch transfer-files ${shared}/some/path/ ${working_dir}/some/path/
sbatch --time=00:20:00 transfer-files ${shared}/some/path/ ${working_dir}/some/path/

Caution: We need to be careful with trailing slashes

Trailing Slash in Source Directory: copies the contents of the source directory, but not the directory itself, into the destination.

If you want to copy the source directory itself into the destination, without merging its contents, you should omit the trailing slash: rsync source destination

Trailing Slash in Destination Directory: copies the source into that directory, preserving its name.

If you don't want the source directory to be included in the destination, use a destination path without a trailing slash: rsync source/ destination