willfurnass

Random Grid Engine tips and tricks

The following work with Son of Grid Engine (SGE) 8.1.9 as configured on the University of Sheffield's ShARC and Iceberg clusters.

Jobs with dependancies

You can use the -hold_jid <<job-name or job-name>> option to make jobs run only when other jobs have finished, rather than having jobs start and sit waiting for other tasks to complete.

willfurnass

#!/bin/sh
# David Jones and Will Furnass (University of Sheffield)

# `qrsh` is often the most convenient utility for starting interactive shell
# sessions on Sun/Son of Grid Engine (SGE) clusters but it has some limitations (as
# configured on the ShARC/Iceberg clusters):
# 
# - none of the standard SGE environment vars (e.g. JOB_ID, PE_HOSTFILE,
#   NSLOTS) that are defined in qsub/qsh batch/interactive sessions are set
#   (annoying but not the end of the world);

willfurnass

#!/bin/bash
# Will Furnass
# Oct 2017

if [[ $# -lt 1 ]]; then
    echo 1>&2 "Extract list of user email addresses from Synology DSS '.dss' config dump"
    exit 1
fi

dss_path="$1"

willfurnass

import tempfile                                                                                                                                                                                                   
import os
import time

import paramiko

# Hostname of a 'submission host' in the Grid Engine cluster we want to submit a job to
hostname = 'sharc.shef.ac.uk'
# Username for logging in to the cluster via SSH
username = "te1st"

willfurnass

Notes on setting up a Python debug environment

Used to investigate a segfault in a Python C extension within a Python package.

1. Install debug Python build using pyenv then check is available for use:

   pyenv install -g 3.8.5
   pyenv versions
   

willfurnass

import os
from typing import Dict
import numpy as np


def export_data(data: Dict[np.ndarray],
                filename: str,
                cols_per_line: int = 8) -> None:
    """Export dictionary of 1d float arrays to a file.
 

willfurnass

#!/usr/bin/tclsh

set sge_root $env(SGE_ROOT)
source "$sge_root/util/resources/jsv/jsv_include.tcl"

proc jsv_on_start {} {
    jsv_send_env
}

proc lexists name {

willfurnass

Initialise a git repository:

cd somedir
git init

Install git subrepo if you haven't done so already:

https://github.com/ingydotnet/git-subrepo#installation

willfurnass

tmux tips

Default prefix: <ctrl>b

Session management

• tmux new -s session_name - creates a new tmux session named session_name
• tmux a - attach to first available session
• tmux attach -t session_name - attaches to an existing tmux session named session_name
• tmux switch -t session_name - switches to an existing session named session_name

willfurnass

#!/usr/bin/env python

# A test of whether Ruffus can be used to submit jobs to a cluster using the
# DRMAA API, sharing the DRMAA session between threads to allow multiple jobs to
# be submitted at once without burdening the head node with multiple DRMAA
# sessions.

# Can run this from a conda environnment created using
# conda create -n drmaatest -c bioconda python=2.7 ruffus drmaa