Some Grid Engine is a fork of Son of Grid Engine at University of Liverpool, with SOME improvement.
We have been using and maintaining this software at Michigan Neuroscience Institute, University of Michigan for over a decade. It is stable and good enough for a small HPC cluster. Here we share it on github, hoping more peoples can benefit from it.
- Support for submitting jobs via SystemD which also allows to enforce memory/cpu limitations via kernel cgroups. Many thanks to fretn:master and ondrejv2:master!
- CMake compiling support. This paved the way for easier maintenance in future. It took 38 seconds to compile in parallel and install on an 8-core old machine, while it took 302 seconds with the legacy SGE way, and 377 seconds with makepkg.
- Fixed a permission error introduced since systemd 241 in 2019 during installation, if SGE is installed as non-root on production system
- Compatible with openssl-1.1.1
- All C compling warning are fixed on Arch Linux and Void Linux. Most of them were caused by 'smarter' gcc, new SSL, new GLIBC, obsolete function such 'sigignore', depreciated function such as 'readdir_r', etc.
- Underscore in port service name 'sge_qmaster/sge_execd' is changed to hyphen in all C files and shell scripts, saving us from modifying /etc/services each time
- Supports both systemd and runit on Void Linux
- Version is changed to the commit version of this github repo
- 5 keystrokes to setup a demo cluster on any Linux machine without root privilege
- Tests for a lot of Linux distros
- Arch Linux, 2022-05-05
pacman -Sy --needed git cmake make gcc openmotif hwloc vi inetutils pkgconf
- Debian Bullseye, 2022-05-05
apt install git build-essential libhwloc-dev libssl-dev libtirpc-dev libmotif-dev libxext-dev libncurses-dev libdb5.3-dev libpam0g-dev pkgconf libsystemd-dev cmake
- Debian Buster, 2021-10-19, cmake 3.21.3 downloaded from cmake.org
apt install git build-essential libhwloc-dev libssl-dev libtirpc-dev libmotif-dev libxext-dev libncurses-dev libdb5.3-dev libpam0g-dev pkgconf libsystemd-dev
- Ubuntu Server 22.04, 2022-05-05
apt install git build-essential libhwloc-dev libssl-dev libtirpc-dev libmotif-dev libxext-dev libncurses-dev libdb5.3-dev libpam0g-dev pkgconf libsystemd-dev cmake
- Ubuntu Server 20.04, 2022-05-05
apt install git build-essential libhwloc-dev libssl-dev libtirpc-dev libmotif-dev libxext-dev libncurses-dev libdb5.3-dev libpam0g-dev pkgconf libsystemd-dev cmake
- Void Linux, 2022-05-05, x86_64, Glibc
xbps-install cmake make gcc openssl-devel motif-devel hwloc libhwloc-devel libtirpc-devel ncurses-devel pam-devel
- CentOS 7.9.2009, 2022-05-05, with SELinux set to permissive
yum groupinstall 'Development Tools'
yum install git hwloc-devel openssl-devel libtirpc-devel motif-devel ncurses-devel libdb-devel pam-devel systemd-devel wget
wget https://github.com/Kitware/CMake/releases/download/v3.23.0/cmake-3.23.0-linux-x86_64.tar.gz
tar xvfz cmake-3.23.0-linux-x86_64.tar.gz
export PATH=$(realpath cmake-3.23.0-linux-x86_64)/bin:$PATH
- Rocky 8.5 and AlmaLinux 8.5, 2022-05-05, with SELinux set to permissive
dnf group install "Development Tools"
dnf --enablerepo=powertools install git hwloc-devel openssl-devel libtirpc-devel motif-devel ncurses-devel libdb-devel pam-devel cmake systemd-devel pkgconf
- openSUSE Leap, 2022-05-05
zypper -n addrepo http://download.opensuse.org/distribution/leap/15.3/repo/oss/ oss
zypper -n install cmake gcc gcc-c++ git hwloc-devel libdb-4_8-devel libtirpc-devel libXext-devel motif-devel ncurses-devel openssl-devel pam-devel pkgconf rsync systemd-devel wget
- openSUSE Tumbleweed, 2022-05-05
zypper -n addrepo http://download.opensuse.org/tumbleweed/repo/oss/ oss
zypper -n install cmake gcc gcc-c++ git hwloc-devel libdb-4_8-devel libtirpc-devel libXext-devel motif-devel ncurses-devel openssl-devel pam-devel pkgconf rsync systemd-devel wget
- CMake, recommended
cmake -S . -B build -DCMAKE_INSTALL_PREFIX=/opt/sge -DSYSTEMD=OFF #or ON if it is not Void Linux or CentOS 7
cmake --build build -j
sudo cmake --install build
Please check the tested environment below in case of any compiling issue.
- Legacy SGE installation on modern Linux distributions
make
sudo make install
Please check the original source/README.BUILD for detail
- Legacy SGE installation on Arch Linux
cp PKGBUILD.in PKGBUILD
makepkg
sudo pacman -U sge-r*.pkg.tar.zst
- step 1, as root.
First of all, change option admin_user in bootstrapfile
useradd -r -d /opt/sge sge
chown -R sge /opt/sge
cd /opt/sge
yes "" | ./install_qmaster
yes "" | ./install_execd
source /opt/sge/default/common/settings.sh
qhost -q #you should be able to see five lines of output
qconf -as $HOSTNAME #add this node as submit host
- step 2, as a regular account
source /opt/sge/default/common/settings.sh
echo hostname | qsub -cwd
watch qstat #check job status
ls STDIN.* #check job output
All SGE services are running under user sge for security reason, as this is production system.
Assuming master node hostname is master, and execution nodes hostnames is node-XX. /etc/hosts on master and all nodes shoud be like it:
10.1.1.1 master
10.1.1.11 node-01
10.1.1.12 node-02
...
10.1.1.1N node-0N
All IP addresses are as an example.
ping master
ping node-XX
useradd -u <UID> -r -d /opt/sge sge
sge UID should be equal on all machines.
First of all, change option admin_user in bootstrapfile
chown -R sge /opt/sge
cd /opt/sge
yes "" | ./install_qmaster
source /opt/sge/default/common/settings.sh
qconf -as master
Next commands it's necessary to run for each nodes in cluster:
qconf -ah node-01
qconf -ah node-02
...
qconf -ah node-0N
mkdir -p /opt/sge/default
chown -R sge /opt/sge/default
scp -pr master:/opt/sge/default/common /opt/sge/default/common
cd /opt/sge
yes "" | ./install_execd
source /opt/sge/default/common/settings.sh
qhost -q
source /opt/sge/default/common/settings.sh
echo hostname | qsub -cwd
watch qstat
cat STDIN.*
Contributions are always welcome!
Written by Manhong Dai Copyright © 2002-2022 University of Michigan. License SISSL
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Thomas Wilson, M.D., Ph.D. Professor of Pathology, UMICH
Ruth Freedman, MPH, former administrator of MNI, UMICH
Fan Meng, Ph.D., Research Associate Professor, Psychiatry, UMICH
Huda Akil, Ph.D., Director of MNI, UMICH
Stanley J. Watson, M.D., Ph.D., Director of MNI, UMICH
Also thanks to https://arc.liv.ac.uk/trac/SGE, Sun, and Oracle