Set up a functional cluster for basic operations

- Include cgroup.conf configuration
- Install slurm-client on compute nodes
- Move home dir to /vagrant for result sharing
- Add Makefile for streamlined setup and cleanup
- Fix idempotency issues

.gitignore

@@ -2,3 +2,4 @@ munge.key
 nodes.rb
 .settings.yml
 .vagrant
+vagrantup.log

Makefile

@@ -0,0 +1,10 @@
+.PHONY: all vagrant clean
+
+all: vagrant
+
+vagrant:
+	vagrant up --no-destroy-on-error --no-color | tee ./vagrantup.log
+
+clean:
+	vagrant destroy -f --no-color
+	rm -rf .vagrant vagrantup.log munge.key ./scratch/submit

cgroup.conf

@@ -0,0 +1,10 @@
+###
+# Slurm cgroup support configuration file.
+###
+CgroupAutomount=yes
+CgroupMountpoint=/sys/fs/cgroup
+ConstrainCores=yes
+ConstrainDevices=yes
+ConstrainKmemSpace=no        #avoid known Kernel issues
+ConstrainRAMSpace=yes
+ConstrainSwapSpace=yes

provision.sh

@@ -16,7 +16,7 @@ apt-get install -y chrony
 systemctl start chrony
 systemctl enable chrony
 
-# Create a dedicated non-privileged user account for MUNGE
+# Create MUNGE user
 getent group munge > /dev/null || groupadd -r -g 900 munge
 id -u munge &>/dev/null || \
   useradd -r -u 900 -g munge -d /var/lib/munge -s /usr/sbin/nologin munge
@@ -26,25 +26,34 @@ getent group slurm > /dev/null || groupadd -g 1001 slurm
 id -u slurm &>/dev/null || \
   useradd -m -u 1001 -g slurm -s /bin/bash slurm
 
-# Create job 'submit' user
+# Create submit user
 getent group submit > /dev/null || groupadd -g 1002 submit
 id -u submit &>/dev/null || \
-  useradd -m -u 1002 -g submit -s /bin/bash submit
+  useradd -M -u 1002 -g submit -s /bin/bash -d /vagrant/scratch/submit submit
+mkdir -p /vagrant/scratch/submit
 
-# Install MUNGE, remove any default key, and stop to another place key later
+# Update APT cache
 apt-get update
-apt-get install -y munge
+
-systemctl stop munge
+# Install MUNGE, remove any default key, and stop to place another key later
-rm -f /etc/munge/munge.key
+if ! dpkg -s munge &>/dev/null; then
+  apt-get install -y munge
+  systemctl stop munge
+  rm -f /etc/munge/munge.key
+fi
+
+# Install slurm client tools
+dpkg -s slurm-client &>/dev/null || apt-get install -y slurm-client
 
 # Create directories for Slurm
 mkdir -p /var/spool/slurm /var/log/slurm /etc/slurm
 chown slurm:slurm /var/spool/slurm /var/log/slurm /etc/slurm
 
-# Copy slurm.conf
+# Copy slurm.conf and cgroup.conf
 cp -u /vagrant/slurm.conf /etc/slurm/slurm.conf
-chown slurm:slurm /etc/slurm/slurm.conf
+cp -u /vagrant/cgroup.conf /etc/slurm/cgroup.conf
-chmod 644 /etc/slurm/slurm.conf
+chown slurm:slurm /etc/slurm/slurm.conf /etc/slurm/cgroup.conf
+chmod 644 /etc/slurm/slurm.conf /etc/slurm/cgroup.conf
 
 # node1 = manager
 if [ "$(hostname)" == "node1" ]; then
@@ -53,34 +62,36 @@ if [ "$(hostname)" == "node1" ]; then
     sudo -u munge /usr/sbin/mungekey --verbose
   fi
 
-  # Set MUNGE key perms
-  chmod 600 /etc/munge/munge.key
-
   # Copy to shared directory for other nodes
   cp /etc/munge/munge.key /vagrant/munge.key
 
   # Enable/start/test munge service
-  systemctl enable munge.service
+  chmod 400 /etc/munge/munge.key
-  systemctl start munge.service
+  systemctl enable munge
-  munge -n | unmunge
+  systemctl start munge
 
   # Install Slurm Workload Manager and doc package for the Slurm config tool
-  apt-get install -y slurm-wlm slurm-wlm-doc
+  if ! dpkg -s slurm-wlm &>/dev/null; then
+    apt-get install -y slurm-wlm slurm-wlm-doc
 
-  # Create directories for slurmctld
+    # Create directories for slurmctld
-  mkdir -p /var/spool/slurmctld
+    systemctl stop slurmctld
-  chown slurm:slurm /var/spool/slurmctld
+    mkdir -p /var/spool/slurmctld
+    chown slurm:slurm /var/spool/slurmctld
+    chmod 755 /var/spool/slurmctld
 
-  # Start Slurm controller
+    # Start Slurm controller
-  systemctl enable slurmctld
+    systemctl enable slurmctld
-  systemctl start slurmctld
+    systemctl start slurmctld
+  fi
 else
   # Initial delay
   sleep 5
 
   # Waits JOIN_TIMEOUT of seconds to find the munge.key file before giving up
   START_TIME="$(date +%s)"
-  # Wait until the munge.key can be found via Vagrant provider file sharing /vagrant
+
+  # Wait until the munge.key can be found via Vagrant provider file sharing
   while [ ! -f /vagrant/munge.key ]; do
     CURRENT_TIME="$(date +%s)"
     DIFF_TIME="$((CURRENT_TIME - START_TIME))"
@@ -100,21 +111,13 @@ else
   cp -f /vagrant/munge.key /etc/munge/munge.key
   chown munge:munge /etc/munge/munge.key
   chmod 400 /etc/munge/munge.key
-  systemctl enable munge.service
+  systemctl enable munge
-  systemctl start munge.service
+  systemctl start munge
-  munge -n | unmunge
 
-  # Submit job as 'submit' on node2
+  # Install SLURM compute node daemon on node[3-4]
-  if [ "$(hostname)" == "node2" ]; then
+  if [[ $(hostname) == node[3-4] ]]; then
-    # Install Slurm client tools
+    mkdir -p /var/spool/slurmd
-    apt-get install -y slurm-client
+    chown slurm:slurm /var/spool/slurmd
-
-    # Submit a test job as the 'submit' user
-    sleep 10
-    sudo -u submit bash -c 'sbatch -N2 --wrap="srun hostname"'
-    sudo -u submit squeue
-  else
-    # Install SLURM compute node daemon on node3+
     apt-get install -y slurmd
     systemctl enable slurmd
     systemctl start slurmd

scratch/.gitignore

@@ -0,0 +1,2 @@
+*
+!.gitignore