diff --git a/roles/apt-get-upgrade/tasks/main.yml b/roles/apt-get-upgrade/tasks/main.yml
new file mode 100644
index 0000000000000000000000000000000000000000..49e0217a1a170d17bee5ef958f7886903e0c1256
--- /dev/null
+++ b/roles/apt-get-upgrade/tasks/main.yml
@@ -0,0 +1,10 @@
+---
+- name: apt-get update
+ apt: update_cache=True
+ sudo: true
+ when: ansible_os_family=="Debian"
+
+- name: apt-get upgrade
+ apt: upgrade=safe
+ sudo: true
+ when: ansible_os_family=="Debian"
diff --git a/roles/nfs-client/tasks/mountFileSystem.yml b/roles/nfs-client/tasks/mountFileSystem.yml
index 555e12b5014dd5c37617903c589a60d090d73d70..0dc48290c5dfa850ef847654c5d9c4708ce61e2c 100644
--- a/roles/nfs-client/tasks/mountFileSystem.yml
+++ b/roles/nfs-client/tasks/mountFileSystem.yml
@@ -4,16 +4,18 @@
sudo: true
+
+
- name: "Mounting NFS mounts"
- mount: name={{ item.name }} src={{ item.ipv4 }}:{{ item.src }} fstype={{ item.fstype }} opts={{ item.opts }} state=mounted
- with_items: exportList
+ mount: name={{ item.mntpt }} src={{ item.src }} fstype={{ item.fstype }} opts={{ item.opts }} state=mounted
+ with_items: nfsMounts
notify: "restart authentication"
notify: "restart rpcbind"
notify: "restart idmap"
sudo: true
ignore_errors: true
register: firstMount
- when: exportList is defined
+ when: nfsMounts is defined
- name: "Wait for nfs to stabailse"
command: sleep 60
@@ -21,13 +23,13 @@
when: firstMount | failed
- name: "Mounting NFS mounts"
- mount: name={{ item.name }} src={{ item.ipv4 }}:{{ item.src }} fstype={{ item.fstype }} opts={{ item.opts }} state=mounted
- with_items: exportList
+ mount: name={{ item.name }} src=" {{ item.ipv4 }}:{{ item.src }} " fstype={{ item.fstype }} opts={{ item.opts }} state=mounted
+ with_items: nfsMounts
notify: "restart authentication"
notify: "restart idmap"
notify: "restart rpcbind"
sudo: true
- when: exportList is defined and firstMount | failed
+ when: nfsMounts is defined and firstMount | failed
- name: "restart fail2ban"
service: name=fail2ban state=started
diff --git a/roles/nfs-common/tasks/aptPackages.yml b/roles/nfs-common/tasks/aptPackages.yml
new file mode 100644
index 0000000000000000000000000000000000000000..364b2b73093be8d9ac7b80c63bd53025d3694713
--- /dev/null
+++ b/roles/nfs-common/tasks/aptPackages.yml
@@ -0,0 +1,9 @@
+---
+-
+ name: "Install nfs-utils"
+ with_items:
+ - nfs-common
+ - nfs-kernel-server
+ apt: "name={{ item }} state=latest"
+ sudo: true
+
diff --git a/roles/nfs-server/tasks/startServer.yml b/roles/nfs-server/tasks/startServer.yml
index e8338d56c7265b4fb04a9f5626198d1f3bfe6846..606b143bda4081e031d4944f63125f0d8036fae2 100644
--- a/roles/nfs-server/tasks/startServer.yml
+++ b/roles/nfs-server/tasks/startServer.yml
@@ -2,6 +2,14 @@
- name: "Starting rpcbind"
service: "name=rpcbind state=restarted"
sudo: true
+ when: ansible_os_family == "RedHat"
+
- name: "Start the Server"
service: "name=nfs state=restarted"
sudo: true
+ when: ansible_os_family == "RedHat"
+
+- name: "Start the Server"
+ service: "name=nfs-kernel-server state=restarted"
+ sudo: true
+ when: ansible_os_family == "Debian"
diff --git a/roles/slurm-from-source/defaults/main.yml b/roles/slurm-from-source/defaults/main.yml
new file mode 100644
index 0000000000000000000000000000000000000000..dc0836cd6b6ca61a8223423f7d4f8072e185abdd
--- /dev/null
+++ b/roles/slurm-from-source/defaults/main.yml
@@ -0,0 +1,14 @@
+---
+munge_dir: /opt/munge
+slurm_dir: /opt/slurm
+slurm_use_vpn: False
+slurmctlddebug: {level: 9, log: '/var/log/slurm/slurmctld.log'}
+slurmddebug: {level: 9, log: '/var/log/slurm/slurmd.log'}
+slurmschedlog: {level: 9, log: '/var/log/slurm/slurmsched.log'}
+slurmdbdlog: {level: 9, log: '/var/log/slurm/slurmdbd.log'}
+slurmfairshare: {def: false, val: 10000}
+slurmdatadir: "/var/spool/slurm"
+slurmselecttype: "select/linear"
+slurmfastschedule: "1"
+slurmschedulertype: "sched/backfill"
+
diff --git a/roles/slurm-from-source/handlers/main.yml b/roles/slurm-from-source/handlers/main.yml
new file mode 100644
index 0000000000000000000000000000000000000000..57f0cb12be532609f9884170cf7aeadf7be037f9
--- /dev/null
+++ b/roles/slurm-from-source/handlers/main.yml
@@ -0,0 +1,18 @@
+---
+ - name: restart munge
+ service: name=munge state=restarted
+ sudo: true
+
+ - name: restart slurm
+ service: name=slurm state=restarted
+ sudo: true
+
+ - name: restart slurmdbd
+ service: name=slurmdbd state=restarted
+ sudo: true
+
+ - name: scontrol reconfigure
+ shell: sleep 10 ; scontrol reconfigure
+ sudo: true
+ delegate_to: "{{ slurmctrl }}"
+ run_once: true
diff --git a/roles/slurm-from-source/tasks/installMungeFromSource.yml b/roles/slurm-from-source/tasks/installMungeFromSource.yml
new file mode 100644
index 0000000000000000000000000000000000000000..62a716e3eda3d4f1ae5f56d66a468ab12d19a726
--- /dev/null
+++ b/roles/slurm-from-source/tasks/installMungeFromSource.yml
@@ -0,0 +1,34 @@
+- name: get munge
+ shell: wget https://munge.googlecode.com/files/munge-{{ munge_version }}.tar.bz2
+ args:
+ chdir: /tmp
+ creates: /tmp/munge-{{ munge_version }}.tar.bz2
+
+- name: remove old
+ shell: rm -rf /tmp/munge-{{ munge_version }} && rm -rf /tmp/slurm-{{ slurm_version }}
+ sudo: true
+
+- name: untar munge
+ shell: tar jxf munge-{{ munge_version }}.tar.bz2
+ args:
+ chdir: /tmp
+
+- name: build munge
+ shell: ./configure --prefix={{ munge_dir }} && make
+ args:
+ chdir: /tmp/munge-{{ munge_version }}
+
+- name: install munge
+ shell: make install
+ sudo: true
+ args:
+ chdir: /tmp/munge-{{ munge_version }}
+
+- name: copy init script
+ template: dest=/etc/init.d/munge src=munge.initd.j2 mode=755
+ sudo: true
+
+- name: start on boot
+ shell: update-rc.d munge defaults
+ sudo: true
+ when: ansible_distribution == "Ubuntu"
diff --git a/roles/slurm-from-source/tasks/installSlurmFromSource.yml b/roles/slurm-from-source/tasks/installSlurmFromSource.yml
new file mode 100644
index 0000000000000000000000000000000000000000..945bf80409b908e1ec337f24e71ff5cbeb493330
--- /dev/null
+++ b/roles/slurm-from-source/tasks/installSlurmFromSource.yml
@@ -0,0 +1,26 @@
+- name: get slurm
+ shell: wget http://www.schedmd.com/download/archive/slurm-{{ slurm_version }}.tar.bz2
+ args:
+ chdir: /tmp
+ creates: /tmp/slurm-{{ slurm_version }}.tar.bz2
+
+- name: untar slurm
+ shell: tar jxf /tmp/slurm-{{ slurm_version }}.tar.bz2
+ args:
+ chdir: /tmp
+
+- name: build slurm
+ shell: ./configure --prefix={{ slurm_dir }} --with-munge={{ munge_dir }} && make
+ args:
+ chdir: /tmp/slurm-{{ slurm_version }}
+
+- name: install slurm
+ shell: make install
+ sudo: true
+ args:
+ chdir: /tmp/slurm-{{ slurm_version }}
+
+- name: copy init script
+ template: dest=/etc/init.d/slurm src=slurm.initd.j2 mode=755
+ sudo: true
+
diff --git a/roles/slurm-from-source/tasks/main.yml b/roles/slurm-from-source/tasks/main.yml
new file mode 100644
index 0000000000000000000000000000000000000000..1fa1d007c4b9177cbdce067781edae6ce3a70f2c
--- /dev/null
+++ b/roles/slurm-from-source/tasks/main.yml
@@ -0,0 +1,84 @@
+---
+- name: create munge group
+ group: name=munge system=yes gid=498
+ sudo: true
+
+- name: create munge user
+ user: name=munge group=munge system=yes createhome=no uid=498
+ sudo: true
+
+- name: create slurm group
+ group: name=slurm system=yes gid=497
+ sudo: true
+
+- name: create slurm user
+ user: name=slurm group=slurm system=yes createhome=no uid=497
+ sudo: true
+
+- name: make sure slurm conf dir exists
+ file: dest={{ slurm_dir }}/etc state=directory
+ sudo: true
+
+- name: make sure slurm lock dir exists
+ file: dest=/var/lock/subsys state=directory owner=root group=root mode=755
+ sudo: true
+
+- name: create data directory
+ file: path={{ slurmdatadir }} state=directory owner=slurm group=slurm mode=755
+ sudo: true
+
+- name: create log directory
+ shell: mkdir -p {{ slurmddebug.log | dirname }}; chown slurm:slurm {{ slurmddebug.log | dirname }}
+ args:
+ creates: "{{ slurmddebug.log | dirname }}"
+ sudo: true
+ when: slurmddebug is defined and slurmddebug.log
+
+- name: install deps
+ yum: name={{ item }} state=latest
+ with_items:
+ - perl
+ - perl-DBI
+ sudo: true
+ when: ansible_os_family == "RedHat"
+
+- name: install deps
+ apt: name={{ item }} state=installed update_cache=yes
+ sudo: true
+ with_items:
+ - gcc
+ - wget
+ - libssl-dev
+ - libpam0g-dev
+ - libbz2-dev
+ - make
+ - perl
+ - libdbi-perl
+ when: ansible_os_family == "Debian"
+
+- include: installMungeFromSource.yml
+
+- name: chown mungedir
+ file: path={{ munge_dir }} state=directory owner=munge recurse=yes
+ sudo: true
+
+- name: make munge logdir
+ file: path={{ munge_dir }}/var/log/munge state=directory owner=munge mode=700
+ sudo: true
+
+- name: install munge key
+ template: src=munge_key.j2 dest={{ munge_dir }}/etc/munge/munge.key owner=munge mode=600
+ sudo: true
+ notify: restart munge
+
+- include: installSlurmFromSource.yml
+
+- name: install slurm.conf
+ template: src=slurm.conf.j2 dest={{ slurm_dir }}/etc/slurm.conf
+ sudo: true
+ when: slurm_use_vpn==False
+
+- name: install slurm.conf
+ template: src=slurm-vpn.conf.j2 dest={{ slurm_dir }}/etc/slurm.conf
+ sudo: true
+ when: slurm_use_vpn==True
diff --git a/roles/slurm-from-source/templates/munge.initd.j2 b/roles/slurm-from-source/templates/munge.initd.j2
new file mode 100755
index 0000000000000000000000000000000000000000..5443aba6b2405bc65ace2895d87c38100b79b011
--- /dev/null
+++ b/roles/slurm-from-source/templates/munge.initd.j2
@@ -0,0 +1,567 @@
+#!/bin/sh
+###############################################################################
+# Written by Chris Dunlap <cdunlap@llnl.gov>.
+# Copyright (C) 2007-2013 Lawrence Livermore National Security, LLC.
+# Copyright (C) 2002-2007 The Regents of the University of California.
+# UCRL-CODE-155910.
+###############################################################################
+# chkconfig: - 66 33
+# description: MUNGE Uid 'N' Gid Emporium authentication service
+###############################################################################
+### BEGIN INIT INFO
+# Provides: munge
+# Required-Start: $local_fs $remote_fs $network $time
+# Required-Stop: $local_fs $remote_fs
+# Should-Start: $named $syslog
+# Should-Stop: $named $syslog
+# Default-Start:
+# Default-Stop:
+# Short-Description: MUNGE Uid 'N' Gid Emporium authentication service
+# Description: MUNGE (MUNGE Uid 'N' Gid Emporium) is a highly scalable
+# authentication service for creating and validating
+# credentials.
+### END INIT INFO
+###############################################################################
+
+unset SERVICE_NAME DAEMON_EXEC DAEMON_ARGS CONFIG PIDFILE NICE USER GROUP \
+ SIGHUP_RELOAD VARRUNDIR
+
+prefix="/opt/munge"
+exec_prefix="${prefix}"
+sbindir="${exec_prefix}/sbin"
+sysconfdir="${prefix}/etc"
+localstatedir="${prefix}/var"
+
+SERVICE_NAME="MUNGE"
+DAEMON_EXEC="$sbindir/munged"
+#DAEMON_ARGS=
+#CONFIG=#_NOT_SUPPORTED_#
+PIDFILE="$localstatedir/run/munge/munged.pid"
+#NICE=
+USER="munge"
+GROUP="munge"
+#SIGHUP_RELOAD=#_NOT_SUPPORTED_#
+VARRUNDIR="$localstatedir/run/munge"
+
+###############################################################################
+
+service_init ()
+{
+# Determine the system type and initialize the environment.
+#
+# Note that the shell positional parameters must be preserved when calling
+# this function in order for SuSE to initialize its environment properly.
+##
+ PATH=/sbin:/usr/sbin:/bin:/usr/bin
+ INIT_NAME="`basename \"$0\" .init | sed 's/^[SK][0-9][0-9]*//'`"
+ DAEMON_NAME="`basename \"$DAEMON_EXEC\"`"
+ SIGTERM_TIMEOUT="3"
+ STATUS=0
+
+ # Read configuration defaults to override variables:
+ # $CONFIG, $DAEMON_ARGS, $PIDFILE, $USER, $NICE, $SIGHUP_RELOAD
+ ##
+ for dir in "$sysconfdir/default" "$sysconfdir/sysconfig"; do
+ [ -r "$dir/$INIT_NAME" ] && . "$dir/$INIT_NAME"
+ done
+ [ -z "$DAEMON_ARGS" -a -n "$OPTIONS" ] && DAEMON_ARGS="$OPTIONS"
+ [ "`id | sed 's/^uid=\([0-9]*\).*/\1/'`" -ne 0 ] && unset USER
+ expr -- "$NICE" : '[0-9]*$' >/dev/null 2>&1 && NICE="+$NICE"
+ [ -n "$SIGHUP_RELOAD" -a "$SIGHUP_RELOAD" != 0 ] \
+ && RELOAD=1 || unset RELOAD
+
+ if [ -f /etc/debian_version -a -x /sbin/start-stop-daemon ]; then
+ SYSTEM="DEBIAN"
+ [ -x "$DAEMON_EXEC" ] || exit 0 # pkg removed but not purged
+ [ -r /etc/default/rcS ] && . /etc/default/rcS
+ [ -r /lib/init/vars.sh ] && . /lib/init/vars.sh
+ [ -r /lib/lsb/init-functions ] && . /lib/lsb/init-functions
+ elif [ -f /etc/redhat-release -a -r /etc/rc.d/init.d/functions ]; then
+ SYSTEM="REDHAT"
+ . /etc/rc.d/init.d/functions
+ RH_LOCK="/var/lock/subsys/$INIT_NAME"
+ elif [ -f /etc/SuSE-release -a -r /etc/rc.status ]; then
+ SYSTEM="SUSE"
+ . /etc/rc.status
+ rc_reset
+ elif [ -r /lib/lsb/init-functions ]; then
+ SYSTEM="LSB"
+ . /lib/lsb/init-functions
+ else
+ SYSTEM="OTHER"
+ fi
+
+ # Exit if the package has been removed.
+ ##
+ [ -x "$DAEMON_EXEC" ] || exit 5 # LSB: program not installed
+
+ # Exit if the configuration has been removed.
+ ##
+ [ -z "$CONFIG" -o -r "$CONFIG" ] || exit 6 # LSB: program not configured
+}
+
+service_fini ()
+{
+# Return the exit status.
+##
+ case $SYSTEM in
+ SUSE)
+ rc_exit
+ ;;
+ DEBIAN|REDHAT|LSB|*)
+ exit $STATUS
+ ;;
+ esac
+}
+
+service_start ()
+{
+# Start the service.
+#
+# Required by LSB, where running "start" on a service already running should be
+# considered successful.
+##
+ log_init "Starting $SERVICE_NAME" "$DAEMON_NAME"
+
+ if [ -n "$VARRUNDIR" -a ! -d "$VARRUNDIR" ]; then
+ mkdir -m 755 -p "$VARRUNDIR"
+ [ -n "$USER" ] && chown "$USER" "$VARRUNDIR"
+ [ -n "$GROUP" ] && chgrp "$GROUP" "$VARRUNDIR"
+ fi
+
+ case $SYSTEM in
+ DEBIAN)
+ if $0 status >/dev/null 2>&1; then
+ STATUS=0
+ else
+ ERRMSG=`start-stop-daemon --start --quiet \
+ ${NICE:+"--nicelevel"} ${NICE:+"$NICE"} \
+ ${USER:+"--chuid"} ${USER:+"$USER"} \
+ ${PIDFILE:+"--pidfile"} ${PIDFILE:+"$PIDFILE"} \
+ --exec "$DAEMON_EXEC" -- $DAEMON_ARGS 2>&1`
+ STATUS=$?
+ echo $ERRMSG
+ fi
+ ;;
+ REDHAT)
+ if $0 status >/dev/null 2>&1; then
+ STATUS=0
+ else
+ daemon ${NICE:+"$NICE"} ${USER:+"--user"} ${USER:+"$USER"} \
+ "$DAEMON_EXEC" $DAEMON_ARGS
+ STATUS=$?
+ fi
+ [ $STATUS -eq 0 ] && touch "$RH_LOCK" >/dev/null 2>&1
+ ;;
+ SUSE)
+ ERRMSG=`startproc ${NICE:+"-n"} ${NICE:+"$NICE"} \
+ ${USER:+"-u"} ${USER:+"$USER"} \
+ ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} \
+ "$DAEMON_EXEC" $DAEMON_ARGS 2>&1`
+ rc_status -v
+ STATUS=$?
+ ;;
+ LSB)
+ if [ -n "$USER" ]; then
+ ERRMSG=`su "$USER" -c "/sbin/start_daemon \
+ ${NICE:+\"-n\"} ${NICE:+\"$NICE\"} \
+ ${PIDFILE:+\"-p\"} ${PIDFILE:+\"$PIDFILE\"} \
+ \"$DAEMON_EXEC\" $DAEMON_ARGS" 2>&1`
+ else
+ ERRMSG=`start_daemon ${NICE:+"-n"} ${NICE:+"$NICE"} \
+ ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} \
+ "$DAEMON_EXEC" $DAEMON_ARGS 2>&1`
+ fi
+ STATUS=$?
+ ;;
+ *)
+ if $0 status >/dev/null 2>&1; then
+ STATUS=0
+ else
+ [ -n "$NICE" ] && nice="nice -n $NICE"
+ if [ -n "$USER" ]; then
+ ERRMSG=`su "$USER" -c "$nice \"$DAEMON_EXEC\" $DAEMON_ARGS" 2>&1`
+ else
+ ERRMSG=`$nice "$DAEMON_EXEC" $DAEMON_ARGS 2>&1`
+ fi
+ STATUS=$?
+ fi
+ ;;
+ esac
+ log_fini "$STATUS" "$ERRMSG"
+}
+
+service_stop ()
+{
+# Stop the service.
+#
+# Required by LSB, where running "stop" on a service already stopped or not
+# running should be considered successful.
+##
+ log_init "Stopping $SERVICE_NAME" "$DAEMON_NAME"
+ case $SYSTEM in
+ DEBIAN)
+ if ! $0 status >/dev/null 2>&1; then
+ STATUS=0
+ else
+ start-stop-daemon --stop --quiet \
+ ${PIDFILE:+"--pidfile"} ${PIDFILE:+"$PIDFILE"} \
+ --name "$DAEMON_NAME" ${SIGTERM_TIMEOUT:+"--retry"} \
+ ${SIGTERM_TIMEOUT:+"$SIGTERM_TIMEOUT"} >/dev/null 2>&1
+ STATUS=$?
+ fi
+ ;;
+ REDHAT)
+ if ! $0 status >/dev/null 2>&1; then
+ STATUS=0
+ else
+ killproc ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} \
+ ${SIGTERM_TIMEOUT:+"-d"} ${SIGTERM_TIMEOUT:+"$SIGTERM_TIMEOUT"} \
+ "$DAEMON_EXEC"
+ STATUS=$?
+ fi
+ [ $STATUS -eq 0 ] && rm -f "$RH_LOCK" >/dev/null 2>&1
+ ;;
+ SUSE)
+ killproc ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} \
+ ${SIGTERM_TIMEOUT:+"-t"} ${SIGTERM_TIMEOUT:+"$SIGTERM_TIMEOUT"} \
+ "$DAEMON_EXEC"
+ rc_status -v
+ ;;
+ LSB)
+ killproc ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} "$DAEMON_EXEC"
+ STATUS=$?
+ ;;
+ *)
+ signal_process "$DAEMON_EXEC"
+ rc=$?
+ [ $rc -eq 0 -o $rc -eq 2 ] && STATUS=0 || STATUS=1
+ ;;
+ esac
+ log_fini "$STATUS"
+ [ -f "$PIDFILE" ] && rm -f "$PIDFILE"
+}
+
+service_restart ()
+{
+# Stop and restart the service if it is already running;
+# otherwise, start the service.
+#
+# Required by LSB, where running "restart" on a service already stopped or not
+# running should be considered successful.
+##
+ if $0 status >/dev/null 2>&1; then
+ $0 stop && $0 start
+ else
+ $0 start
+ fi
+
+ case $SYSTEM in
+ SUSE)
+ rc_status
+ ;;
+ DEBIAN|REDHAT|LSB|*)
+ STATUS=$?
+ ;;
+ esac
+}
+
+service_try_restart ()
+{
+# Restart the service if it is already running.
+#
+# Optional for LSB, where running "try-restart" on a service already stopped or
+# not running should be considered successful.
+# Also known as "condrestart" by RedHat.
+##
+ case $SYSTEM in
+ REDHAT)
+ [ -f "$RH_LOCK" ] && $0 restart || :
+ STATUS=$?
+ ;;
+ SUSE)
+ $0 status >/dev/null 2>&1 && $0 restart || rc_reset
+ rc_status
+ ;;
+ DEBIAN|LSB|*)
+ $0 status >/dev/null 2>&1 && $0 restart || :
+ STATUS=$?
+ ;;
+ esac
+}
+
+service_reload ()
+{
+# Reload the configuration without stopping and restarting the service.
+#
+# Optional for LSB.
+##
+ [ -z "$RELOAD" ] && STATUS=3 # LSB: unimplemented feature
+
+ log_init "Reloading $SERVICE_NAME" "$DAEMON_NAME"
+ case $SYSTEM in
+ DEBIAN)
+ if [ -n "$RELOAD" ]; then
+ start-stop-daemon --stop --quiet --signal HUP \
+ ${PIDFILE:+"--pidfile"} ${PIDFILE:+"$PIDFILE"} \
+ --name "$DAEMON_NAME" >/dev/null 2>&1
+ STATUS=$?
+ fi
+ ;;
+ REDHAT)
+ if [ -n "$RELOAD" ]; then
+ killproc ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} "$DAEMON_EXEC" -HUP
+ STATUS=$?
+ else
+ echo_failure
+ fi
+ ;;
+ SUSE)
+ if [ -n "$RELOAD" ]; then
+ killproc -HUP ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} "$DAEMON_EXEC"
+ else
+ rc_failed $STATUS
+ fi
+ rc_status -v
+ ;;
+ LSB)
+ if [ -n "$RELOAD" ]; then
+ killproc ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} "$DAEMON_EXEC" -HUP
+ STATUS=$?
+ fi
+ ;;
+ *)
+ if [ -n "$RELOAD" ]; then
+ signal_process "$DAEMON_EXEC" "HUP"
+ STATUS=$?
+ fi
+ ;;
+ esac
+ log_fini "$STATUS"
+}
+
+service_force_reload ()
+{
+# Reload the configuration if the service supports this;
+# otherwise, restart the service if it is already running.
+#
+# Required by LSB, where running "force-reload" on a service already stopped or
+# not running should be considered successful.
+##
+ if [ -n "$RELOAD" ]; then
+ $0 reload
+ else
+ $0 try-restart
+ fi
+
+ case $SYSTEM in
+ SUSE)
+ rc_status
+ ;;
+ DEBIAN|REDHAT|LSB|*)
+ STATUS=$?
+ ;;
+ esac
+}
+
+service_status ()
+{
+# Print the current status of the service.
+#
+# Required by LSB.
+##
+ case $SYSTEM in
+ REDHAT)
+ status ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} "$DAEMON_EXEC"
+ STATUS=$?
+ ;;
+ SUSE)
+ printf "Checking for service $SERVICE_NAME: "
+ checkproc ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} "$DAEMON_EXEC"
+ rc_status -v
+ ;;
+ LSB)
+ printf "Checking status of $SERVICE_NAME: "
+ pids=`pidofproc ${PIDFILE:+"-p"} ${PIDFILE:+"$PIDFILE"} \
+ "$DAEMON_EXEC" 2>/dev/null`
+ STATUS=$?
+ if [ $STATUS -eq 0 -a -n "$pids" ]; then
+ echo "running."
+ elif [ $STATUS -ne 0 -a -s "$PIDFILE" ]; then
+ echo "dead."
+ else
+ echo "stopped."
+ fi
+ ;;
+ DEBIAN|*)
+ printf "Checking status of $SERVICE_NAME: "
+ pids=`query_pids "$DAEMON_EXEC" "$PIDFILE"`
+ rc=$?
+ if [ $rc -eq 0 -a -n "$pids" ]; then
+ echo "running."
+ STATUS=0 # LSB: program is running
+ elif [ $rc -ne 0 -a -s "$PIDFILE" ]; then
+ echo "dead."
+ STATUS=1 # LSB: program is dead & pidfile exists
+ elif [ $rc -ne 0 ]; then
+ echo "stopped."
+ STATUS=3 # LSB: program is not running
+ else
+ echo "unknown."
+ STATUS=4 # LSB: program status unknown
+ fi
+ ;;
+ esac
+}
+
+query_pids ()
+{
+# Writes the matching PIDs to stdout.
+# Returns 0 on success (ie, pids found).
+##
+ PROCNAME="$1"
+ PIDFILE="$2"
+
+ if type pgrep >/dev/null 2>&1; then
+ pids=`pgrep -d ' ' -x "\`basename \"$PROCNAME\"\`" 2>/dev/null`
+ rc=$?
+ elif type pidof >/dev/null 2>&1; then
+ pids=`pidof -o $$ -x "$PROCNAME" 2>/dev/null`
+ rc=$?
+ else
+ pids=`(ps awx -o pid -o command || ps -e -f -o pid -o args) 2>/dev/null \
+ | tail +2 | egrep "( |/)$PROCNAME( |$)" | grep -v egrep \
+ | sed 's/ *\([0-9]*\).*/\1/' | sort -n | tr '\012' ' '`
+ [ -n "$pids" ] && rc=0 || rc=1
+ fi
+
+ unset pids_running
+ if [ -n "$pids" -a -r "$PIDFILE" ]; then
+ read pid_line < "$PIDFILE"
+ for pid in $pid_line; do
+ expr -- "$pid" : '[0-9]*$' >/dev/null 2>&1 \
+ && expr -- " $pids " : ".* $pid .*" >/dev/null 2>&1 \
+ && pids_running="$pids_running $pid"
+ done
+ [ -n "$pids_running" ] && pids=$pids_running
+ fi
+
+ echo $pids
+ return $rc
+}
+
+signal_process ()
+{
+# Returns 0 on success, 1 if kill failed, 2 if PROCNAME is not running.
+##
+ PROCNAME="$1"
+ SIGNUM="$2"
+
+ pids=`query_pids "$DAEMON_EXEC" "$PIDFILE"`
+ [ $? -ne 0 -o -z "$pids" ] && return 2
+
+ kill ${SIGNUM:+"-$SIGNUM"} $pids >/dev/null 2>&1
+ [ $? -ne 0 ] && return 1
+ [ -n "$SIGNUM" ] && return 0
+
+ sleep 1
+ pids=`query_pids "$DAEMON_EXEC" "$PIDFILE"`
+ [ $? -ne 0 -o -z "$pids" ] && return 0
+ [ -z "$SIGTERM_TIMEOUT" ] && return 1
+
+ sleep "$SIGTERM_TIMEOUT"
+ kill -KILL $pids >/dev/null 2>&1
+ pids=`query_pids "$DAEMON_EXEC" "$PIDFILE"`
+ [ $? -ne 0 -o -z "$pids" ] && return 0
+ return 1
+}
+
+log_init ()
+{
+# Output informational message at beginning of action.
+##
+ MESSAGE="$1"
+ PROCNAME="$2"
+
+ case $SYSTEM in
+ DEBIAN)
+ if [ "$VERBOSE" != no ]; then
+ if type log_daemon_msg >/dev/null 2>&1; then
+ log_daemon_msg "$MESSAGE" "$PROCNAME"
+ else
+ printf "$MESSAGE: $PROCNAME"
+ fi
+ fi
+ ;;
+ REDHAT|SUSE|LSB|*)
+ printf "$MESSAGE: $PROCNAME"
+ ;;
+ esac
+}
+
+log_fini ()
+{
+# Output informational/error message at end of action.
+##
+ STATUS="$1"
+ ERRMSG="$2"
+
+ case $SYSTEM in
+ DEBIAN)
+ if [ "$VERBOSE" != no ]; then
+ if ( type log_end_msg && type log_failure_msg ) >/dev/null 2>&1; then
+ log_end_msg "$STATUS"
+ [ $STATUS -eq 0 -o -z "$ERRMSG" ] || log_failure_msg "$ERRMSG"
+ else
+ [ $STATUS -eq 0 ] && echo "." || echo " (failed)."
+ [ $STATUS -eq 0 -o -z "$ERRMSG" ] || echo "$ERRMSG" >&2
+ fi
+ fi
+ ;;
+ REDHAT)
+ echo
+ ;;
+ SUSE)
+ [ $STATUS -eq 0 -o -z "$ERRMSG" ] || echo "$ERRMSG" >&2
+ ;;
+ LSB|*)
+ [ $STATUS -eq 0 ] && echo "." || echo " (failed)."
+ [ $STATUS -eq 0 -o -z "$ERRMSG" ] || echo "$ERRMSG" >&2
+ ;;
+ esac
+}
+
+###############################################################################
+
+service_init "$@"
+
+case "$1" in
+ start)
+ service_start
+ ;;
+ stop)
+ service_stop
+ ;;
+ restart)
+ service_restart
+ ;;
+ try-restart|condrestart)
+ service_try_restart
+ ;;
+ reload)
+ service_reload
+ ;;
+ force-reload)
+ service_force_reload
+ ;;
+ status)
+ service_status
+ ;;
+ *)
+ echo "Usage: `basename \"$0\"`" \
+ "(start|stop|restart|try-restart|reload|force-reload|status)" >&2
+ exit 2 # LSB: invalid or excess argument(s)
+ ;;
+esac
+
+service_fini
diff --git a/roles/slurm-from-source/templates/munge_key.j2 b/roles/slurm-from-source/templates/munge_key.j2
new file mode 100644
index 0000000000000000000000000000000000000000..83d3483ee198fffce76dd82dee5cbe1fb8c0ab8f
--- /dev/null
+++ b/roles/slurm-from-source/templates/munge_key.j2
@@ -0,0 +1 @@
+{{ mungekey }}
diff --git a/roles/slurm-from-source/templates/slurm-vpn.conf.j2 b/roles/slurm-from-source/templates/slurm-vpn.conf.j2
new file mode 100644
index 0000000000000000000000000000000000000000..671840e07831bac8de9b440c3c9e90ca65ff3fd3
--- /dev/null
+++ b/roles/slurm-from-source/templates/slurm-vpn.conf.j2
@@ -0,0 +1,109 @@
+#
+# Example slurm.conf file. Please run configurator.html
+# (in doc/html) to build a configuration file customized
+# for your environment.
+#
+#
+# slurm.conf file generated by configurator.html.
+#
+# See the slurm.conf man page for more information.
+#
+ClusterName=CIAB
+ControlMachine={{ slurmctrl }}
+ControlAddr={{ slurmctrl }}-vpn
+#BackupController=
+#BackupAddr=
+#
+SlurmUser=slurm
+#SlurmdUser=root
+SlurmctldPort=6817
+SlurmdPort=6818
+AuthType=auth/munge
+#JobCredentialPrivateKey=
+#JobCredentialPublicCertificate=
+StateSaveLocation=/tmp
+SlurmdSpoolDir=/tmp/slurmd
+SwitchType=switch/none
+MpiDefault=none
+SlurmctldPidFile=/var/run/slurmctld.pid
+SlurmdPidFile=/var/run/slurmd.pid
+ProctrackType=proctrack/pgid
+#PluginDir=
+CacheGroups=0
+#FirstJobId=
+ReturnToService=0
+#MaxJobCount=
+#PlugStackConfig=
+#PropagatePrioProcess=
+#PropagateResourceLimits=
+#PropagateResourceLimitsExcept=
+#Prolog=
+#Epilog=
+#SrunProlog=
+#SrunEpilog=
+#TaskProlog=
+#TaskEpilog=
+#TaskPlugin=
+#TrackWCKey=no
+#TreeWidth=50
+#TmpFS=
+#UsePAM=
+#
+# TIMERS
+SlurmctldTimeout=300
+SlurmdTimeout=300
+InactiveLimit=0
+MinJobAge=300
+KillWait=30
+Waittime=0
+#
+# SCHEDULING
+SchedulerType=sched/backfill
+#SchedulerAuth=
+#SchedulerPort=
+#SchedulerRootFilter=
+SelectType=select/linear
+FastSchedule=1
+#PriorityType=priority/multifactor
+#PriorityDecayHalfLife=14-0
+#PriorityUsageResetPeriod=14-0
+#PriorityWeightFairshare=100000
+#PriorityWeightAge=1000
+#PriorityWeightPartition=10000
+#PriorityWeightJobSize=1000
+#PriorityMaxAge=1-0
+#
+# LOGGING
+SlurmctldDebug=3
+#SlurmctldLogFile=
+SlurmdDebug=3
+#SlurmdLogFile=
+JobCompType=jobcomp/none
+#JobCompLoc=
+#
+# ACCOUNTING
+#JobAcctGatherType=jobacct_gather/linux
+#JobAcctGatherFrequency=30
+#
+#AccountingStorageType=accounting_storage/slurmdbd
+#AccountingStorageHost=
+#AccountingStorageLoc=
+#AccountingStoragePass=
+#AccountingStorageUser=
+#
+MpiParams=ports=12000-12999
+# COMPUTE NODES
+{% set nodelist = [] %}
+{% for queue in slurmqueues %}
+{% for node in groups[queue.group] %}
+{% if nodelist.append(node) %}
+{% endif %}
+{% endfor %}
+{% endfor %}
+{% for node in nodelist|unique %}
+NodeName={{ node }} NodeAddr={{ node }}-vpn Procs={{ hostvars[node]['ansible_processor_cores'] }} State=UNKNOWN
+{% endfor %}
+
+{% for queue in slurmqueues %}
+PartitionName={{ queue.name }} {% if queue.default %}Default=yes{% endif %} Nodes={{ groups[queue.group]|join(',') }}
+{% endfor %}
diff --git a/roles/slurm-from-source/templates/slurm.conf.j2 b/roles/slurm-from-source/templates/slurm.conf.j2
new file mode 100644
index 0000000000000000000000000000000000000000..7832ac9a5be7ec84409c754f6eeed9f84c2d3143
--- /dev/null
+++ b/roles/slurm-from-source/templates/slurm.conf.j2
@@ -0,0 +1,136 @@
+#
+# Example slurm.conf file. Please run configurator.html
+# (in doc/html) to build a configuration file customized
+# for your environment.
+#
+#
+# slurm.conf file generated by configurator.html.
+#
+# See the slurm.conf man page for more information.
+#
+ClusterName={{ clustername }}
+ControlMachine={{ slurmctrl }}
+#ControlAddr=
+#BackupController=
+#BackupAddr=
+#
+SlurmUser=slurm
+SlurmdUser=root
+SlurmctldPort=6817
+SlurmdPort=6818
+AuthType=auth/munge
+#JobCredentialPrivateKey=
+#JobCredentialPublicCertificate=
+StateSaveLocation={{ slurmdatadir }}
+SlurmdSpoolDir={{ slurmdatadir }}
+SwitchType=switch/none
+MpiDefault=none
+SlurmctldPidFile=/var/run/slurmctld.pid
+SlurmdPidFile=/var/run/slurmd.pid
+ProctrackType=proctrack/pgid
+#PluginDir=
+CacheGroups=0
+#FirstJobId=
+ReturnToService=0
+#MaxJobCount=
+#PlugStackConfig=
+#PropagatePrioProcess=
+#PropagateResourceLimits=
+#PropagateResourceLimitsExcept=
+#Prolog=
+#Epilog=
+#SrunProlog=
+#SrunEpilog=
+#TaskProlog=
+#TaskEpilog=
+#TaskPlugin=
+#TrackWCKey=no
+#TreeWidth=50
+#TmpFS=
+#UsePAM=
+#
+# TIMERS
+SlurmctldTimeout=300
+SlurmdTimeout=300
+InactiveLimit=0
+MinJobAge=300
+KillWait=30
+Waittime=0
+#
+# SCHEDULING
+SchedulerType={{ slurmschedulertype }}
+#SchedulerAuth=
+#SchedulerPort=
+#SchedulerRootFilter=
+SelectType={{ slurmselecttype }}
+FastSchedule={{ slurmfastschedule }}
+#PriorityType=priority/multifactor
+#PriorityDecayHalfLife=14-0
+#PriorityUsageResetPeriod=14-0
+#PriorityWeightFairshare=100000
+#PriorityWeightAge=1000
+#PriorityWeightPartition=10000
+#PriorityWeightJobSize=1000
+#PriorityMaxAge=1-0
+#
+# LOGGING
+{% if slurmctlddebug %}
+SlurmctldDebug={{ slurmctlddebug.level }}
+SlurmctldLogFile={{ slurmctlddebug.log }}
+{% else %}
+#SlurmctldDebug=
+#SlurmctldLogFile=
+{% endif %}
+{% if slurmddebug %}
+SlurmdDebug={{ slurmddebug.level }}
+SlurmdLogFile={{ slurmddebug.log }}
+{% else %}
+#SlurmdDebug=
+#SlurmdLogFile=
+{% endif %}
+{% if slurmschedlog %}
+SlurmSchedlogLevel={{ slurmschedlog.level }}
+SlurmSchedLogFile={{ slurmschedlog.log }}
+{% else %}
+#SlurmSchedlogLevel=
+#SlurmSchedLogFile=
+{% endif %}
+JobCompType=jobcomp/none
+#JobCompLoc=
+#
+# ACCOUNTING
+#JobAcctGatherType=jobacct_gather/linux
+#JobAcctGatherFrequency=30
+#
+AccountingStorageType=accounting_storage/slurmdbd
+AccountingStorageHost={{ slurmctrl }}
+#AccountingStorageEnforce=limits,safe
+#AccountingStorageLoc=
+#AccountingStoragePass=
+#AccountingStorageUser=
+#
+#GRES
+GresTypes=gpu
+
+# Fair share
+{% if slurmfairshare.def %}
+PriorityWeightFairshare={{ slurmfairshare.val }}
+{% endif %}
+
+DisableRootJobs=YES
+MpiParams=ports=12000-12999
+# COMPUTE NODES
+{% set nodelist = [] %}
+{% for queue in slurmqueues %}
+{% for node in groups[queue.group] %}
+{% if nodelist.append(node) %}
+{% endif %}
+{% endfor %}
+{% endfor %}
+{% for node in nodelist|unique %}
+NodeName={{ node }} Procs={{ hostvars[node]['ansible_processor_vcpus'] }} State=UNKNOWN
+{% endfor %}
+
+{% for queue in slurmqueues %}
+PartitionName={{ queue.name }} {% if queue.default %}Default=yes{% endif %} Nodes={{ groups[queue.group]|join(',') }}
+{% endfor %}
diff --git a/roles/slurm-from-source/templates/slurm.initd.j2 b/roles/slurm-from-source/templates/slurm.initd.j2
new file mode 100644
index 0000000000000000000000000000000000000000..8e0b0ce083689dfdee92efeb2b8c55142b48830a
--- /dev/null
+++ b/roles/slurm-from-source/templates/slurm.initd.j2
@@ -0,0 +1,338 @@
+#!/bin/bash
+#
+# chkconfig: 345 90 10
+# description: SLURM is a simple resource management system which \
+# manages exclusive access to a set of compute \
+# resources and distributes work to those resources.
+#
+# processname: ${exec_prefix}/sbin/slurmd
+# pidfile: /var/run/slurmd.pid
+#
+# processname: ${exec_prefix}/sbin/slurmctld
+# pidfile: /var/run/slurmctld.pid
+#
+# config: /etc/sysconfig/slurm
+#
+### BEGIN INIT INFO
+# Provides: slurm
+# Required-Start: $remote_fs $syslog $network munge
+# Required-Stop: $remote_fs $syslog $network munge
+# Should-Start: $named
+# Should-Stop: $named
+# Default-Start: 2 3 4 5
+# Default-Stop: 0 1 6
+# Short-Description: slurm daemon management
+# Description: Start slurm to provide resource management
+### END INIT INFO
+
+exec_prefix="{{ slurm_dir }}"
+prefix="{{ slurm_dir }}"
+BINDIR="${exec_prefix}/bin"
+CONFDIR="${prefix}/etc"
+LIBDIR="${exec_prefix}/lib"
+SBINDIR="${exec_prefix}/sbin"
+
+# Source function library.
+if [ -f /etc/rc.status ]; then
+ . /etc/rc.status
+ SUSE=1
+ STARTPROC=startproc
+
+ rc_reset
+else
+
+ # Read configuration defaults to override variables:
+ # $CONFIG, $DAEMON_ARGS, $PIDFILE, $USER, $NICE, $SIGHUP_RELOAD
+ ##
+ for dir in "$sysconfdir/default" "$sysconfdir/sysconfig"; do
+ [ -r "$dir/$INIT_NAME" ] && . "$dir/$INIT_NAME"
+ done
+ [ -z "$DAEMON_ARGS" -a -n "$OPTIONS" ] && DAEMON_ARGS="$OPTIONS"
+ [ "`id | sed 's/^uid=\([0-9]*\).*/\1/'`" -ne 0 ] && unset USER
+ expr -- "$NICE" : '[0-9]*$' >/dev/null 2>&1 && NICE="+$NICE"
+ [ -n "$SIGHUP_RELOAD" -a "$SIGHUP_RELOAD" != 0 ] \
+ && RELOAD=1 || unset RELOAD
+
+ if [ -f /etc/debian_version -a -x /sbin/start-stop-daemon ]; then
+ SYSTEM="DEBIAN"
+ [ -r /etc/default/rcS ] && . /etc/default/rcS
+ [ -r /lib/init/vars.sh ] && . /lib/init/vars.sh
+ [ -r /lib/lsb/init-functions ] && . /lib/lsb/init-functions
+ STARTPROC="start_daemon"
+ elif [ -f /etc/redhat-release -a -r /etc/rc.d/init.d/functions ]; then
+ SYSTEM="REDHAT"
+ . /etc/rc.d/init.d/functions
+ RH_LOCK="/var/lock/subsys/$INIT_NAME"
+ SUSE=0
+ STARTPROC=daemon
+ elif [ -f /etc/SuSE-release -a -r /etc/rc.status ]; then
+ SYSTEM="SUSE"
+ . /etc/rc.status
+ rc_reset
+ elif [ -r /lib/lsb/init-functions ]; then
+ SYSTEM="LSB"
+ . /lib/lsb/init-functions
+ else
+ SYSTEM="OTHER"
+ fi
+
+ function rc_status() {
+ RETVAL=$?
+ }
+ function rc_exit () {
+ exit $RETVAL
+ }
+ RETVAL=0
+fi
+
+# We can not use a starter program without losing environment
+# variables that are critical on Blue Gene systems
+if [ -d /bgl/BlueLight/ppcfloor ]; then
+ STARTPROC=""
+fi
+
+# Source slurm specific configuration
+# This can be used to alter limits for users jobs or set daemon options.
+# For example, the limits for user jobs could be higher or lower than the
+# default limits for user root (e.g. "ulimit -t unlimited" sets an unlimited
+# CPU time limit for spawned user jobs).
+# SLURMCTLD_OPTIONS defines slurmctld command line options. See "man slurmctld"
+# SLURMD_OPTIONS defines slurmd command line options. See "man slurmd"
+if [ -f /etc/sysconfig/slurm ] ; then
+ . /etc/sysconfig/slurm
+else
+ SLURMCTLD_OPTIONS=""
+ SLURMD_OPTIONS=""
+fi
+
+if [ ! -x $BINDIR/scontrol ]; then
+ echo "Could not find $BINDIR/scontrol. Bad path?"
+ exit 1
+fi
+
+if [ ! -f $CONFDIR/slurm.conf ]; then
+ echo "Could not find $CONFDIR/slurm.conf. Bad path?"
+ exit 1
+fi
+
+# setup library paths for slurm and munge support
+export LD_LIBRARY_PATH=$LIBDIR${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}
+
+start() {
+ prog=$1
+ shift
+ echo -n "starting $prog: "
+ unset HOME MAIL USER USERNAME
+ $STARTPROC $SBINDIR/$prog $*
+ rc_status -v
+ echo
+ touch /var/lock/subsys/slurm
+}
+
+stop() {
+ echo -n "stopping $1: "
+ killproc $1 -TERM
+ rc_status -v
+ echo
+ rm -f /var/lock/subsys/slurm
+}
+
+startall() {
+ for prog in `$BINDIR/scontrol show daemons`; do
+ optvar=`echo ${prog}_OPTIONS | tr "a-z" "A-Z"`
+ if [[ ${MULTIPLE_SLURMD} == yes ]] && [[ ${prog} == slurmd ]]
+ then
+ for node in $($BINDIR/scontrol show aliases)
+ do
+ start $prog -N ${node} ${!optvar}
+ done
+ else
+ start $prog ${!optvar}
+ fi
+ done
+}
+
+#
+# status() with slight modifications to take into account
+# instantiations of job manager slurmd's, which should not be
+# counted as "running"
+#
+slurmstatus() {
+ local base=${1##*/}
+ local pid
+ local rpid
+ local pidfile
+ local pidfiles
+ local rc
+
+ pidfile=`grep -i ${base}pid $CONFDIR/slurm.conf | grep -v '^ *#'`
+ if [ $? = 0 ]; then
+ pidfile=${pidfile##*=}
+ pidfile=${pidfile%#*}
+ pidfile=${pidfile//\"/}
+ else
+ pidfile=/var/run/${base}.pid
+ fi
+
+ pid=`pidof -o $$ -o $$PPID -o %PPID -x $1 || \
+ pidof -o $$ -o $$PPID -o %PPID -x ${base}`
+
+ if [ "$base" == "slurmd" ] ; then
+ echo ${pidfile} | grep -q %n
+ if [[ $? -eq 0 ]]
+ then
+ for n in $($BINDIR/scontrol show aliases)
+ do
+ pidfiles="${pidfiles} $(echo ${pidfile} | sed "s/%n/$n/g")"
+ done
+ else
+ pidfiles=${pidfile}
+ fi
+ else
+ pidfiles=${pidfile}
+ fi
+
+ RETVAL=0
+ for pidfile in ${pidfiles}
+ do
+ rc=1
+ if [ -f $pidfile ]; then
+ read rpid < $pidfile
+ if [ "$rpid" != "" -a "$pid" != "" ]; then
+ for i in $pid ; do
+ if [ "$i" = "$rpid" ]; then
+ echo $"${base} (pid $rpid) is running..."
+ rc=0
+ break
+ fi
+ done
+ elif [ "$rpid" != "" -a "$pid" = "" ]; then
+# Due to change in user id, pid file may persist
+# after slurmctld terminates
+ if [ "$base" != "slurmctld" ] ; then
+ echo $"${base} dead but pid file exists"
+ else
+ echo $"${base} is stopped"
+ fi
+ RETVAL=1
+ fi
+ fi
+
+ if [[ $rc -eq 0 ]]
+ then
+ continue
+ fi
+
+ if [ "$base" = "slurmctld" -a "$pid" != "" ] ; then
+ echo $"${base} (pid $pid) is running..."
+ continue
+ fi
+
+ echo $"${base} is stopped"
+ if [ "$RETVAL" == "0" ]; then
+ RETVAL=3
+ fi
+ done
+
+ return $RETVAL
+}
+
+#
+# stop slurm daemons,
+# wait for termination to complete (up to 10 seconds) before returning
+#
+slurmstop() {
+ for prog in `$BINDIR/scontrol show daemons`; do
+ stop $prog
+
+ for i in 1 2 3 4
+ do
+ sleep $i
+ slurmstatus $prog
+ if [ $? != 0 ]; then
+ break
+ fi
+ done
+ done
+
+ # slurmstatus return 1 in case of stopped daemon
+ # and that is what we are looking for here
+ if [[ ${RETVAL} == "1" ]]
+ then
+ RETVAL=0
+ else
+ RETVAL=1
+ fi
+}
+
+#
+# The pathname substitution in daemon command assumes prefix and
+# exec_prefix are same. This is the default, unless the user requests
+# otherwise.
+#
+# Any node can be a slurm controller and/or server.
+#
+case "$1" in
+ start)
+ startall
+ ;;
+ startclean)
+ SLURMCTLD_OPTIONS="-c $SLURMCTLD_OPTIONS"
+ SLURMD_OPTIONS="-c $SLURMD_OPTIONS"
+ startall
+ ;;
+ stop)
+ slurmstop
+ ;;
+ status)
+ anystop=0
+ for prog in `$BINDIR/scontrol show daemons`; do
+ slurmstatus $prog
+ rc=$?
+ if [ $rc != 0 ] ; then
+ anystop=$rc
+ fi
+ done
+ RETVAL=$anystop
+ ;;
+ restart)
+ $0 stop
+ $0 start
+ ;;
+ condrestart)
+ if [ -f /var/lock/subsys/slurm ]; then
+ for prog in `$BINDIR/scontrol show daemons`; do
+ stop $prog
+ sleep 1
+ optvar=`echo ${prog}_OPTIONS | tr "a-z" "A-Z"`
+ if [[ ${MULTIPLE_SLURMD} == yes ]] && [[ ${prog} == slurmd ]]
+ then
+ for node in $($BINDIR/scontrol show aliases)
+ do
+ start $prog -N ${node}
+ done
+ else
+ start $prog ${!optvar}
+ fi
+ done
+ fi
+ ;;
+ reconfig|reload)
+ for prog in `$BINDIR/scontrol show daemons`; do
+ echo -n $"Reloading $prog daemon configuration: "
+ killproc $prog -HUP
+ echo
+ done
+ ;;
+ test)
+ for prog in `$BINDIR/scontrol show daemons`; do
+ echo "$prog runs here"
+ done
+ ;;
+ *)
+ echo "Usage: $0 {start|startclean|stop|status|restart|reconfig|condrestart|test}"
+ exit 1
+ ;;
+esac
+
+rc_exit
diff --git a/roles/slurm-start/tasks/main.yml b/roles/slurm-start/tasks/main.yml
new file mode 100644
index 0000000000000000000000000000000000000000..29f3bc2e623188acbb4fdb843076bd7a675c2c97
--- /dev/null
+++ b/roles/slurm-start/tasks/main.yml
@@ -0,0 +1,13 @@
+---
+- name: start munge
+ service: name=munge state=started
+ sudo: true
+
+- name: start slurmdbd
+ service: name=slurmdbd state=started
+ sudo: true
+ ignore_errors: true
+
+- name: start slurm
+ service: name=slurm state=started
+ sudo: true
diff --git a/roles/slurmdb/handlers/main.yml b/roles/slurmdb/handlers/main.yml
index cbbaae73ee7bef6df63023906addb9f499ecd1f2..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 100644
--- a/roles/slurmdb/handlers/main.yml
+++ b/roles/slurmdb/handlers/main.yml
@@ -1,8 +0,0 @@
----
- - name: restart slurmdbd
- service: name={{ item }} state=restarted
- with_items:
- - slurmdbd
- - slurm
- sudo: true
-
diff --git a/roles/slurmdb/tasks/main.yml b/roles/slurmdb/tasks/main.yml
index f00f19cc6ceb3176603f3efff50fd7da98a9a183..14fbda28113a39d049684311656b7aa0fdca9790 100644
--- a/roles/slurmdb/tasks/main.yml
+++ b/roles/slurmdb/tasks/main.yml
@@ -13,12 +13,23 @@
sudo: true
- name: install mysql local root password
- mysql_user: check_implicit_admin=True login_user=root login_password="{{ sqlrootPasswd }}" name=root password="{{ sqlrootPasswd }}" state=present
+ mysql_user: check_implicit_admin=True login_user=root login_password={{ sqlrootPasswd }} name=root password={{ sqlrootPasswd }} state=present
sudo: true
+- name: install slurmdbd init
+ template: src=slurmdbd.initd.j2 dest=/etc/init.d/slurmdbd mode=755
+ sudo: true
+
+- name: install slurmdb.conf init
+ template: src=slurmdbd.conf.j2 dest={{ slurm_dir }}/etc/slurmdbd.conf
+ sudo: true
+ when: slurm_dir is defined
+
+
- name: install slurmdbd.conf
template: src=slurmdbd.conf.j2 dest=/etc/slurm/slurmdbd.conf
sudo: true
+ when: slurm_dir is not defined
- name: configure database slurmdb localhost
mysql_user: login_user=root login_password="{{ sqlrootPasswd }}" name=slurmdb password="{{ slurmdb_passwd }}" host=localhost priv=*.*:ALL,GRANT state=present
@@ -27,7 +38,6 @@
- name: configure database slurmdb domain
mysql_user: login_user=root login_password="{{ sqlrootPasswd }}" name=slurmdb password="{{ slurmdb_passwd }}" host="{{ ansible_hostname }}"."{{ ansible_domain }}" priv=*.*:ALL,GRANT state=present
sudo: true
-# notify: restart slurmdb
- name: sanity check slrumdbd service
service: "name=slurmdbd enabled=yes state=started"
diff --git a/roles/slurmdb/templates/slurmdbd.initd.j2 b/roles/slurmdb/templates/slurmdbd.initd.j2
new file mode 100644
index 0000000000000000000000000000000000000000..afeb6cab36371c3a773c206a1d57848cbe5b05cc
--- /dev/null
+++ b/roles/slurmdb/templates/slurmdbd.initd.j2
@@ -0,0 +1,223 @@
+#!/bin/bash
+#
+# chkconfig: 345 90 10
+# description: SLURMDBD is a database server interface for \
+# SLURM (Simple Linux Utility for Resource Management).
+#
+# processname: ${exec_prefix}/sbin/slurmdbd
+# pidfile: /var/run/slurmdbd.pid
+#
+# config: /etc/sysconfig/slurm
+#
+### BEGIN INIT INFO
+# Provides: slurmbd
+# Required-Start: $remote_fs $syslog $network munge
+# Required-Stop: $remote_fs $syslog $network munge
+# Should-Start: $named
+# Should-Stop: $named
+# Default-Start: 2 3 4 5
+# Default-Stop: 0 1 6
+# Short-Description: SLURM database daemon
+# Description: Start slurm to provide database server for SLURM
+### END INIT INFO
+
+exec_prefix="{{ slurm_dir }}"
+prefix="{{ slurm_dir }}"
+CONFDIR="${prefix}/etc"
+LIBDIR="${exec_prefix}/lib"
+SBINDIR="${exec_prefix}/sbin"
+
+#Source function library.
+if [ -f /etc/rc.status ]; then
+ . /etc/rc.status
+ SUSE=1
+ STARTPROC=startproc
+
+ rc_reset
+else
+ # Read configuration defaults to override variables:
+ # $CONFIG, $DAEMON_ARGS, $PIDFILE, $USER, $NICE, $SIGHUP_RELOAD
+ ##
+ for dir in "$sysconfdir/default" "$sysconfdir/sysconfig"; do
+ [ -r "$dir/$INIT_NAME" ] && . "$dir/$INIT_NAME"
+ done
+ [ -z "$DAEMON_ARGS" -a -n "$OPTIONS" ] && DAEMON_ARGS="$OPTIONS"
+ [ "`id | sed 's/^uid=\([0-9]*\).*/\1/'`" -ne 0 ] && unset USER
+ expr -- "$NICE" : '[0-9]*$' >/dev/null 2>&1 && NICE="+$NICE"
+ [ -n "$SIGHUP_RELOAD" -a "$SIGHUP_RELOAD" != 0 ] \
+ && RELOAD=1 || unset RELOAD
+
+ if [ -f /etc/debian_version -a -x /sbin/start-stop-daemon ]; then
+ SYSTEM="DEBIAN"
+ [ -r /etc/default/rcS ] && . /etc/default/rcS
+ [ -r /lib/init/vars.sh ] && . /lib/init/vars.sh
+ [ -r /lib/lsb/init-functions ] && . /lib/lsb/init-functions
+ STARTPROC="start_daemon"
+ elif [ -f /etc/redhat-release -a -r /etc/rc.d/init.d/functions ]; then
+ SYSTEM="REDHAT"
+ . /etc/rc.d/init.d/functions
+ RH_LOCK="/var/lock/subsys/$INIT_NAME"
+ SUSE=0
+ STARTPROC=daemon
+ elif [ -f /etc/SuSE-release -a -r /etc/rc.status ]; then
+ SYSTEM="SUSE"
+ . /etc/rc.status
+ rc_reset
+ elif [ -r /lib/lsb/init-functions ]; then
+ SYSTEM="LSB"
+ . /lib/lsb/init-functions
+ else
+ SYSTEM="OTHER"
+ fi
+
+
+ function rc_status() {
+ RETVAL=$?
+ }
+ function rc_exit () {
+ exit $RETVAL
+ }
+ RETVAL=0
+fi
+
+# We can not use a starter program without losing environment
+# variables that are critical on Blue Gene systems
+if [ -d /bgl/BlueLight/ppcfloor ]; then
+ STARTPROC=""
+fi
+
+# Source slurm specific configuration
+# SLURMDBD_OPTIONS defines slurmdbd command line options. See "man slurmdbd"
+if [ -f /etc/sysconfig/slurm ] ; then
+ . /etc/sysconfig/slurm
+else
+ SLURMDBD_OPTIONS=""
+fi
+
+if [ ! -f $CONFDIR/slurmdbd.conf ]; then
+ echo "Could not find $CONFDIR/slurmdbd.conf. Bad path?"
+ exit 1
+fi
+
+# setup library paths for slurm and munge support
+export LD_LIBRARY_PATH=$LIBDIR${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}
+
+start() {
+ prog=$1
+ shift
+ echo -n "starting $prog: "
+ unset HOME MAIL USER USERNAME
+ $STARTPROC $SBINDIR/$prog $SLURMDBD_OPTIONS
+ rc_status -v
+ echo
+ touch /var/lock/subsys/slurmdbd
+}
+
+stop() {
+ echo -n "stopping $1: "
+ killproc $1 -TERM
+ rc_status -v
+ echo
+ rm -f /var/lock/subsys/slurmdbd
+}
+
+slurmstatus() {
+ local base=${1##*/}
+ local pid
+ local rpid
+ local pidfile
+
+ pidfile=`grep -i pidfile $CONFDIR/slurmdbd.conf | grep -v '^ *#'`
+ if [ $? = 0 ]; then
+ pidfile=${pidfile##*=}
+ pidfile=${pidfile%#*}
+ pidfile=${pidfile//\"/}
+ else
+ pidfile=/var/run/slurmdbd.pid
+ fi
+
+ pid=`pidof -o $$ -o $$PPID -o %PPID -x $1 || \
+ pidof -o $$ -o $$PPID -o %PPID -x ${base}`
+
+ if [ -f $pidfile ]; then
+ read rpid < $pidfile
+ if [ "$rpid" != "" -a "$pid" != "" ]; then
+ for i in $pid ; do
+ if [ "$i" = "$rpid" ]; then
+ echo $"${base} (pid $pid) is running..."
+ return 0
+ fi
+ done
+ elif [ "$rpid" != "" -a "$pid" = "" ]; then
+ echo $"${base} dead but pid file exists"
+ return 1
+ fi
+
+ fi
+
+ if [ "$base" = "slurmdbd" -a "$pid" != "" ] ; then
+ echo $"${base} (pid $pid) is running..."
+ return 0
+ fi
+
+ echo $"${base} is stopped"
+
+ return 3
+}
+
+#
+# stop slurm daemons,
+# wait for termination to complete (up to 10 seconds) before returning
+#
+slurmstop() {
+ stop $1
+
+ for i in 1 2 3 4
+ do
+ sleep $i
+ slurmstatus $1
+ if [ $? != 0 ]; then
+ break
+ fi
+ done
+}
+#
+# The pathname substitution in daemon command assumes prefix and
+# exec_prefix are same. This is the default, unless the user requests
+# otherwise.
+#
+# Any node can be a slurm controller and/or server.
+#
+case "$1" in
+ start)
+ start slurmdbd
+ ;;
+ stop)
+ slurmstop slurmdbd
+ ;;
+ status)
+ slurmstatus slurmdbd
+ rc_status -v
+ ;;
+ restart)
+ $0 stop
+ $0 start
+ ;;
+ condrestart)
+ if [ -f /var/lock/subsys/slurm ]; then
+ stop slurmdbd
+ start slurmdbd
+ fi
+ ;;
+ reconfig|reload)
+ echo -n $"Reloading slurmdbd daemon configuration: "
+ killproc slurmdbd -HUP
+ echo
+ ;;
+ *)
+ echo "Usage: $0 {start|stop|status|restart|condrestart|reconfig}"
+ exit 1
+ ;;
+esac
+
+rc_exit
diff --git a/roles/syncExports/tasks/addExports.yml b/roles/syncExports/tasks/addExports.yml
index f91dd861f324e07105e2ad587c0ce2fe2ed8a526..8853541bcd81ebbe805949044d2d030b89dc5bdf 100644
--- a/roles/syncExports/tasks/addExports.yml
+++ b/roles/syncExports/tasks/addExports.yml
@@ -2,26 +2,18 @@
- name: "Create exports if necessary"
file: dest={{ item.src }} state=directory mode=755 owner=root group=root
sudo: true
- delegate_to: "{{ nfs_server }}"
- run_once: true
with_items: exportList
- name: "Templating /etc/exports"
template: src=exports.j2 dest=/etc/exports owner=root group=root mode=644
- delegate_to: "{{ nfs_server }}"
- run_once: true
sudo: true
# Do not do this as a handler, instead do this here as a task so that it happens imediatly after the exports file is created before any clients
# attempt a mount
- name : "Reload exports"
command: exportfs -ra
- delegate_to: "{{ nfs_server }}"
- run_once: true
sudo: true
- name : "Pause ... clients sometimes have errors"
command: sleep 60
- delegate_to: "{{ nfs_server }}"
- run_once: true
sudo: true
diff --git a/roles/syncExports/templates/exports.j2 b/roles/syncExports/templates/exports.j2
index 21c4f552610f7c382d3159c85cc6788786c8df85..b480290d68f0b775b0a7f3ef9b67b58e1de20325 100644
--- a/roles/syncExports/templates/exports.j2
+++ b/roles/syncExports/templates/exports.j2
@@ -1,13 +1,10 @@
{% set iplist = [] %}
{% for export in exportList %}
-{% for group in groupList %}
-{% for node in groups[group.name] %}
-{% if hostvars[node]['ansible_'+group.interface] is defined %}
-{% if iplist.append(hostvars[node]['ansible_'+group.interface]['ipv4']['address']) %}
+{% for node in groups[export.group] %}
+{% if hostvars[node]['ansible_'+export.interface] is defined %}
+{% if iplist.append(hostvars[node]['ansible_'+export.interface]['ipv4']['address']) %}
{% endif %}
{% endif %}
{% endfor %}
-{% endfor %}
{{ export.src }} {% for ip in iplist|unique %}{{ ip }}({{ export.srvopts }}) {% endfor %}
-
{% endfor %}