#!/bin/sh
#
# torque.setup - create pbs_server database and default queue
# 

# print out basic usage info

function usage
  {
  echo "USAGE:  torque.setup [-h|--help] [-d config_path|--config-path=config_path]
                [--bin_path=bin_path] [--sbin_path=sbin_path] [--syslog]
                <USERNAME> [<HOSTNAME.DOMAIN>]

       -h|--help                                Show command help.

       -d config_path|--config-path=CONFIG_PATH Specifies the path of the directory
                                                which is home to the server\'s configuration
                                                files. Default is [$PBS_HOME]

       --prefix=PATH                            Specifies the path to executables in PATH/bin
                                                and PATH/sbin. Default is [$PREFIX]

       --syslog=true|false                      Log errors to syslog. Default is false.
" >&2
  }

#
# logit - write a message to a file descriptor and syslog if desired
#
# usage: logit msg [fd [pri]]
#  msg - Message.
#  fd  - File descriptor to write msg to. Default: 2.
#  pri - Syslog priority. Default: user.error. See -p argument listed in logger man page.
#
# Note: message is written to syslog only if WANTSYSLOG is true.
#

function logit
  {
  # need a message
  [ $# -eq 0 ] && return 1

  # get the message
  msg=$1
  shift

  # get the fd for non-syslog output
  if [ $# -gt 0 ]; then
    fd=$1
    shift
  else
    # default
    fd=2
  fi

  # write to syslog?
  if [ $WANTSYSLOG = true ]; then
    if [ $# -gt 0 ]; then
      pri=$1
      shift
    else
      # default
      pri=user.error
    fi

    # write to syslog
    logger -t $PROG -p $pri $msg
    rc=$?

    if [ $rc -ne 0 ]; then
      echo "$PROG: logger failed with $rc" >&2
      exit $rc 
    fi
  fi

  # write to fd
  echo "$msg" >&$fd

  return 0
  }

 
# return list of ip addresses (one from each interface on this host)
function get_ip_addrs
  {
  ip addr show | awk '/inet [0-9]+/ {split($2, f, "/"); print f[1]}'
  }
 
# return hostname of given ip addr
function ip2hostname
  {
  if [ ! -z "$1" ]; then
    getent hosts $1 | awk '$1 == "'$1'" {print $2; exit}'
  fi
  }
 
# return list containing the hostname of each interface on this machine
function get_hostnames
  {
  for ip in `get_ip_addrs`; do
    hostname=`ip2hostname $ip`
    if [ ! -z "$hostname" ]; then
      # add hostname to the list
      echo "$hostname"
    fi
  done | sort -u
  }

# remove localhost from hostname list if it is not the hostname of this host
function remove_localhost
  {
  hostname_list="$1"
  this_hostname=`hostname`

  for hostname in $hostname_list; do
    echo "$hostname" | egrep \^localhost &>/dev/null
    if [ $? -eq 0 ]; then
      # do not add localhost if it is not the name of this host
      if [ "$hostname" != "$this_hostname" ]; then
        continue
      fi
    fi

    # add hostname to the list
    echo "$hostname"
  done
  }


PROG=`basename $0`

PREFIX=@exec_prefix@
BIN_PATH=$PREFIX/bin
SBIN_PATH=$PREFIX/sbin
PBS_HOME=@PBS_HOME@
WANTSYSLOG=false

#export POSIXLY_CORRECT=1
OPTS=`getopt -o hd: -l config-path: -l prefix: -l syslog: -l help -- "$@"`

if [ $? != 0 ]; then
  logit "getop failed"
  usage
  exit 1
fi

eval set -- "$OPTS"

while true; do
  case "$1" in
    -d|--config-path)

      PBS_HOME="$2"
      shift 2
      ;;

    --prefix)

      PREFIX="$2"
      BIN_PATH=$PREFIX/bin
      SBIN_PATH=$PREFIX/sbin
      shift 2
      ;;

    --syslog)

      WANTSYSLOG=`echo "$2" | tr '[:upper:]' '[:lower:]'`
      shift 2
      ;;

    -h|--help)

      usage
      exit 0
      ;;

    --)
      shift
      break
      ;;

    *)
      break
      ;;

  esac
done

# now check options

if [ $# -lt 1 -o $# -gt 2 ]; then
  usage
  exit 1
fi

if [ ! -d "$BIN_PATH" ]; then
  logit "$BIN_PATH does not exist."
  exit 1
fi

if [ ! -d "$SBIN_PATH" ]; then
  logit "$SBIN_PATH does not exist."
  exit 1
fi

if [ "$WANTSYSLOG" != true -a "$WANTSYSLOG" != false ]; then
  usage
  exit 1
fi

# make sure logger command present if syslog desired
if [ $WANTSYSLOG = true ]; then
  logger -V &>/dev/null
  if [ $? -ne 0 ]; then
    WANTSYSLOG=false
    logit "cannot find logger command."
    exit 1
  fi
fi

export PATH=$SBIN_PATH:$BIN_PATH:$PATH

unset PBSDEBUG
pgrep trqauthd >/dev/null

if [ $? -ne 0 ]; then
  trqauthd
  if [ $? -ne 0 ]; then
    logit "trqauthd failed to start!!! exiting setup"
    exit 1
  else
    logit "trqauthd successfully started" 1 user.info
  fi
fi
  
# user and optional hostname
USER=$1
HOSTNAME=$2

logit "initializing TORQUE (admin: $USER)" 1 user.info

pgrep pbs_server >/dev/null

if [ $? -eq 0 ]; then
  logit "pbs_server already running... run 'qterm' to stop pbs_server and rerun"
  exit 1
fi

pbs_server -t create -d $PBS_HOME
# Starting in TORQUE 3.1 the server is multi-threaded.
# We need to pause a second to allow the server to finish coming
# up. If we go to qmgr right away it will fail.
sleep 2

pgrep pbs_server >/dev/null

if [ $? -ne 0 ]; then
  logit "ERROR: pbs_server failed to start, check syslog and server logs for more information"
  exit 1
fi

# setup admins

if [ ! -z "$HOSTNAME" ]; then
  # supplied to script by caller
  hostname_list=$HOSTNAME
else
  # look up hostnames for all interfaces on this host
  hostname_list_all=`get_hostnames`

  # remove localhost if necessary
  hostname_list=`remove_localhost "$hostname_list_all"`
fi
 
for hostname in $hostname_list; do
  for admin_type in operators managers; do 
    qmgr -c "s s $admin_type += $USER@$hostname"

    if [ $? -ne 0 ] ; then
      logit "ERROR: cannot set $USER@$hostname in $admin_type list"
      qterm
      exit 1
    fi
  done
done

# setup basic parameters

qmgr -c 'set server scheduling = true'
qmgr -c 'set server keep_completed = 300'
qmgr -c 'set server mom_job_sync = true'

# create default queue

qmgr -c 'create queue batch'
qmgr -c 'set queue batch queue_type = execution'
qmgr -c 'set queue batch started = true'
qmgr -c 'set queue batch enabled = true'
qmgr -c 'set queue batch resources_default.walltime = 1:00:00'
qmgr -c 'set queue batch resources_default.nodes = 1'

qmgr -c 'set server default_queue = batch'