#!/bin/sh
### BEGIN INIT INFO
# Provides:          google-fluentd
# Required-Start:    $network $local_fs
# Required-Stop:     $network $local_fs
# Default-Start:     2 3 4 5
# Default-Stop:      0 1 6
# Short-Description: data collector for Treasure Data
# Description:       google-fluentd is a data collector
### END INIT INFO


export PATH=/sbin:/usr/sbin:/bin:/usr/bin

TD_AGENT_NAME=google-fluentd
TD_AGENT_HOME=/opt/google-fluentd
TD_AGENT_DEFAULT=/etc/default/google-fluentd
TD_AGENT_NAME_AND_VERSION="$(head -1 ${TD_AGENT_HOME}/version-manifest.txt)"
TD_AGENT_USER=root
TD_AGENT_GROUP=root
TD_AGENT_RUBY=/opt/google-fluentd/embedded/bin/ruby
TD_AGENT_BIN_FILE=/usr/sbin/google-fluentd
TD_AGENT_LOG_FILE=/var/log/google-fluentd/google-fluentd.log
TD_AGENT_PID_FILE=/var/run/google-fluentd/google-fluentd.pid
TD_AGENT_OPTIONS="--use-v1-config --suppress-repeated-stacktrace"

# timeout can be overridden from /etc/default/google-fluentd
STOPTIMEOUT=120

# Read configuration variable file if it is present
if [ -f "${TD_AGENT_DEFAULT}" ]; then
  . "${TD_AGENT_DEFAULT}"
fi

if [ -n "${NAME}" ]; then
  # backward compatibility with omnibus-td-agent <= 2.2.0. will be deleted from future release.
  echo "Warning: Declaring \$NAME in ${TD_AGENT_DEFAULT} for customizing \$PIDFILE has been deprecated. Use \$TD_AGENT_PID_FILE instead." 1>&2
  TD_AGENT_PID_FILE="//var/run/${NAME}/${NAME}.pid"
fi

if [ -n "${DAEMON_ARGS}" ]; then
# TODO: Show warning on use of `DAEMON_ARGS`
# echo "Warning: Declaring \$DAEMON_ARGS in ${TD_AGENT_DEFAULT} has been deprecated. Use \$TD_AGENT_OPTIONS instead." 1>&2
  TD_AGENT_OPTIONS="${DAEMON_ARGS} ${TD_AGENT_OPTIONS}"
fi

# Arguments to run the daemon with
TD_AGENT_ARGS="${TD_AGENT_ARGS:-${TD_AGENT_BIN_FILE} --log ${TD_AGENT_LOG_FILE} ${TD_AGENT_OPTIONS}}"
START_STOP_DAEMON_ARGS=""

# Exit if the package is not installed
[ -x "${TD_AGENT_RUBY}" ] || exit 0

# Source function library.
# Define LSB log_* functions.
# Depend on lsb-base (>= 3.0-6) to ensure that this file is present.
. /lib/lsb/init-functions

# Check the user
if [ -n "${TD_AGENT_USER}" ]; then
  if ! getent passwd | grep -q "^${TD_AGENT_USER}:"; then
    echo "$0: user for running ${TD_AGENT_NAME} doesn't exist: ${TD_AGENT_USER}" >&2
    exit 1
  fi
  mkdir -p "$(dirname "${TD_AGENT_PID_FILE}")"
  chown -R "${TD_AGENT_USER}" "$(dirname "${TD_AGENT_PID_FILE}")"
  START_STOP_DAEMON_ARGS="${START_STOP_DAEMON_ARGS} -c ${TD_AGENT_USER}"
fi

if [ -n "${TD_AGENT_GROUP}" ]; then
  if ! getent group | grep -q "^${TD_AGENT_GROUP}:"; then
    echo "$0: group for running ${TD_AGENT_NAME} doesn't exist: ${TD_AGENT_GROUP}" >&2
    exit 1
  fi
  START_STOP_DAEMON_ARGS="${START_STOP_DAEMON_ARGS} --group ${TD_AGENT_GROUP}"
fi

if [ -n "${TD_AGENT_PID_FILE}" ]; then
  mkdir -p "$(dirname "${TD_AGENT_PID_FILE}")"
  chown -R "${TD_AGENT_USER}" "$(dirname "${TD_AGENT_PID_FILE}")"
  # Work around https://github.com/fluent/fluentd/issues/2735
  # TD_AGENT_ARGS="${TD_AGENT_ARGS} --daemon ${TD_AGENT_PID_FILE}"
  START_STOP_DAEMON_ARGS="${START_STOP_DAEMON_ARGS} --background --make-pidfile"
fi

# This brings the memory usage down which is a good tradeoff for a daemon
# deployed widely. It may slow the program down by increasing lock contention,
# but that seems unlikely in a low-thread application, and we haven't seen CPU
# increase in our load tests.
#
# More details about this change:
# https://www.speedshop.co/2017/12/04/malloc-doubles-ruby-memory.html
export MALLOC_ARENA_MAX=2

kill_by_file() {
  local sig="$1"
  shift 1
  local pid="$(cat "$@" 2>/dev/null || echo "")"
  if [ -n "${pid}" ]; then
    if /bin/kill "${sig}" "${pid}" 1>/dev/null 2>&1; then
      return 0
    else
      return 2
    fi
  else
    return 1
  fi
}

#
# Function that starts the daemon/service
#
do_start() {
  # Set Max number of file descriptors for the safety sake
  # see http://docs.fluentd.org/en/articles/before-install
  ulimit -n 65536 1>/dev/null 2>&1 || true

  # Return
  #   0 if daemon has been started
  #   1 if daemon was already running
  #   2 if daemon could not be started
  #   3 if daemon was not supposed to be started
  if kill_by_file -0 "${TD_AGENT_PID_FILE}"; then
    return 1
  else
    GOOGLE_LOGGING_ENABLE=$(curl --silent --connect-timeout 1 -f -H "Metadata-Flavor: Google" http://169.254.169.254/computeMetadata/v1/instance/attributes/google-logging-enable 2>/dev/null || echo "")
    if [ -n "$GOOGLE_LOGGING_ENABLE" -a "$GOOGLE_LOGGING_ENABLE" = "0" ]; then
        log_warning_msg "Disabled via metadata"
        return 3
    fi

    start-stop-daemon --start --quiet --pidfile "${TD_AGENT_PID_FILE}" --exec "${TD_AGENT_RUBY}" \
      ${START_STOP_DAEMON_ARGS} -- ${TD_AGENT_ARGS} --no-supervisor \
      || return 2
  fi
  # Add code here, if necessary, that waits for the process to be ready
  # to handle requests from services started subsequently which depend
  # on this one.  As a last resort, sleep for some time.
}

#
# Function that stops the daemon/service
#
do_stop() {
  # Return
  #   0 if daemon has been stopped
  #   1 if daemon was already stopped
  #   2 if daemon could not be stopped
  #   other if a failure occurred
  if [ -e "${TD_AGENT_PID_FILE}" ]; then
    # Use own process termination instead of killproc because killproc can't wait SIGTERM
    if kill_by_file -TERM "${TD_AGENT_PID_FILE}"; then
      local i
      for i in $(seq "${STOPTIMEOUT}"); do
        if kill_by_file -0 "${TD_AGENT_PID_FILE}"; then
          sleep 1
        else
          break
        fi
      done
      if kill_by_file -0 "${TD_AGENT_PID_FILE}"; then
        echo -n "Timeout error occurred trying to stop ${TD_AGENT_NAME}..."
        return 2
      else
        rm -f "${TD_AGENT_PID_FILE}"
      fi
    else
      return 1
    fi
  else
    if killproc "${TD_AGENT_PROG_NAME:-${TD_AGENT_NAME}}"; then
      rm -f "${TD_AGENT_PID_FILE}"
    else
      return 2
    fi
  fi
}

#
# Function that sends a SIGHUP to the daemon/service
#
do_reload() {
  kill_by_file -HUP "${TD_AGENT_PID_FILE}"
}

do_restart() {
  if ! do_configtest; then
    return 1
  fi
  local val=0
  do_stop || val="$?"
  case "${val}" in
  0 )
    if ! do_start; then
      return 1
    fi
    ;;
  * ) # Failed to stop
    return 1
    ;;
  esac
}

do_configtest() {
  eval "${TD_AGENT_ARGS} --user ${TD_AGENT_USER} --group ${TD_AGENT_GROUP} --dry-run -q"
}

RETVAL=0
case "$1" in
"start" )
  echo -n "Starting ${TD_AGENT_NAME_AND_VERSION}: "
  do_start || RETVAL="$?"
  case "$RETVAL" in
  0 | 1 )
    log_success_msg "${TD_AGENT_NAME}"
    ;;
  * )
    log_failure_msg "${TD_AGENT_NAME}"
    exit 1
    ;;
  esac
  ;;
"stop" )
  echo -n "Stopping ${TD_AGENT_NAME}: "
  do_stop || RETVAL="$?"
  case "$RETVAL" in
  0 )
    log_success_msg "${TD_AGENT_NAME}"
    ;;
  * )
    log_failure_msg "${TD_AGENT_NAME}"
    exit 1
    ;;
  esac
  ;;
"reload" | "force-reload" )
  echo -n "Reloading ${TD_AGENT_NAME}: "
  if ! do_configtest; then
    log_failure_msg "${TD_AGENT_NAME}"
    exit 1
  fi
  if do_reload; then
    log_success_msg "${TD_AGENT_NAME}"
  else
    log_failure_msg "${TD_AGENT_NAME}"
    exit 1
  fi
  ;;
"restart" )
  echo -n "Restarting ${TD_AGENT_NAME}: "
  if do_restart; then
    log_success_msg "${TD_AGENT_NAME}"
  else
    log_failure_msg "${TD_AGENT_NAME}"
    exit 1
  fi
  ;;
"status" )
  if kill_by_file -0 "${TD_AGENT_PID_FILE}"; then
    log_success_msg "${TD_AGENT_NAME} is running"
  else
    log_failure_msg "${TD_AGENT_NAME} is not running"
    exit 1
  fi
  ;;
"configtest" )
  if do_configtest; then
    log_success_msg "${TD_AGENT_NAME}"
  else
    log_failure_msg "${TD_AGENT_NAME}"
    exit 1
  fi
  ;;
* )
  echo "Usage: $0 {start|stop|reload|restart|force-reload|status|configtest}" >&2
  exit 1
  ;;
esac