git.rot13.org Git - sysadmin-cookbook/blob - recepies/lxc/lxc-watchdog.sh

   1 #! /bin/sh
   2 ### BEGIN INIT INFO
   3 # Provides:          lxc-watchdog
   4 # Required-Start:    $remote_fs $named $network $time
   5 # Required-Stop:     $remote_fs $named $network
   6 # Required-Start:
   7 # Required-Stop:
   8 # Default-Start:     2 3 4 5
   9 # Default-Stop:      0 1 6
  10 # Short-Description: Manage Linux Containers startup/shutdown
  11 # Description:       Uses clever inotify hack to monitor container's
  12 #                    halt/reboot events watching /var/run/utmp
  13 ### END INIT INFO
  14
  15 # Author: Dobrica Pavlinusic <dpavlin@rot13.org>
  16 #
  17 # based on Tony Risinger post to lxc-users mailing list
  18 # http://www.mail-archive.com/lxc-users@lists.sourceforge.net/msg00074.html
  19 #
  20 # Install with:
  21 # ln -sf /srv/sysadmin-cookbook/recepies/lxc/lxc-watchdog.sh /etc/init.d/lxc-watchdog
  22 # update-rc.d lxc-watchdog defaults
  23
  24
  25 which inotifywait >/dev/null || apt-get install inotify-tools
  26
  27
  28 lxc_exists() {
  29         name=$1
  30
  31         if [ ! -e /var/lib/lxc/$name/config ] ; then
  32                 echo "Usage: $0 name"
  33                 lxc_status
  34                 exit 1
  35         fi
  36 }
  37
  38
  39 lxc_rootfs() {
  40         grep '^ *lxc\.rootfs *=' "/var/lib/lxc/$1/config" | cut -d= -f2 | sed 's/^ *//'
  41 }
  42
  43
  44 lxc_status() {
  45         ( find /var/lib/lxc/ -name "config" | cut -d/ -f5 | sort -u | xargs -i lxc-info -n {} | sed "s/'//g" | while read name is status ; do
  46                 boot="-"
  47                 hostname=`cat $(lxc_rootfs $name)/etc/hostname`
  48                 ip=`grep address $(lxc_rootfs $name)/etc/network/interfaces | sed 's/.*address //'`
  49                 test -s /var/lib/lxc/$name/on_boot && boot="boot"
  50                 echo "$name $status $boot $(lxc_rootfs $name) $ip $hostname"
  51         done ) | column -t
  52 }
  53
  54
  55 cleanup_init_scripts() {
  56         rootfs=$(lxc_rootfs $1)
  57
  58         ls \
  59                 $rootfs/etc/rc?.d/*umountfs \
  60                 $rootfs/etc/rc?.d/*umountroot \
  61                 $rootfs/etc/rc?.d/*hwclock* \
  62                 $rootfs/etc/rc?.d/*udev* \
  63         2>/dev/null | xargs -i rm -v {}
  64
  65         echo $1 > $rootfs/etc/hostname
  66 }
  67
  68
  69 setup_inittab() {
  70         rootfs=$(lxc_rootfs $1)
  71         remove=$2
  72         add=$3
  73
  74         # let container respond to kill -SIGPWR
  75         inittab=$rootfs/etc/inittab
  76         if ! grep "$add" ${inittab} >/dev/null ; then
  77                 grep -v "$remove" ${inittab} > ${inittab}.new
  78                 echo $add >> ${inittab}.new
  79                 mv ${inittab}.new ${inittab}
  80                 echo "$inittab modified with $add"
  81         fi
  82 }
  83
  84
  85 lxc_log() {
  86         echo `date +%Y-%m-%dT%H:%M:%S` $*
  87 }
  88
  89
  90 lxc_kill() {
  91         name=$1
  92         sig=$2
  93
  94         init_pid=`lxc-ps -C init -o pid | grep "^$name" | cut -d" " -f2-`
  95         if [ -z "$init_pid" ] ; then
  96                 lxc-info -n $name
  97                 exit 1
  98         fi
  99         lxc_log "$name kill $sig $init_pid"
 100         /bin/kill $sig $init_pid
 101 }
 102
 103 lxc_stop() {
 104         lxc_log "$name stop"
 105         lxc_kill $name -SIGPWR
 106         lxc-wait -n $name -s STOPPED
 107         lxc_log "$name stoped"
 108 #       rm -f /var/lib/lxc/${name}/on_boot
 109 }
 110
 111
 112 lxc_start() {
 113         name=$1
 114
 115         if ! lxc-info -n $name | grep RUNNING ; then
 116                 lxc_log "$name start"
 117                 lxc-start -n $name -o /tmp/${name}.log -d
 118                 lxc-wait  -n $name -s RUNNING
 119                 lxc-info  -n $name
 120                 test -f /var/lib/lxc/${name}/on_boot || echo $name > /var/lib/lxc/${name}/on_boot
 121         fi
 122 }
 123
 124 lxc_watchdog() {
 125 name=$1
 126 rootfs=$(lxc_rootfs $1)
 127
 128 while true; do
 129         vps_utmp=${rootfs}/var/run/utmp
 130         tasks=`wc -l < /cgroup/${name}/tasks`
 131         test -z "$tasks" && exit 1
 132         if [ "$tasks" -eq 1 ]; then
 133
 134                 runlevel="$(runlevel ${vps_utmp})"
 135                 lxc_log "$name runlevel $runlevel"
 136
 137                 case $runlevel in
 138                 N*)
 139                         # nothing for new boot state
 140                 ;;
 141                 ??0)
 142                         lxc_log "$name halt"
 143                         lxc-stop -n "${name}"
 144                         lxc-wait -n ${name} -s STOPPED
 145                         break
 146                 ;;
 147                 ??6)
 148                         lxc_log "$name reboot";
 149                         lxc-stop -n ${name}
 150                         lxc-wait -n ${name} -s STOPPED
 151                         lxc-start -d -n ${name} -o /tmp/${name}.log
 152                 ;;
 153                 *)
 154                         # make sure vps is still running
 155                         state="$(lxc-info -n "${name}" | sed -e 's/.* is //')"
 156                         [ "$state" = "RUNNING" ] || break
 157                 ;;
 158                 esac
 159         else
 160                 lxc_log "$name $tasks tasks"
 161         fi
 162
 163         # time of 5 minutes on it JUST IN CASE...
 164         inotifywait -qqt 300 ${vps_utmp}
 165 done
 166
 167 lxc_log "$name watchdog exited"
 168
 169 }
 170
 171
 172 usage() {
 173         echo "Usage: $0 {start|stop|restart|status|boot|disable} [name name ... ]" >&2
 174         exit 3
 175 }
 176
 177 command_on_lxc() {
 178 command=$1
 179 shift
 180
 181 echo "# $command $1"
 182
 183 case "$command" in
 184
 185 start)
 186         lxc_exists $1
 187         cleanup_init_scripts $1
 188         setup_inittab $1 ":respawn:/sbin/getty.*tty1"   "c1:12345:respawn:/sbin/getty 38400 tty1 linux"
 189         setup_inittab $1 "::power"                      "p0::powerfail:/sbin/init 0"
 190         setup_inittab $1 "::ctrlaltdel"                 "p6::ctrlaltdel:/sbin/init 6"
 191         lxc_start $1
 192         # give container 5 seconds to start more than one process
 193         ( sleep 5 ; nohup $0 watchdog $1 >> /tmp/$1.log 2>/dev/null ) &
 194         ;;
 195 stop|halt)
 196         lxc_exists $1
 197         lxc_stop $1
 198         ;;
 199 reload|force-reload|restart|reboot)
 200         lxc_kill $1 -SIGINT
 201         ;;
 202 watchdog)
 203         lxc_watchdog $1
 204         ;;
 205 boot)
 206         echo $1 > /var/lib/lxc/$1/on_boot
 207         ;;
 208 disable)
 209         echo -n > /var/lib/lxc/$1/on_boot
 210         ;;
 211 *)
 212         usage
 213         ;;
 214
 215 esac
 216
 217 }
 218
 219 command=$1
 220 test -z "$command" && usage
 221 test "$command" = "status" && lxc_status && exit
 222 shift
 223
 224 if [ -z "$1" ] ; then
 225         ls /var/lib/lxc/*/on_boot | while read path ; do
 226                 name=`echo $path | cut -d/ -f5`
 227                 if [ "$command" != "start" -o "$command" = "start" -a -s $path ] ; then
 228                         command_on_lxc $command $name
 229                 else
 230                         echo "# skip $command $name"
 231                 fi
 232         done
 233 else
 234         while [ ! -z "$1" ] ; do
 235                 command_on_lxc $command $1
 236                 shift
 237         done
 238 fi
 239