check.sh

   1 #!/bin/sh
   2 ##
   3 ## Pings specified host and runs specified program if no responding
   4 ## Copyright (c) 2005-2008 by Michal Nazarewicz (mina86/AT/mina86.com)
   5 ##
   6 ## This program is free software; you can redistribute it and/or modify
   7 ## it under the terms of the GNU General Public License as published by
   8 ## the Free Software Foundation; either version 3 of the License, or
   9 ## (at your option) any later version.
  10 ##
  11 ## This program is distributed in the hope that it will be useful,
  12 ## but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 ## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14 ## GNU General Public License for more details.
  15 ##
  16 ## You should have received a copy of the GNU General Public License
  17 ## along with this program; if not, see <http://www.gnu.org/licenses/>.
  18 ##
  19 ## This is part of Tiny Applications Collection
  20 ##   -> http://tinyapps.sourceforge.net/
  21 ##
  22
  23 ##
  24 ## I use it to shut down my PC when my brother shuts down his PC which
  25 ## is a router.  It may be also used to run a reconnect script when
  26 ## connection is loast.  It can be also used to monitor another PC and
  27 ## send (let say) SMS when it goes down.
  28 ##
  29
  30 ##
  31 ## Version
  32 ##
  33 version () {
  34         echo 'Check v0.2  (c) 2005,2006 by Michał Nazarewicz'
  35         echo
  36 }
  37
  38
  39 ##
  40 ## Usage
  41 ##
  42 usage () {
  43         cat <<EOF
  44 usage: check.sh [ <options> ] [ [--] <action> [ <arg> ... ] ]
  45 <options> are:
  46   -h --help            Displays help screen and exits
  47   -H --long-help       Displays long help screen and exits
  48   -V --version         Displays vesion and exits
  49   -q --quiet           Displays less messages
  50   -r --retry=<count>   Runs <action> after <count> failed checkings       [10]
  51   -i --interval=<int>  Waits <int> seconds betwenn each checking          [10]
  52   -c --check=<check>   Runs <check> to check       [check_ping www.google.com]
  53   -k --keep-going      Restarts after running action                      [no]
  54   -t --trap=<trpap>    Executes <trap> when signal is trapped         [exit 0]
  55   -T --ignore-sig      Ignores signals                                    [no]
  56
  57   Single letter options may not be given as one argument (eg. -kTq).
  58
  59 <action>  Program to be run when check faileds                          [exit]
  60 <arg> ... Arguments to the <action>                                        [1]
  61
  62 EOF
  63         if [ "X$1" != "Xlong" ]; then return 0; fi
  64 cat <<EOF
  65
  66 <check> must exit with 0 exit code if check succeeded or non-zero otherwise.
  67 Script contains the fallowing built-in check functions:
  68  check_ping <host>             -- checks if <host> answers to ping
  69  check_proc_load <proc> <load> -- checks if <proc>'s CPU load is below <load>
  70                                   (<load> must be an integer betwen 1 and 100)
  71  check_proc <proc>             -- checks if <proc> is running
  72 Each built-in function has also an alias with the first c letter removed.
  73
  74 The script exports variables CHECK_QUIET (empty if false, y if true),
  75 CHECK_INGORESIG (empty if false, y if true), (in case of <action>)
  76 CHECK_EXIT_CODE (exit code of <check>) and CHECK_TMP (path to a temporary
  77 file) which both <check> and <action> should take into account.
  78
  79 Examples:
  80
  81   ./check.sh '-check_proc_load X 95' -k -- killall X -q &
  82     Kills X each time it uses more then 95% of CPU for 5 minutes.  Usefull
  83     when executed from system start scripts since it keeps running silently in
  84     the background.
  85
  86   ./check.sh || /sbin/halt
  87     Halts computer if www.google.com stops responding to pings (most likely
  88     connection was lsot).  May be canceled with Ctrl+C.
  89
  90   ./check.sh '-check_proc foo' -k -q -r100 -i1 -- /path/to/foo' &
  91     Checks whether foo is runing and if not executes it.  Useful when executed
  92     from system start scripts since it keeps running silently in the
  93     background.
  94
  95 EOF
  96 }
  97
  98
  99 ##
 100 ## Init variables
 101 ##
 102 umask 077
 103 SLEEP=10
 104 RETRY=10
 105
 106 KEEPGOING=
 107 CHECK_QUIET=
 108 CHECK_IGNORESIG=
 109 export CHECK_QUIET CHECK_IGNORESIG
 110 TRAPCMD="exit 0"
 111 CMD="check_ping www.google.com"
 112
 113
 114 ##
 115 ## Parse arguments
 116 ##
 117 while [ $# -ne 0 ]; do
 118         case "$1" in
 119         -h|--help)      version; usage     ; exit 0; ;;
 120         -H|--long-help) version; usage long; exit 0; ;;
 121         -V|--version)   version;             exit 0; ;;
 122
 123         -q|--quiet)      CHECK_QUIET=yes  ; ;;
 124         -k|--keep-going) KEEPGOING=yes    ; ;;
 125         -T|--ignore-sig) CHECK_IGNORESIG=y; ;;
 126
 127         -r|--retry)    RETRY=$2  ; shift; ;;
 128         -i|--interval) SLEEP=$2  ; shift; ;;
 129         -c|--count)    CMD=$2    ; shift; ;;
 130         -t|--trap)     TRAPCMD=$2; shift; ;;
 131
 132         -r*) RETRY=${1#-?}  ; ;; (--retry=*)    RETRY=${1#-*=}  ; ;;
 133         -i*) SLEEP=${1#-?}  ; ;; (--interval=*) SLEEP=${1#-*=}  ; ;;
 134         -c*) CMD=${1#-?}    ; ;; (--check=*)    CMD=${1#-*=}    ; ;;
 135         -t*) TRAPCMD=${1#-?}; ;; (--trap=*)     TRAPCMD=${1#-*=}; ;;
 136
 137         --) shift; break; ;;
 138         -*) echo Unknown option: "$1"; exit 1; ;;
 139         *) break; ;;
 140         esac
 141         shift
 142 done
 143 [ $# -eq 0 ] && set -- exit 1
 144
 145
 146 [ -n "$CHECK_QUIET" ] || version
 147
 148
 149 ##
 150 ## Check arguments
 151 ##
 152 if [ "$RETRY" -lt 1 ]; then
 153         echo '<count> must be integer >= 1'
 154         exit 2;
 155 fi
 156 if [ "$SLEEP" -lt 1 ]; then
 157         echo '<interval> must be integer >= 1'
 158         exit 2;
 159 fi
 160
 161
 162 ##
 163 ## Checking function
 164 ##
 165 check_ping () {
 166         ping -c 1 "$1" 2>/dev/null >"$CHECK_TMP" || return 1
 167         [ -n "$CHECK_QUIET" ] || head -n 2 "$CHECK_TMP" |tail -n 1
 168         return 0
 169 }
 170 heck_ping () { check_ping "$@"; }
 171
 172 check_proc_load () {
 173         LOAD="$(ps -o %cpu -C "$1" |tail -n+2)"
 174         if [ ! "$LOAD" ]; then
 175                 [ -n "$CHECK_QUIET" ] || printf '%s: not running\n' "$1"
 176                 return 0
 177         fi
 178         [ -n "$CHECK_QUIET" ] || printf '%s: CPU Load %5s%%\n' "$1" "$LOAD"
 179
 180         [ "${LOAD%%.*}" -lt "$2" ]
 181 }
 182 heck_proc_load () { check_proc_load "$@"; }
 183
 184 check_proc () {
 185         if [ -z "$__CHECK_PROC_PKILL_PRESENT" ]; then
 186                 which pkill >/dev/null 2>&1
 187                 __CHECK_PROC_PKILL_PRESENT=$?
 188         fi
 189
 190         if [ $__CHECK_PROC_PKILL_PRESENT -eq 0 ]; then
 191                 if ! pkill -0 "^$1\$"; then return 1; fi
 192         elif ! ps -C "$1" >/dev/null 2>/dev/null; then
 193                 return 1
 194         fi
 195
 196         if [ -n "$CHECK_QUIET" ]; then return 0; fi
 197         case "$__CHECK_PROC_GLOBAL_VAR" in
 198         (1) printf '%s: running  (/)\r'  "$1"; __CHECK_PROC_GLOBAL_VAR=2; ;;
 199         (2) printf '%s: running  (-)\r'  "$1"; __CHECK_PROC_GLOBAL_VAR=3; ;;
 200         (3) printf '%s: running  (\\)\r' "$1"; __CHECK_PROC_GLOBAL_VAR=0; ;;
 201         (*) printf '%s: running  (|)\r'  "$1"; __CHECK_PROC_GLOBAL_VAR=1; ;;
 202         esac
 203         return 0
 204 }
 205 heck_proc () { check_proc "$@"; }
 206
 207
 208 ##
 209 ## Temporary file
 210 ##
 211
 212 # Find temp dir
 213 CHECK_TMP=.
 214 for N in "$TMPDIR" "$TMP" "$TEMP" ~/tmp /tmp; do
 215         if [ -n "$N" ] && [ -d "$N" ] && [ -w "$N" ]; then
 216                 CHECK_TMP="$N"
 217                 break
 218         fi
 219 done
 220
 221 # tempfile exists
 222 if which mktemp >/dev/null 2>&1; then
 223         export CHECK_TMP=$(mktemp "$CHECK_TMP/check-XXXXXXXX")
 224 elif which tempfile >/dev/null 2>&1; then
 225         export CHECK_TMP=$(tempfile -d "$CHECK_TMP" -p "check-")
 226 else
 227         echo unable to create temporary file, mktemp or tempfile required >&2
 228         exit 0
 229 fi
 230
 231 trap '[ -f "$CHECK_TMP" ] && rm -f -- "$CHECK_TMP"' 0
 232
 233
 234 ##
 235 ## Signals
 236 ##
 237 catch_sig () {
 238         if [ -n "$CHECK_IGNORESIG" ] || [ "X$1" = XUSR1 ] ||
 239                 [ "X$1" = XUSR2 ] || [ "X$1" = XALRM ]; then
 240                 [ -n "$CHECK_QUIET" ] || echo "Got SIG$1; ignoring..."
 241                 return 0
 242         fi
 243
 244         [ -n "$CHECK_QUIET" ] || echo "Got SIG$1"
 245         $TRAPCMD
 246 }
 247
 248 for e in HUP INT QUIT ABRT SEGV PIPE ALRM TERM USR1 USR2; do
 249         # shellcheck disable=SC2064
 250         trap "catch_sig $e" "$e"
 251 done
 252
 253
 254
 255 ##
 256 ## Loop
 257 ##
 258 count=0
 259 while :; do
 260         ## OK
 261         if $CMD; then
 262                 count=0
 263                 sleep "$SLEEP"
 264                 continue
 265         fi
 266
 267         ## Not OK
 268         export CHECK_EXIT_CODE=$?
 269         count=$(( count + 1 ))
 270         [ -n "$CHECK_QUIET" ] || printf " !!! %3d / %3d\r" "$count" "$RETRY"
 271
 272         ## Limit reached
 273         if [ "$count" -ge "$RETRY" ]; then
 274                 [ -n "$CHECK_QUIET" ] || echo ' !!! Checking failed.'
 275                 [ -f "$CHECK_TMP" ] && rm -f -- "$CHECK_TMP"
 276                 "$@"
 277                 [ -n "$KEEPGOING" ] || exit 0;
 278                 count=0
 279         fi
 280
 281         sleep "$SLEEP"
 282 done