ctdb/config/events/legacy/01.reclock.script

   1 #!/bin/sh
   2 # script to check accessibility to the reclock file on a node
   3
   4 [ -n "$CTDB_BASE" ] || \
   5     CTDB_BASE=$(d=$(dirname "$0") ; cd -P "$d" ; dirname "$PWD")
   6
   7 . "${CTDB_BASE}/functions"
   8
   9 load_script_options
  10
  11 ctdb_setup_state_dir "service" "reclock"
  12
  13 ctdb_get_reclock_option ()
  14 {
  15         _reclock_opt_file="${CTDB_SCRIPT_VARDIR}/recovery_lock.cache"
  16
  17         if [ -f "$_reclock_opt_file" ] ; then
  18                 . "$_reclock_opt_file"
  19                 return
  20         fi
  21
  22         ctdb_translate_option "cluster" \
  23                               "recovery lock" \
  24                               "CTDB_RECOVERY_LOCK" >"$_reclock_opt_file"
  25
  26         . "$_reclock_opt_file"
  27 }
  28
  29 ctdb_get_reclock_option
  30
  31 # If CTDB_RECOVERY_LOCK specifies a helper then exit because this
  32 # script can't do anything useful.
  33 case "$CTDB_RECOVERY_LOCK" in
  34 !*) exit 0 ;;
  35 esac
  36
  37 case "$1" in
  38 init)
  39         ctdb_counter_init
  40
  41         if [ -n "$CTDB_RECOVERY_LOCK" ] ; then
  42             d=$(dirname "$CTDB_RECOVERY_LOCK")
  43             mkdir -vp "$d"
  44         fi
  45         ;;
  46
  47 monitor)
  48         # Early exit if not using a reclock file
  49         [ -n "$CTDB_RECOVERY_LOCK" ] || exit 0
  50
  51         # Try to stat the reclock file as a background process so that
  52         # we don't block in case the cluster filesystem is unavailable
  53         (
  54             if stat "$CTDB_RECOVERY_LOCK" ; then
  55                 # We could stat the file, reset the counter
  56                 ctdb_counter_init
  57             fi
  58         ) >/dev/null 2>&1 &
  59
  60         ctdb_counter_incr
  61         num_fails=$(ctdb_counter_get)
  62         if [ "$num_fails" -ge 200 ] ; then
  63             echo "Reclock file \"$CTDB_RECOVERY_LOCK\" can not be accessed. Shutting down."
  64             df
  65             sleep 1
  66             $CTDB shutdown
  67             exit 1
  68         elif [ "$num_fails" -ge 4 ] ; then
  69                 die "ERROR: ${num_fails} consecutive failures checking reclock"
  70         fi
  71         ;;
  72 esac
  73
  74 exit 0