ctdb/config/debug-hung-script.sh

   1 #!/bin/sh
   2
   3 [ -n "$CTDB_BASE" ] || \
   4     export CTDB_BASE=$(cd -P $(dirname "$0") ; echo "$PWD")
   5
   6 . "$CTDB_BASE/functions"
   7
   8 loadconfig ctdb
   9
  10 # Testing hook
  11 if [ -n "$CTDB_DEBUG_HUNG_SCRIPT_LOGFILE" ] ; then
  12     exec >>"$CTDB_DEBUG_HUNG_SCRIPT_LOGFILE" 2>&1
  13 fi
  14
  15 (
  16     flock --wait 2 9 || exit 1
  17
  18     echo "===== Start of hung script debug for PID=\"$1\", event=\"$2\" ====="
  19
  20     echo "pstree -p -a ${1}:"
  21     out=$(pstree -p -a $1)
  22     echo "$out"
  23
  24     # Check for processes matching a regular expression and print
  25     # stack staces.  This could help confirm that certain processes
  26     # are stuck in certain places such as the cluster filesystem.  The
  27     # regexp should separate items with "\|" and should not contain
  28     # parentheses.  The default pattern can be replaced for testing.
  29     default_pat='exportfs\|rpcinfo'
  30     pat="${CTDB_DEBUG_HUNG_SCRIPT_STACKPAT:-${default_pat}}"
  31     echo "$out" |
  32     sed -n "s@.*-\(.*${pat}.*\),\([0-9]*\).*@\2 \1@p" |
  33     while read pid name ; do
  34         trace=$(cat "/proc/${pid}/stack" 2>/dev/null)
  35         if [ $? -eq 0 ] ; then
  36             echo "---- Stack trace of interesting process ${pid}[${name}] ----"
  37             echo "$trace"
  38         fi
  39     done
  40
  41     if [ "$2" = "init" ] ; then
  42         exit 0
  43     fi
  44
  45     echo "---- ctdb scriptstatus ${2}: ----"
  46     # No use running several of these in parallel if, say, "releaseip"
  47     # event hangs for multiple IPs.  In that case the output would be
  48     # interleaved in the log and would just be confusing.
  49     ctdb scriptstatus "$2"
  50
  51     echo "===== End of hung script debug for PID=\"$1\", event=\"$2\" ====="
  52
  53 ) 9>"${CTDB_VARDIR}/debug-hung-script.lock"