tools/ctdb: "ctdb runstate" now accepts optional expected run state arguments
[obnox/ctdb.git] / config / ctdb.init
index 19a7ba0fa8d020bf8c792c56ff0b69b9bf34894f..70dcfa642169408d32bf773942764e59ecb8c110 100755 (executable)
@@ -6,7 +6,7 @@
 # chkconfig:           - 90 01
 #
 # description:                 Starts and stops the clustered tdb daemon
-# pidfile:             /var/run/ctdbd/ctdbd.pid
+# pidfile:             /var/run/ctdb/ctdbd.pid
 #
 
 ### BEGIN INIT INFO
@@ -32,6 +32,10 @@ fi
     LC_ALL=en_US.UTF-8
 }
 
+if [ -f /lib/lsb/init-functions ] ; then
+    . /lib/lsb/init-functions
+fi
+
 # Avoid using root's TMPDIR
 unset TMPDIR
 
@@ -50,6 +54,7 @@ detect_init_style
 export CTDB_INIT_STYLE
 
 ctdbd=${CTDBD:-/usr/sbin/ctdbd}
+pidfile="/var/run/ctdb/ctdbd.pid"
 
 if [ "$CTDB_VALGRIND" = "yes" ]; then
     init_style="valgrind"
@@ -84,6 +89,9 @@ build_ctdb_options () {
     }
     maybe_set "--reclock"                "$CTDB_RECOVERY_LOCK"
 
+    mkdir -p $(dirname "$pidfile")
+    maybe_set "--pidfile"                "$pidfile"
+
     # build up CTDB_OPTIONS variable from optional parameters
     maybe_set "--logfile"                "$CTDB_LOGFILE"
     maybe_set "--nlist"                  "$CTDB_NODES"
@@ -95,7 +103,6 @@ build_ctdb_options () {
     maybe_set "--event-script-dir"       "$CTDB_EVENT_SCRIPT_DIR"
     maybe_set "--transport"              "$CTDB_TRANSPORT"
     maybe_set "-d"                       "$CTDB_DEBUGLEVEL"
-    maybe_set "--debug-hung-script"      "$CTDB_DEBUG_HUNG_SCRIPT"
     maybe_set "--notification-script"    "$CTDB_NOTIFY_SCRIPT"
     maybe_set "--start-as-disabled"      "$CTDB_START_AS_DISABLED"    "yes"
     maybe_set "--start-as-stopped "      "$CTDB_START_AS_STOPPED"     "yes"
@@ -108,41 +115,32 @@ build_ctdb_options () {
     maybe_set "--max-persistent-check-errors" "$CTDB_MAX_PERSISTENT_CHECK_ERRORS"
 }
 
+export_debug_variables ()
+{
+    export CTDB_DEBUG_HUNG_SCRIPT CTDB_EXTERNAL_TRACE
+}
+
 # Log given message or stdin to either syslog or a CTDB log file
 do_log ()
 {
-    if [ "$CTDB_SYSLOG" = "yes" -o \
-       "${CTDB_OPTIONS#*--syslog}" != "$CTDB_OPTIONS" ] ; then
-
-       logger -t "ctdb.init" "$@"
-    else
-       _l="${CTDB_LOGFILE:-/var/log/log.ctdb}"
-       {
-           date
-           if [ -n "$*" ] ; then
-               echo "$*"
-           else
-               cat
-           fi
-       } >>"$_l"
-    fi
+    script_log "ctdb.init" "$@"
 }
 
 select_tdb_checker ()
 {
     # Find the best TDB consistency check available.
     use_tdb_tool_check=false
-    if [ -x /usr/bin/tdbtool ] && \
-       echo "help" | /usr/bin/tdbtool | grep -q check ; then
+    if which tdbtool >/dev/null 2>&1 && \
+       echo "help" | tdbtool | grep -q check ; then
 
        use_tdb_tool_check=true
-    elif [ -x /usr/bin/tdbtool -a -x /usr/bin/tdbdump ] ; then
+    elif which tdbtool >/dev/null 2>&1 && which tdbdump >/dev/null 2>&1 ; then
            do_log <<EOF
 WARNING: The installed 'tdbtool' does not offer the 'check' subcommand.
  Using 'tdbdump' for database checks.
  Consider updating 'tdbtool' for better checks!
 EOF
-    elif [ -x /usr/bin/tdbdump ] ; then
+    elif which tdbdump >/dev/null 2>&1 ; then
        do_log <<EOF
 WARNING: 'tdbtool' is not available.
  Using 'tdbdump' to check the databases.
@@ -222,7 +220,7 @@ wait_until_ready () {
     _timeout="${1:-10}" # default is 10 seconds
 
     _count=0
-    while ! ctdb ping >/dev/null 2>&1 ; do
+    while ! ctdb runstate startup running >/dev/null 2>&1 ; do
        if [ $_count -ge $_timeout ] ; then
            return 1
        fi
@@ -231,14 +229,6 @@ wait_until_ready () {
     done
 }
 
-ctdbd=${CTDBD:-/usr/sbin/ctdbd}
-
-drop_all_public_ips() {
-    while read _ip _iface _rest; do
-       ip addr del $_ip dev $_iface >/dev/null 2>&1
-    done <"${CTDB_PUBLIC_ADDRESSES:-/dev/null}"
-}
-
 start() {
     echo -n $"Starting ctdbd service: "
 
@@ -255,9 +245,11 @@ start() {
 
     build_ctdb_options
 
+    export_debug_variables
+
     # make sure we drop any ips that might still be held if previous
     # instance of ctdb got killed with -9 or similar
-    drop_all_public_ips
+    drop_all_public_ips "ctdb.init"
 
     if select_tdb_checker ; then
        check_persistent_databases || return $?
@@ -296,6 +288,7 @@ start() {
     if [ $RETVAL -eq 0 ] ; then
        if ! wait_until_ready ; then
            RETVAL=1
+           echo "Timed out waiting for initialisation - killing CTDB"
            pkill -9 -f $ctdbd >/dev/null 2>&1
        fi
     fi
@@ -343,6 +336,8 @@ stop() {
     # make sure all ips are dropped, pfkill -9 might leave them hanging around
     drop_all_public_ips
 
+    rm -f "$pidfile"
+
     case $init_style in
        suse)
            # re-set the return code to the recorded RETVAL in order
@@ -365,30 +360,47 @@ restart() {
     start
 }
 
-status() {
-    echo -n $"Checking for ctdbd service: "
-    ctdb ping >/dev/null 2>&1 || {
-       RETVAL=$?
-       echo -n "  ctdbd not running. "
-       case $init_style in
-           suse)
-               set_retval $RETVAL
-               rc_status -v
-               ;;
-           redhat)
-               if [ -f /var/lock/subsys/ctdb ]; then
-                       echo $"ctdb dead but subsys locked"
-                       RETVAL=2
-               else
-                       echo $"ctdb is stopped"
-                       RETVAL=3
-               fi
-               ;;
-       esac
-       return $RETVAL
-    }
-    echo ""
-    ctdb status
+# Given that CTDB_VALGRIND is a debug option we don't support the pid
+# file.  We just do a quick and dirty hack instead.  Otherwise we just
+# end up re-implementing each distro's pidfile support...
+check_status_valgrind ()
+{
+    if pkill -0 -f "valgrind.*${ctdbd}" ; then
+       echo "ctdbd is running under valgrind..."
+       return 0
+    else
+       echo "ctdbd is not running"
+       return 1
+    fi
+}
+
+check_status ()
+{
+    # Backward compatibility.  When we arrange to pass --pidfile to
+    # ctdbd we also create the directory that will contain it.  If
+    # that directory is missing then we don't use the pidfile to check
+    # status.
+    if [ -d $(dirname "$pidfile") ] ; then
+       _pf_opt="-p $pidfile"
+    else
+       _pf_opt=""
+    fi
+
+    case "$init_style" in
+       valgrind)
+           check_status_valgrind
+           ;;
+       suse)
+           checkproc $_pf_opt "$ctdbd"
+           rc_status -v
+           ;;
+       redhat)
+           status $_pf_opt -l "ctdb" "$ctdbd"
+           ;;
+       debian)
+           status_of_proc $_pf_opt "$ctdbd" "ctdb"
+           ;;
+    esac
 }
 
 
@@ -405,14 +417,16 @@ case "$1" in
        restart
        ;;
     status)
-       status
+       check_status
        ;;
     condrestart|try-restart)
-       ctdb status > /dev/null && restart || :
+       if check_status >/dev/null ; then
+           restart
+       fi
        ;;
     cron)
        # used from cron to auto-restart ctdb
-       ctdb status > /dev/null || restart
+       check_status >/dev/null || restart
        ;;
     *)
        echo $"Usage: $0 {start|stop|restart|reload|force-reload|status|cron|condrestart|try-restart}"