1 # Hey Emacs, this is a -*- shell-script -*- !!!
3 # utility functions for ctdb event scripts
5 if [ -z "$CTDB_BASE" ] ; then
6 echo 'CTDB_BASE unset in CTDB functions file'
10 CTDB_VARDIR="/usr/local/var/lib/ctdb"
11 ctdb_rundir="/usr/local/var/run/ctdb"
13 # Only (and always) override these variables in test code
15 if [ -z "$CTDB_SCRIPT_VARDIR" ] ; then
16 CTDB_SCRIPT_VARDIR="/usr/local/var/lib/ctdb/state"
19 if [ -z "$CTDB_SYS_ETCDIR" ] ; then
20 CTDB_SYS_ETCDIR="/etc"
23 if [ -z "$CTDB_HELPER_BINDIR" ] ; then
24 CTDB_HELPER_BINDIR="/usr/local/libexec/ctdb"
27 #######################################
28 # pull in a system config file, if any
30 rewrite_ctdb_options ()
34 _opts_defaults="mode=700"
35 # Get any extra options specified after colon
36 if [ "$CTDB_DBDIR" = "tmpfs" ] ; then
39 _opts="${CTDB_DBDIR#tmpfs:}"
41 # This is an internal variable, only used by ctdbd_wrapper.
42 # It is OK to repeat mount options - last value wins
43 CTDB_DBDIR_TMPFS_OPTIONS="${_opts_defaults}${_opts:+,}${_opts}"
45 CTDB_DBDIR="${ctdb_rundir}/CTDB_DBDIR"
48 CTDB_DBDIR_TMPFS_OPTIONS=""
55 foo="${service_config:-${service_name}}"
56 if [ -n "$foo" ] ; then
62 if [ "$1" != "ctdb" ] ; then
70 if [ -f $CTDB_SYS_ETCDIR/sysconfig/$1 ]; then
71 . $CTDB_SYS_ETCDIR/sysconfig/$1
72 elif [ -f $CTDB_SYS_ETCDIR/default/$1 ]; then
73 . $CTDB_SYS_ETCDIR/default/$1
74 elif [ -f $CTDB_BASE/sysconfig/$1 ]; then
75 . $CTDB_BASE/sysconfig/$1
78 if [ "$1" = "ctdb" ] ; then
79 _config="${CTDBD_CONF:-${CTDB_BASE}/ctdbd.conf}"
80 if [ -r "$_config" ] ; then
91 ##############################################################
93 # CTDB_SCRIPT_DEBUGLEVEL can be overwritten by setting it in a
97 if [ ${CTDB_SCRIPT_DEBUGLEVEL:-2} -ge 4 ] ; then
98 # If there are arguments then echo them. Otherwise expect to
99 # use stdin, which allows us to pass lots of debug using a
101 if [ -n "$1" ] ; then
104 sed -e 's@^@DEBUG: @'
107 if [ -z "$1" ] ; then
122 # Log given message or stdin to either syslog or a CTDB log file
123 # $1 is the tag passed to logger if syslog is in use.
128 case "$CTDB_LOGGING" in
130 if [ -n "$CTDB_LOGGING" ] ; then
131 _file="${CTDB_LOGGING#file:}"
133 _file="/usr/local/var/log/log.ctdb"
136 if [ -n "$*" ] ; then
144 # Handle all syslog:* variants here too. There's no tool to do
145 # the lossy things, so just use logger.
146 logger -t "ctdbd: ${_tag}" $*
151 # When things are run in the background in an eventscript then logging
152 # output might get lost. This is the "solution". :-)
153 background_with_logging ()
156 "$@" 2>&1 </dev/null |
157 script_log "${script_name}&"
163 ##############################################################
164 # check number of args for different events
170 echo "ERROR: must supply interface, IP and maskbits"
176 echo "ERROR: must supply old interface, new interface, IP and maskbits"
183 ##############################################################
184 # determine on what type of system (init style) we are running
187 # only do detection if not already set:
188 [ -z "$CTDB_INIT_STYLE" ] || return
190 if [ -x /sbin/startproc ]; then
191 CTDB_INIT_STYLE="suse"
192 elif [ -x /sbin/start-stop-daemon ]; then
193 CTDB_INIT_STYLE="debian"
195 CTDB_INIT_STYLE="redhat"
199 ######################################################
200 # simulate /sbin/service on platforms that don't have it
201 # _service() makes it easier to hook the service() function for
208 # do nothing, when no service was specified
209 [ -z "$_service_name" ] && return
211 if [ -x /sbin/service ]; then
212 $_nice /sbin/service "$_service_name" "$_op"
213 elif [ -x /usr/sbin/service ]; then
214 $_nice /usr/sbin/service "$_service_name" "$_op"
215 elif [ -x /bin/systemctl ]; then
216 $_nice /bin/systemctl "$_op" "$_service_name"
217 elif [ -x $CTDB_SYS_ETCDIR/init.d/$_service_name ]; then
218 $_nice $CTDB_SYS_ETCDIR/init.d/$_service_name "$_op"
219 elif [ -x $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name ]; then
220 $_nice $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name "$_op"
230 ######################################################
231 # simulate /sbin/service (niced) on platforms that don't have it
238 ######################################################
239 # Cached retrieval of PNN from local node. This never changes so why
240 # open a client connection to the server each time this is needed?
241 # This sets $pnn - this avoid an unnecessary subprocess.
244 _pnn_file="${CTDB_SCRIPT_VARDIR}/my-pnn"
245 if [ ! -f "$_pnn_file" ] ; then
246 ctdb pnn | sed -e 's@.*:@@' >"$_pnn_file"
249 read pnn <"$_pnn_file"
252 # Cached retrieval of private IP address from local node. This never
253 # changes. Sets $ip_address to avoid an unnecessary subprocess.
254 ctdb_get_ip_address ()
256 _ip_addr_file="${CTDB_SCRIPT_VARDIR}/my-ip-address"
257 if [ ! -f "$_ip_addr_file" ] ; then
259 awk -F '|' 'NR == 2 { print $3 }' >"$_ip_addr_file"
262 read ip_address <"$_ip_addr_file"
265 ######################################################
266 # wrapper around /proc/ settings to allow them to be hooked
268 # 1st arg is relative path under /proc/, 2nd arg is value to set
271 echo "$2" >"/proc/$1"
276 if [ -w "/proc/$1" ] ; then
281 ######################################################
282 # wrapper around getting file contents from /proc/ to allow
283 # this to be hooked for testing
284 # 1st arg is relative path under /proc/
290 ######################################################
291 # Print up to $_max kernel stack traces for processes named $_program
292 program_stack_traces ()
298 for _pid in $(pidof "$_prog") ; do
299 [ $_count -le $_max ] || break
301 # Do this first to avoid racing with process exit
302 _stack=$(get_proc "${_pid}/stack" 2>/dev/null)
303 if [ -n "$_stack" ] ; then
304 echo "Stack trace for ${_prog}[${_pid}]:"
306 _count=$(($_count + 1))
311 ######################################################
312 # Ensure $service_name is set
313 assert_service_name ()
315 [ -n "$service_name" ] || die "INTERNAL ERROR: \$service_name not set"
318 ######################################################
319 # check a set of directories is available
320 # return 1 on a missing directory
321 # directories are read from stdin
322 ######################################################
323 ctdb_check_directories_probe()
325 while IFS="" read d ; do
331 [ -d "${d}/." ] || return 1
336 ######################################################
337 # check a set of directories is available
338 # directories are read from stdin
339 ######################################################
340 ctdb_check_directories()
342 ctdb_check_directories_probe || {
343 echo "ERROR: $service_name directory \"$d\" not available"
348 ######################################################
349 # check a set of tcp ports
350 # usage: ctdb_check_tcp_ports <ports...>
351 ######################################################
353 # This flag file is created when a service is initially started. It
354 # is deleted the first time TCP port checks for that service succeed.
355 # Until then ctdb_check_tcp_ports() prints a more subtle "error"
356 # message if a port check fails.
357 _ctdb_check_tcp_common ()
360 _d="${CTDB_SCRIPT_VARDIR}/failcount"
361 _ctdb_service_started_file="${_d}/${service_name}.started"
364 ctdb_check_tcp_init ()
366 _ctdb_check_tcp_common
367 mkdir -p "${_ctdb_service_started_file%/*}" # dirname
368 touch "$_ctdb_service_started_file"
371 # Check whether something is listening on all of the given TCP ports
372 # using the "ctdb checktcpport" command.
373 ctdb_check_tcp_ports()
375 if [ -z "$1" ] ; then
376 echo "INTERNAL ERROR: ctdb_check_tcp_ports - no ports specified"
380 for _p ; do # process each function argument (port)
381 _cmd="ctdb checktcpport $_p"
386 _ctdb_check_tcp_common
387 if [ ! -f "$_ctdb_service_started_file" ] ; then
388 echo "ERROR: $service_name tcp port $_p is not responding"
389 debug "\"ctdb checktcpport $_p\" was able to bind to port"
391 echo "INFO: $service_name tcp port $_p is not responding"
397 # Couldn't bind, something already listening, next port...
401 echo "ERROR: unexpected error running \"ctdb checktcpport\""
403 ctdb checktcpport (exited with $_ret) with output:
410 # All ports listening
411 _ctdb_check_tcp_common
412 rm -f "$_ctdb_service_started_file"
416 ######################################################
417 # check a unix socket
418 # usage: ctdb_check_unix_socket SERVICE_NAME <socket_path>
419 ######################################################
420 ctdb_check_unix_socket() {
422 [ -z "$socket_path" ] && return
424 if ! netstat --unix -a -n | grep -q "^unix.*LISTEN.*${socket_path}$"; then
425 echo "ERROR: $service_name socket $socket_path not found"
430 ######################################################
431 # check a command returns zero status
432 # usage: ctdb_check_command <command>
433 ######################################################
434 ctdb_check_command ()
436 _out=$("$@" 2>&1) || {
437 echo "ERROR: $* returned error"
443 ################################################
444 # kill off any TCP connections with the given IP
445 ################################################
446 kill_tcp_connections ()
452 if [ "$3" = "oneway" ] ; then
456 get_tcp_connections_for_ip "$_ip" | {
461 while read _dst _src; do
462 _destport="${_dst##*:}"
465 # we only do one-way killtcp for CIFS
466 139|445) __oneway=true ;;
469 echo "Killing TCP connection $_src $_dst"
470 _connections="${_connections}${_nl}${_src} ${_dst}"
471 if ! $__oneway ; then
472 _connections="${_connections}${_nl}${_dst} ${_src}"
475 _killcount=$(($_killcount + 1))
478 if [ $_killcount -eq 0 ] ; then
482 echo "$_connections" | \
483 "${CTDB_HELPER_BINDIR}/ctdb_killtcp" "$_iface" || {
484 echo "Failed to kill TCP connections"
488 _remaining=$(get_tcp_connections_for_ip $_ip | wc -l)
490 if [ $_remaining -eq 0 ] ; then
491 echo "Killed $_killcount TCP connections to released IP $_ip"
495 _t="${_remaining}/${_killcount}"
496 echo "Failed to kill TCP connections for IP $_ip (${_t} remaining)"
500 ##################################################################
501 # kill off the local end for any TCP connections with the given IP
502 ##################################################################
503 kill_tcp_connections_local_only ()
505 kill_tcp_connections "$@" "oneway"
508 ##################################################################
509 # tickle any TCP connections with the given IP
510 ##################################################################
511 tickle_tcp_connections ()
515 get_tcp_connections_for_ip "$_ip" |
519 while read dest src; do
520 echo "Tickle TCP connection $src $dest"
521 ctdb tickle $src $dest >/dev/null 2>&1 || _failed=true
522 echo "Tickle TCP connection $dest $src"
523 ctdb tickle $dest $src >/dev/null 2>&1 || _failed=true
527 echo "Failed to send tickle control"
532 get_tcp_connections_for_ip ()
536 ss -tn state established "src [$_ip]" | awk 'NR > 1 {print $3, $4}'
539 ########################################################
547 # Ensure interface is up
548 ip link set "$_iface" up || \
549 die "Failed to bringup interface $_iface"
551 # Only need to define broadcast for IPv4
557 ip addr add "$_ip/$_maskbits" $_bcast dev "$_iface" || {
558 echo "Failed to add $_ip/$_maskbits on dev $_iface"
562 # Wait 5 seconds for IPv6 addresses to stop being tentative...
563 if [ -z "$_bcast" ] ; then
564 for _x in $(seq 1 10) ; do
565 ip addr show to "${_ip}/128" | grep -q "tentative" || break
569 # If the address was a duplicate then it won't be on the
570 # interface so flag an error.
571 _t=$(ip addr show to "${_ip}/128")
574 echo "Failed to add $_ip/$_maskbits on dev $_iface"
577 *tentative*|*dadfailed*)
578 echo "Failed to add $_ip/$_maskbits on dev $_iface"
579 ip addr del "$_ip/$_maskbits" dev "$_iface"
586 delete_ip_from_iface()
592 # This could be set globally for all interfaces but it is probably
593 # better to avoid surprises, so limit it the interfaces where CTDB
594 # has public IP addresses. There isn't anywhere else convenient
595 # to do this so just set it each time. This is much cheaper than
596 # remembering and re-adding secondaries.
597 set_proc "sys/net/ipv4/conf/${_iface}/promote_secondaries" 1
599 ip addr del "$_ip/$_maskbits" dev "$_iface" || {
600 echo "Failed to del $_ip on dev $_iface"
605 # If the given IP is hosted then print 2 items: maskbits and iface
615 ip addr show to "${_addr}/${_bits}" 2>/dev/null | \
616 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
617 sub("@.*", "", iface) } \
618 $1 ~ /inet/ { mask = $2; sub(".*/", "", mask); \
624 _addr="${1%/*}" # Remove optional maskbits
626 set -- $(ip_maskbits_iface $_addr)
627 if [ -n "$1" ] ; then
630 echo "Removing public address $_addr/$_maskbits from device $_iface"
631 delete_ip_from_iface $_iface $_addr $_maskbits >/dev/null 2>&1
635 drop_all_public_ips ()
637 while read _ip _x ; do
639 done <"${CTDB_PUBLIC_ADDRESSES:-/dev/null}"
644 set_proc_maybe sys/net/ipv4/route/flush 1
645 set_proc_maybe sys/net/ipv6/route/flush 1
648 ########################################################
649 # Interface monitoring
651 # If the interface is a virtual one (e.g. VLAN) then get the
652 # underlying interface
653 interface_get_real ()
655 # Output of "ip link show <iface>"
658 # Extract the full interface description to see if it is a VLAN
659 _t=$(echo "$_iface_info" |
660 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
664 # VLAN: use the underlying interface, after the '@'
668 # Not a regular VLAN. For backward compatibility, assume
669 # there is some other sort of VLAN that doesn't have the
670 # '@' in the output and only use what is before a '.'. If
671 # there is no '.' then this will be the whole interface
677 # Check whether an interface is operational
682 _iface_info=$(ip link show "$_iface" 2>&1) || {
683 echo "ERROR: Monitored interface ${_iface} does not exist"
688 # If the interface is a virtual one (e.g. VLAN) then get the
689 # underlying interface.
690 _realiface=$(interface_get_real "$_iface_info")
692 if _bi=$(get_proc "net/bonding/${_realiface}" 2>/dev/null) ; then
693 # This is a bond: various monitoring strategies
694 echo "$_bi" | grep -q 'Currently Active Slave: None' && {
695 echo "ERROR: No active slaves for bond device ${_realiface}"
698 echo "$_bi" | grep -q '^MII Status: up' || {
699 echo "ERROR: public network interface ${_realiface} is down"
702 echo "$_bi" | grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' && {
703 # This works around a bug in the driver where the
704 # overall bond status can be up but none of the actual
705 # physical interfaces have a link.
706 echo "$_bi" | grep 'MII Status:' | tail -n +2 | grep -q '^MII Status: up' || {
707 echo "ERROR: No active slaves for 802.ad bond device ${_realiface}"
717 # loopback is always working
721 # we don't know how to test ib links
725 ethtool "$_iface" | grep -q 'Link detected: yes' || {
726 # On some systems, this is not successful when a
727 # cable is plugged but the interface has not been
728 # brought up previously. Bring the interface up
730 ip link set "$_iface" up
731 ethtool "$_iface" | grep -q 'Link detected: yes' || {
732 echo "ERROR: No link on the public network interface ${_iface}"
742 ########################################################
744 _ctdb_counter_common () {
745 _service_name="${1:-${service_name:-${script_name}}}"
746 _counter_file="${CTDB_SCRIPT_VARDIR}/failcount/${_service_name}"
747 mkdir -p "${_counter_file%/*}" # dirname
749 ctdb_counter_init () {
750 _ctdb_counter_common "$1"
754 ctdb_counter_incr () {
755 _ctdb_counter_common "$1"
758 echo -n 1 >> "$_counter_file"
760 ctdb_counter_get () {
761 _ctdb_counter_common "$1"
763 stat -c "%s" "$_counter_file" 2>/dev/null || echo 0
765 ctdb_check_counter () {
766 _msg="${1:-error}" # "error" - anything else is silent on fail
767 _op="${2:--ge}" # an integer operator supported by test
768 _limit="${3:-${service_fail_limit}}"
771 _size=$(ctdb_counter_get "$1")
774 if [ "$_op" != "%" ] ; then
775 if [ $_size $_op $_limit ] ; then
779 if [ $(($_size $_op $_limit)) -eq 0 ] ; then
784 if [ "$_msg" = "error" ] ; then
785 echo "ERROR: $_size consecutive failures for $_service_name, marking node unhealthy"
793 ########################################################
795 ctdb_setup_service_state_dir ()
797 service_state_dir="${CTDB_SCRIPT_VARDIR}/service_state/${1:-${service_name}}"
798 mkdir -p "$service_state_dir" || {
799 echo "Error creating state dir \"$service_state_dir\""
804 ########################################################
805 # Managed status history, for auto-start/stop
807 _ctdb_managed_common ()
809 _ctdb_managed_file="${CTDB_SCRIPT_VARDIR}/managed_history/${service_name}"
812 ctdb_service_managed ()
815 mkdir -p "${_ctdb_managed_file%/*}" # dirname
816 touch "$_ctdb_managed_file"
819 ctdb_service_unmanaged ()
822 rm -f "$_ctdb_managed_file"
825 is_ctdb_previously_managed_service ()
828 [ -f "$_ctdb_managed_file" ]
831 ##################################################################
832 # Reconfigure a service on demand
834 _ctdb_service_reconfigure_common ()
836 _d="${CTDB_SCRIPT_VARDIR}/service_status/${service_name}"
838 _ctdb_service_reconfigure_flag="$_d/reconfigure"
841 ctdb_service_needs_reconfigure ()
843 _ctdb_service_reconfigure_common
844 [ -e "$_ctdb_service_reconfigure_flag" ]
847 ctdb_service_set_reconfigure ()
849 _ctdb_service_reconfigure_common
850 >"$_ctdb_service_reconfigure_flag"
853 ctdb_service_unset_reconfigure ()
855 _ctdb_service_reconfigure_common
856 rm -f "$_ctdb_service_reconfigure_flag"
859 ctdb_service_reconfigure ()
861 echo "Reconfiguring service \"${service_name}\"..."
862 ctdb_service_unset_reconfigure
863 service_reconfigure || return $?
867 # Default service_reconfigure() function does nothing.
868 service_reconfigure ()
873 ctdb_reconfigure_take_lock ()
875 _ctdb_service_reconfigure_common
876 _lock="${_d}/reconfigure_lock"
877 mkdir -p "${_lock%/*}" # dirname
882 # This is overkill but will work if we need to extend this to
883 # allow certain events to run multiple times in parallel
884 # (e.g. takeip) and write multiple PIDs to the file.
886 if [ -n "$_locker_event" ] ; then
888 if [ -n "$_pid" -a "$_pid" != $$ ] && \
889 kill -0 "$_pid" 2>/dev/null ; then
895 printf "%s\n%s\n" "$event_name" $$ >"$_lock"
900 ctdb_reconfigure_release_lock ()
902 _ctdb_service_reconfigure_common
903 _lock="${_d}/reconfigure_lock"
908 ctdb_replay_monitor_status ()
910 echo "Replaying previous status for this script due to reconfigure..."
911 # Leading separator ('|') is missing in some versions...
912 _out=$(ctdb scriptstatus -X | grep -E "^\|?monitor\|${script_name}\|")
913 # Output looks like this:
914 # |monitor|60.nfs|1|ERROR|1314764004.030861|1314764004.035514|foo bar|
915 # This is the cheapest way of getting fields in the middle.
916 set -- $(IFS="|" ; echo $_out)
919 # The error output field can include colons so we'll try to
920 # preserve them. The weak checking at the beginning tries to make
921 # this work for both broken (no leading '|') and fixed output.
923 _err_out="${_out#*monitor|${script_name}|*|*|*|*|}"
925 OK) : ;; # Do nothing special.
927 # Recast this as an error, since we can't exit with the
928 # correct negative number.
930 _err_out="[Replay of TIMEDOUT scriptstatus - note incorrect return code.] ${_err_out}"
933 # Recast this as an OK, since we can't exit with the
934 # correct negative number.
936 _err_out="[Replay of DISABLED scriptstatus - note incorrect return code.] ${_err_out}"
938 *) : ;; # Must be ERROR, do nothing special.
940 if [ -n "$_err_out" ] ; then
946 ctdb_service_check_reconfigure ()
950 # We only care about some events in this function. For others we
952 case "$event_name" in
953 monitor|ipreallocated|reconfigure) : ;;
957 if ctdb_reconfigure_take_lock ; then
958 # No events covered by this function are running, so proceed
960 case "$event_name" in
962 (ctdb_service_reconfigure)
966 if ctdb_service_needs_reconfigure ; then
967 ctdb_service_reconfigure
972 ctdb_reconfigure_release_lock
974 # Somebody else is running an event we don't want to collide
975 # with. We proceed with caution.
976 case "$event_name" in
978 # Tell whoever called us to retry.
982 # Defer any scheduled reconfigure and just run the
983 # rest of the ipreallocated event, as per the
984 # eventscript. There's an assumption here that the
985 # event doesn't depend on any scheduled reconfigure.
986 # This is true in the current code.
990 # There is most likely a reconfigure in progress so
991 # the service is possibly unstable. As above, we
992 # defer any scheduled reconfigured. We also replay
993 # the previous monitor status since that's the best
994 # information we have.
995 ctdb_replay_monitor_status
1001 ##################################################################
1002 # Does CTDB manage this service? - and associated auto-start/stop
1004 ctdb_compat_managed_service ()
1006 if [ "$1" = "yes" -a "$2" = "$service_name" ] ; then
1007 CTDB_MANAGED_SERVICES="$CTDB_MANAGED_SERVICES $2"
1011 is_ctdb_managed_service ()
1015 # $t is used just for readability and to allow better accurate
1016 # matching via leading/trailing spaces
1017 t=" $CTDB_MANAGED_SERVICES "
1019 # Return 0 if "<space>$service_name<space>" appears in $t
1020 if [ "${t#* ${service_name} }" != "${t}" ] ; then
1024 # If above didn't match then update $CTDB_MANAGED_SERVICES for
1025 # backward compatibility and try again.
1026 ctdb_compat_managed_service "$CTDB_MANAGES_VSFTPD" "vsftpd"
1027 ctdb_compat_managed_service "$CTDB_MANAGES_SAMBA" "samba"
1028 ctdb_compat_managed_service "$CTDB_MANAGES_WINBIND" "winbind"
1029 ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "apache2"
1030 ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "httpd"
1031 ctdb_compat_managed_service "$CTDB_MANAGES_ISCSI" "iscsi"
1032 ctdb_compat_managed_service "$CTDB_MANAGES_CLAMD" "clamd"
1033 ctdb_compat_managed_service "$CTDB_MANAGES_NFS" "nfs"
1035 t=" $CTDB_MANAGED_SERVICES "
1037 # Return 0 if "<space>$service_name<space>" appears in $t
1038 [ "${t#* ${service_name} }" != "${t}" ]
1041 ctdb_start_stop_service ()
1045 # Allow service-start/service-stop pseudo-events to start/stop
1046 # services when we're not auto-starting/stopping and we're not
1048 case "$event_name" in
1050 if is_ctdb_managed_service ; then
1051 die 'service-start event not permitted when service is managed'
1053 if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
1054 die 'service-start event not permitted with $CTDB_SERVICE_AUTOSTARTSTOP = yes'
1060 if is_ctdb_managed_service ; then
1061 die 'service-stop event not permitted when service is managed'
1063 if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
1064 die 'service-stop event not permitted with $CTDB_SERVICE_AUTOSTARTSTOP = yes'
1071 # Do nothing unless configured to...
1072 [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] || return 0
1074 [ "$event_name" = "monitor" ] || return 0
1076 if is_ctdb_managed_service ; then
1077 if ! is_ctdb_previously_managed_service ; then
1078 echo "Starting service \"$service_name\" - now managed"
1079 background_with_logging ctdb_service_start
1083 if is_ctdb_previously_managed_service ; then
1084 echo "Stopping service \"$service_name\" - no longer managed"
1085 background_with_logging ctdb_service_stop
1091 ctdb_service_start ()
1093 # The service is marked managed if we've ever tried to start it.
1094 ctdb_service_managed
1096 service_start || return $?
1102 ctdb_service_stop ()
1104 ctdb_service_unmanaged
1108 # Default service_start() and service_stop() functions.
1110 # These may be overridden in an eventscript.
1113 service "$service_name" start
1118 service "$service_name" stop
1121 ##################################################################
1123 ctdb_standard_event_handler ()
1130 _family="$1" ; shift
1131 if [ "$_family" = "inet6" ] ; then
1132 _iptables_cmd="ip6tables"
1134 _iptables_cmd="iptables"
1137 # iptables doesn't like being re-entered, so flock-wrap it.
1138 flock -w 30 "${CTDB_SCRIPT_VARDIR}/iptables.flock" "$_iptables_cmd" "$@"
1141 # AIX (and perhaps others?) doesn't have mktemp
1142 if ! type mktemp >/dev/null 2>&1 ; then
1146 if [ "$1" = "-d" ] ; then
1150 _d="${TMPDIR:-/tmp}"
1151 _hex10=$(dd if=/dev/urandom count=20 2>/dev/null | \
1153 sed -e 's@\(..........\).*@\1@')
1154 _t="${_d}/tmp.${_hex10}"
1167 ########################################################
1169 ########################################################
1175 tickledir="${CTDB_SCRIPT_VARDIR}/tickles"
1176 mkdir -p "$tickledir"
1180 # What public IPs do I hold?
1181 _ips=$(ctdb -X ip | awk -F'|' -v pnn=$pnn '$3 == pnn {print $2}')
1183 # IPs and port as ss filters
1185 for _ip in $_ips ; do
1186 _ip_filter="${_ip_filter}${_ip_filter:+ || }src [${_ip}]"
1188 _port_filter="sport == :${_port}"
1190 # Record connections to our public IPs in a temporary file.
1191 # This temporary file is in CTDB's private state directory and
1192 # $$ is used to avoid a very rare race involving CTDB's script
1193 # debugging. No security issue, nothing to see here...
1194 _my_connections="${tickledir}/${_port}.connections.$$"
1195 # Parentheses are needed around the filters for precedence but
1196 # the parentheses can't be empty!
1197 ss -tn state established \
1198 "${_ip_filter:+( ${_ip_filter} )}" \
1199 "${_port_filter:+( ${_port_filter} )}" |
1200 awk 'NR > 1 {print $4, $3}' |
1201 sort >"$_my_connections"
1203 # Record our current tickles in a temporary file
1204 _my_tickles="${tickledir}/${_port}.tickles.$$"
1205 for _i in $_ips ; do
1206 ctdb -X gettickles $_i $_port |
1207 awk -F'|' 'NR > 1 { printf "%s:%s %s:%s\n", $2, $3, $4, $5 }'
1209 sort >"$_my_tickles"
1211 # Add tickles for connections that we haven't already got tickles for
1212 comm -23 "$_my_connections" "$_my_tickles" |
1213 while read _src _dst ; do
1214 ctdb addtickle $_src $_dst
1217 # Remove tickles for connections that are no longer there
1218 comm -13 "$_my_connections" "$_my_tickles" |
1219 while read _src _dst ; do
1220 ctdb deltickle $_src $_dst
1223 rm -f "$_my_connections" "$_my_tickles"
1225 # Remove stale files from killed scripts
1226 find "$tickledir" -type f -mmin +10 | xargs -r rm
1229 ########################################################
1230 # load a site local config file
1231 ########################################################
1233 [ -n "$CTDB_RC_LOCAL" -a -x "$CTDB_RC_LOCAL" ] && {
1237 [ -x $CTDB_BASE/rc.local ] && {
1238 . $CTDB_BASE/rc.local
1241 [ -d $CTDB_BASE/rc.local.d ] && {
1242 for i in $CTDB_BASE/rc.local.d/* ; do
1243 [ -x "$i" ] && . "$i"
1247 script_name="${0##*/}" # basename
1248 service_fail_limit=1