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 $CTDB_SYS_ETCDIR/init.d/$_service_name ]; then
216 $_nice $CTDB_SYS_ETCDIR/init.d/$_service_name "$_op"
217 elif [ -x $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name ]; then
218 $_nice $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name "$_op"
228 ######################################################
229 # simulate /sbin/service (niced) on platforms that don't have it
236 ######################################################
237 # Cached retrieval of PNN from local node. This never changes so why
238 # open a client connection to the server each time this is needed?
239 # This sets $pnn - this avoid an unnecessary subprocess.
242 _pnn_file="${CTDB_SCRIPT_VARDIR}/my-pnn"
243 if [ ! -f "$_pnn_file" ] ; then
244 ctdb pnn | sed -e 's@.*:@@' >"$_pnn_file"
247 read pnn <"$_pnn_file"
250 ######################################################
251 # wrapper around /proc/ settings to allow them to be hooked
253 # 1st arg is relative path under /proc/, 2nd arg is value to set
256 echo "$2" >"/proc/$1"
261 if [ -w "/proc/$1" ] ; then
266 ######################################################
267 # wrapper around getting file contents from /proc/ to allow
268 # this to be hooked for testing
269 # 1st arg is relative path under /proc/
275 ######################################################
276 # Print up to $_max kernel stack traces for processes named $_program
277 program_stack_traces ()
283 for _pid in $(pidof "$_prog") ; do
284 [ $_count -le $_max ] || break
286 # Do this first to avoid racing with process exit
287 _stack=$(get_proc "${_pid}/stack" 2>/dev/null)
288 if [ -n "$_stack" ] ; then
289 echo "Stack trace for ${_prog}[${_pid}]:"
291 _count=$(($_count + 1))
296 ######################################################
297 # Ensure $service_name is set
298 assert_service_name ()
300 [ -n "$service_name" ] || die "INTERNAL ERROR: \$service_name not set"
303 ######################################################
304 # check a set of directories is available
305 # return 1 on a missing directory
306 # directories are read from stdin
307 ######################################################
308 ctdb_check_directories_probe()
310 while IFS="" read d ; do
316 [ -d "${d}/." ] || return 1
321 ######################################################
322 # check a set of directories is available
323 # directories are read from stdin
324 ######################################################
325 ctdb_check_directories()
327 ctdb_check_directories_probe || {
328 echo "ERROR: $service_name directory \"$d\" not available"
333 ######################################################
334 # check a set of tcp ports
335 # usage: ctdb_check_tcp_ports <ports...>
336 ######################################################
338 # This flag file is created when a service is initially started. It
339 # is deleted the first time TCP port checks for that service succeed.
340 # Until then ctdb_check_tcp_ports() prints a more subtle "error"
341 # message if a port check fails.
342 _ctdb_check_tcp_common ()
345 _d="${CTDB_SCRIPT_VARDIR}/failcount"
346 _ctdb_service_started_file="${_d}/${service_name}.started"
349 ctdb_check_tcp_init ()
351 _ctdb_check_tcp_common
352 mkdir -p "${_ctdb_service_started_file%/*}" # dirname
353 touch "$_ctdb_service_started_file"
356 # Check whether something is listening on all of the given TCP ports
357 # using the "ctdb checktcpport" command.
358 ctdb_check_tcp_ports()
360 if [ -z "$1" ] ; then
361 echo "INTERNAL ERROR: ctdb_check_tcp_ports - no ports specified"
365 for _p ; do # process each function argument (port)
366 _cmd="ctdb checktcpport $_p"
371 _ctdb_check_tcp_common
372 if [ ! -f "$_ctdb_service_started_file" ] ; then
373 echo "ERROR: $service_name tcp port $_p is not responding"
374 debug "\"ctdb checktcpport $_p\" was able to bind to port"
376 echo "INFO: $service_name tcp port $_p is not responding"
382 # Couldn't bind, something already listening, next port...
386 echo "ERROR: unexpected error running \"ctdb checktcpport\""
388 ctdb checktcpport (exited with $_ret) with output:
395 # All ports listening
396 _ctdb_check_tcp_common
397 rm -f "$_ctdb_service_started_file"
401 ######################################################
402 # check a unix socket
403 # usage: ctdb_check_unix_socket SERVICE_NAME <socket_path>
404 ######################################################
405 ctdb_check_unix_socket() {
407 [ -z "$socket_path" ] && return
409 if ! netstat --unix -a -n | grep -q "^unix.*LISTEN.*${socket_path}$"; then
410 echo "ERROR: $service_name socket $socket_path not found"
415 ######################################################
416 # check a command returns zero status
417 # usage: ctdb_check_command <command>
418 ######################################################
419 ctdb_check_command ()
421 _out=$("$@" 2>&1) || {
422 echo "ERROR: $* returned error"
428 ################################################
429 # kill off any TCP connections with the given IP
430 ################################################
431 kill_tcp_connections ()
436 if [ "$2" = "oneway" ] ; then
440 get_tcp_connections_for_ip "$_ip" | {
445 while read _dst _src; do
446 _destport="${_dst##*:}"
449 # we only do one-way killtcp for CIFS
450 139|445) __oneway=true ;;
453 echo "Killing TCP connection $_src $_dst"
454 _connections="${_connections}${_nl}${_src} ${_dst}"
455 if ! $__oneway ; then
456 _connections="${_connections}${_nl}${_dst} ${_src}"
459 _killcount=$(($_killcount + 1))
462 if [ $_killcount -eq 0 ] ; then
466 echo "$_connections" | ctdb killtcp || {
467 echo "Failed to send killtcp control"
473 _remaining=$(get_tcp_connections_for_ip $_ip | wc -l)
475 if [ $_remaining -eq 0 ] ; then
476 echo "Killed $_killcount TCP connections to released IP $_ip"
480 _count=$(($_count + 1))
481 if [ $_count -gt 3 ] ; then
482 echo "Timed out killing tcp connections for IP $_ip ($_remaining remaining)"
486 echo "Waiting for $_remaining connections to be killed for IP $_ip"
492 ##################################################################
493 # kill off the local end for any TCP connections with the given IP
494 ##################################################################
495 kill_tcp_connections_local_only ()
497 kill_tcp_connections "$1" "oneway"
500 ##################################################################
501 # tickle any TCP connections with the given IP
502 ##################################################################
503 tickle_tcp_connections ()
507 get_tcp_connections_for_ip "$_ip" |
511 while read dest src; do
512 echo "Tickle TCP connection $src $dest"
513 ctdb tickle $src $dest >/dev/null 2>&1 || _failed=true
514 echo "Tickle TCP connection $dest $src"
515 ctdb tickle $dest $src >/dev/null 2>&1 || _failed=true
519 echo "Failed to send tickle control"
524 get_tcp_connections_for_ip ()
528 netstat -tn | awk -v ip=$_ip \
529 'index($1, "tcp") == 1 && \
530 (index($4, ip ":") == 1 || index($4, "::ffff:" ip ":") == 1) \
531 && $6 == "ESTABLISHED" \
535 ########################################################
543 # Ensure interface is up
544 ip link set "$_iface" up || \
545 die "Failed to bringup interface $_iface"
547 # Only need to define broadcast for IPv4
553 ip addr add "$_ip/$_maskbits" $_bcast dev "$_iface" || {
554 echo "Failed to add $_ip/$_maskbits on dev $_iface"
558 # Wait 5 seconds for IPv6 addresses to stop being tentative...
559 if [ -z "$_bcast" ] ; then
560 for _x in $(seq 1 10) ; do
561 ip addr show to "${_ip}/128" | grep -q "tentative" || break
565 # If the address was a duplicate then it won't be on the
566 # interface so flag an error.
567 _t=$(ip addr show to "${_ip}/128")
570 echo "Failed to add $_ip/$_maskbits on dev $_iface"
573 *tentative*|*dadfailed*)
574 echo "Failed to add $_ip/$_maskbits on dev $_iface"
575 ip addr del "$_ip/$_maskbits" dev "$_iface"
582 delete_ip_from_iface()
588 # This could be set globally for all interfaces but it is probably
589 # better to avoid surprises, so limit it the interfaces where CTDB
590 # has public IP addresses. There isn't anywhere else convenient
591 # to do this so just set it each time. This is much cheaper than
592 # remembering and re-adding secondaries.
593 set_proc "sys/net/ipv4/conf/${_iface}/promote_secondaries" 1
595 ip addr del "$_ip/$_maskbits" dev "$_iface" || {
596 echo "Failed to del $_ip on dev $_iface"
601 # If the given IP is hosted then print 2 items: maskbits and iface
607 *:*) _family="inet6" ; _bits=128 ;;
608 *) _family="inet" ; _bits=32 ;;
611 ip addr show to "${_addr}/${_bits}" 2>/dev/null | \
612 awk -v family="${_family}" \
613 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
614 sub("@.*", "", iface) } \
615 $1 ~ /inet/ { mask = $2; sub(".*/", "", mask); \
616 print mask, iface, family }'
621 _addr="${1%/*}" # Remove optional maskbits
623 set -- $(ip_maskbits_iface $_addr)
624 if [ -n "$1" ] ; then
627 echo "Removing public address $_addr/$_maskbits from device $_iface"
628 delete_ip_from_iface $_iface $_addr $_maskbits >/dev/null 2>&1
632 drop_all_public_ips ()
634 while read _ip _x ; do
636 done <"${CTDB_PUBLIC_ADDRESSES:-/dev/null}"
641 set_proc_maybe sys/net/ipv4/route/flush 1
642 set_proc_maybe sys/net/ipv6/route/flush 1
645 ########################################################
646 # Interface monitoring
648 # If the interface is a virtual one (e.g. VLAN) then get the
649 # underlying interface
650 interface_get_real ()
652 # Output of "ip link show <iface>"
655 # Extract the full interface description to see if it is a VLAN
656 _t=$(echo "$_iface_info" |
657 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
661 # VLAN: use the underlying interface, after the '@'
665 # Not a regular VLAN. For backward compatibility, assume
666 # there is some other sort of VLAN that doesn't have the
667 # '@' in the output and only use what is before a '.'. If
668 # there is no '.' then this will be the whole interface
674 # Check whether an interface is operational
679 _iface_info=$(ip link show "$_iface" 2>&1) || {
680 echo "ERROR: Monitored interface ${_iface} does not exist"
685 # If the interface is a virtual one (e.g. VLAN) then get the
686 # underlying interface.
687 _realiface=$(interface_get_real "$_iface_info")
689 if _bi=$(get_proc "net/bonding/${_realiface}" 2>/dev/null) ; then
690 # This is a bond: various monitoring strategies
691 echo "$_bi" | grep -q 'Currently Active Slave: None' && {
692 echo "ERROR: No active slaves for bond device ${_realiface}"
695 echo "$_bi" | grep -q '^MII Status: up' || {
696 echo "ERROR: public network interface ${_realiface} is down"
699 echo "$_bi" | grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' && {
700 # This works around a bug in the driver where the
701 # overall bond status can be up but none of the actual
702 # physical interfaces have a link.
703 echo "$_bi" | grep 'MII Status:' | tail -n +2 | grep -q '^MII Status: up' || {
704 echo "ERROR: No active slaves for 802.ad bond device ${_realiface}"
714 # loopback is always working
718 # we don't know how to test ib links
722 ethtool "$_iface" | grep -q 'Link detected: yes' || {
723 # On some systems, this is not successful when a
724 # cable is plugged but the interface has not been
725 # brought up previously. Bring the interface up
727 ip link set "$_iface" up
728 ethtool "$_iface" | grep -q 'Link detected: yes' || {
729 echo "ERROR: No link on the public network interface ${_iface}"
739 ########################################################
741 _ctdb_counter_common () {
742 _service_name="${1:-${service_name:-${script_name}}}"
743 _counter_file="${CTDB_SCRIPT_VARDIR}/failcount/${_service_name}"
744 mkdir -p "${_counter_file%/*}" # dirname
746 ctdb_counter_init () {
747 _ctdb_counter_common "$1"
751 ctdb_counter_incr () {
752 _ctdb_counter_common "$1"
755 echo -n 1 >> "$_counter_file"
757 ctdb_counter_get () {
758 _ctdb_counter_common "$1"
760 stat -c "%s" "$_counter_file" 2>/dev/null || echo 0
762 ctdb_check_counter () {
763 _msg="${1:-error}" # "error" - anything else is silent on fail
764 _op="${2:--ge}" # an integer operator supported by test
765 _limit="${3:-${service_fail_limit}}"
768 _size=$(ctdb_counter_get "$1")
771 if [ "$_op" != "%" ] ; then
772 if [ $_size $_op $_limit ] ; then
776 if [ $(($_size $_op $_limit)) -eq 0 ] ; then
781 if [ "$_msg" = "error" ] ; then
782 echo "ERROR: $_size consecutive failures for $_service_name, marking node unhealthy"
790 ########################################################
792 ctdb_setup_service_state_dir ()
794 service_state_dir="${CTDB_SCRIPT_VARDIR}/service_state/${1:-${service_name}}"
795 mkdir -p "$service_state_dir" || {
796 echo "Error creating state dir \"$service_state_dir\""
801 ########################################################
802 # Managed status history, for auto-start/stop
804 _ctdb_managed_common ()
806 _ctdb_managed_file="${CTDB_SCRIPT_VARDIR}/managed_history/${service_name}"
809 ctdb_service_managed ()
812 mkdir -p "${_ctdb_managed_file%/*}" # dirname
813 touch "$_ctdb_managed_file"
816 ctdb_service_unmanaged ()
819 rm -f "$_ctdb_managed_file"
822 is_ctdb_previously_managed_service ()
825 [ -f "$_ctdb_managed_file" ]
828 ##################################################################
829 # Reconfigure a service on demand
831 _ctdb_service_reconfigure_common ()
833 _d="${CTDB_SCRIPT_VARDIR}/service_status/${service_name}"
835 _ctdb_service_reconfigure_flag="$_d/reconfigure"
838 ctdb_service_needs_reconfigure ()
840 _ctdb_service_reconfigure_common
841 [ -e "$_ctdb_service_reconfigure_flag" ]
844 ctdb_service_set_reconfigure ()
846 _ctdb_service_reconfigure_common
847 >"$_ctdb_service_reconfigure_flag"
850 ctdb_service_unset_reconfigure ()
852 _ctdb_service_reconfigure_common
853 rm -f "$_ctdb_service_reconfigure_flag"
856 ctdb_service_reconfigure ()
858 echo "Reconfiguring service \"${service_name}\"..."
859 ctdb_service_unset_reconfigure
860 service_reconfigure || return $?
864 # Default service_reconfigure() function does nothing.
865 service_reconfigure ()
870 ctdb_reconfigure_take_lock ()
872 _ctdb_service_reconfigure_common
873 _lock="${_d}/reconfigure_lock"
874 mkdir -p "${_lock%/*}" # dirname
879 # This is overkill but will work if we need to extend this to
880 # allow certain events to run multiple times in parallel
881 # (e.g. takeip) and write multiple PIDs to the file.
883 if [ -n "$_locker_event" ] ; then
885 if [ -n "$_pid" -a "$_pid" != $$ ] && \
886 kill -0 "$_pid" 2>/dev/null ; then
892 printf "%s\n%s\n" "$event_name" $$ >"$_lock"
897 ctdb_reconfigure_release_lock ()
899 _ctdb_service_reconfigure_common
900 _lock="${_d}/reconfigure_lock"
905 ctdb_replay_monitor_status ()
907 echo "Replaying previous status for this script due to reconfigure..."
908 # Leading separator ('|') is missing in some versions...
909 _out=$(ctdb scriptstatus -X | grep -E "^\|?monitor\|${script_name}\|")
910 # Output looks like this:
911 # |monitor|60.nfs|1|ERROR|1314764004.030861|1314764004.035514|foo bar|
912 # This is the cheapest way of getting fields in the middle.
913 set -- $(IFS="|" ; echo $_out)
916 # The error output field can include colons so we'll try to
917 # preserve them. The weak checking at the beginning tries to make
918 # this work for both broken (no leading '|') and fixed output.
920 _err_out="${_out#*monitor|${script_name}|*|*|*|*|}"
922 OK) : ;; # Do nothing special.
924 # Recast this as an error, since we can't exit with the
925 # correct negative number.
927 _err_out="[Replay of TIMEDOUT scriptstatus - note incorrect return code.] ${_err_out}"
930 # Recast this as an OK, since we can't exit with the
931 # correct negative number.
933 _err_out="[Replay of DISABLED scriptstatus - note incorrect return code.] ${_err_out}"
935 *) : ;; # Must be ERROR, do nothing special.
937 if [ -n "$_err_out" ] ; then
943 ctdb_service_check_reconfigure ()
947 # We only care about some events in this function. For others we
949 case "$event_name" in
950 monitor|ipreallocated|reconfigure) : ;;
954 if ctdb_reconfigure_take_lock ; then
955 # No events covered by this function are running, so proceed
957 case "$event_name" in
959 (ctdb_service_reconfigure)
963 if ctdb_service_needs_reconfigure ; then
964 ctdb_service_reconfigure
969 ctdb_reconfigure_release_lock
971 # Somebody else is running an event we don't want to collide
972 # with. We proceed with caution.
973 case "$event_name" in
975 # Tell whoever called us to retry.
979 # Defer any scheduled reconfigure and just run the
980 # rest of the ipreallocated event, as per the
981 # eventscript. There's an assumption here that the
982 # event doesn't depend on any scheduled reconfigure.
983 # This is true in the current code.
987 # There is most likely a reconfigure in progress so
988 # the service is possibly unstable. As above, we
989 # defer any scheduled reconfigured. We also replay
990 # the previous monitor status since that's the best
991 # information we have.
992 ctdb_replay_monitor_status
998 ##################################################################
999 # Does CTDB manage this service? - and associated auto-start/stop
1001 ctdb_compat_managed_service ()
1003 if [ "$1" = "yes" -a "$2" = "$service_name" ] ; then
1004 CTDB_MANAGED_SERVICES="$CTDB_MANAGED_SERVICES $2"
1008 is_ctdb_managed_service ()
1012 # $t is used just for readability and to allow better accurate
1013 # matching via leading/trailing spaces
1014 t=" $CTDB_MANAGED_SERVICES "
1016 # Return 0 if "<space>$service_name<space>" appears in $t
1017 if [ "${t#* ${service_name} }" != "${t}" ] ; then
1021 # If above didn't match then update $CTDB_MANAGED_SERVICES for
1022 # backward compatibility and try again.
1023 ctdb_compat_managed_service "$CTDB_MANAGES_VSFTPD" "vsftpd"
1024 ctdb_compat_managed_service "$CTDB_MANAGES_SAMBA" "samba"
1025 ctdb_compat_managed_service "$CTDB_MANAGES_WINBIND" "winbind"
1026 ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "apache2"
1027 ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "httpd"
1028 ctdb_compat_managed_service "$CTDB_MANAGES_ISCSI" "iscsi"
1029 ctdb_compat_managed_service "$CTDB_MANAGES_CLAMD" "clamd"
1030 ctdb_compat_managed_service "$CTDB_MANAGES_NFS" "nfs"
1032 t=" $CTDB_MANAGED_SERVICES "
1034 # Return 0 if "<space>$service_name<space>" appears in $t
1035 [ "${t#* ${service_name} }" != "${t}" ]
1038 ctdb_start_stop_service ()
1042 # Allow service-start/service-stop pseudo-events to start/stop
1043 # services when we're not auto-starting/stopping and we're not
1045 case "$event_name" in
1047 if is_ctdb_managed_service ; then
1048 die 'service-start event not permitted when service is managed'
1050 if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
1051 die 'service-start event not permitted with $CTDB_SERVICE_AUTOSTARTSTOP = yes'
1057 if is_ctdb_managed_service ; then
1058 die 'service-stop event not permitted when service is managed'
1060 if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
1061 die 'service-stop event not permitted with $CTDB_SERVICE_AUTOSTARTSTOP = yes'
1068 # Do nothing unless configured to...
1069 [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] || return 0
1071 [ "$event_name" = "monitor" ] || return 0
1073 if is_ctdb_managed_service ; then
1074 if ! is_ctdb_previously_managed_service ; then
1075 echo "Starting service \"$service_name\" - now managed"
1076 background_with_logging ctdb_service_start
1080 if is_ctdb_previously_managed_service ; then
1081 echo "Stopping service \"$service_name\" - no longer managed"
1082 background_with_logging ctdb_service_stop
1088 ctdb_service_start ()
1090 # The service is marked managed if we've ever tried to start it.
1091 ctdb_service_managed
1093 service_start || return $?
1099 ctdb_service_stop ()
1101 ctdb_service_unmanaged
1105 # Default service_start() and service_stop() functions.
1107 # These may be overridden in an eventscript.
1110 service "$service_name" start
1115 service "$service_name" stop
1118 ##################################################################
1120 ctdb_standard_event_handler ()
1127 _family="$1" ; shift
1128 if [ "$_family" = "inet6" ] ; then
1129 _iptables_cmd="ip6tables"
1131 _iptables_cmd="iptables"
1134 # iptables doesn't like being re-entered, so flock-wrap it.
1135 flock -w 30 "${CTDB_SCRIPT_VARDIR}/iptables.flock" "$_iptables_cmd" "$@"
1138 # AIX (and perhaps others?) doesn't have mktemp
1139 if ! type mktemp >/dev/null 2>&1 ; then
1143 if [ "$1" = "-d" ] ; then
1147 _d="${TMPDIR:-/tmp}"
1148 _hex10=$(dd if=/dev/urandom count=20 2>/dev/null | \
1150 sed -e 's@\(..........\).*@\1@')
1151 _t="${_d}/tmp.${_hex10}"
1164 ########################################################
1166 ########################################################
1172 tickledir="${CTDB_SCRIPT_VARDIR}/tickles"
1173 mkdir -p "$tickledir"
1177 # What public IPs do I hold?
1178 _ips=$(ctdb -X ip | awk -F'|' -v pnn=$pnn '$3 == pnn {print $2}')
1180 # IPs as a regexp choice
1181 _ipschoice="($(echo $_ips | sed -e 's/ /|/g' -e 's/\./\\\\./g'))"
1183 # Record connections to our public IPs in a temporary file.
1184 # This temporary file is in CTDB's private state directory and
1185 # $$ is used to avoid a very rare race involving CTDB's script
1186 # debugging. No security issue, nothing to see here...
1187 _my_connections="${tickledir}/${_port}.connections.$$"
1189 awk -v destpat="^${_ipschoice}:${_port}\$" \
1190 '$1 == "tcp" && $6 == "ESTABLISHED" && $4 ~ destpat {print $5, $4}' |
1191 sort >"$_my_connections"
1193 # Record our current tickles in a temporary file
1194 _my_tickles="${tickledir}/${_port}.tickles.$$"
1195 for _i in $_ips ; do
1196 ctdb -X gettickles $_i $_port |
1197 awk -F'|' 'NR > 1 { printf "%s:%s %s:%s\n", $2, $3, $4, $5 }'
1199 sort >"$_my_tickles"
1201 # Add tickles for connections that we haven't already got tickles for
1202 comm -23 "$_my_connections" "$_my_tickles" |
1203 while read _src _dst ; do
1204 ctdb addtickle $_src $_dst
1207 # Remove tickles for connections that are no longer there
1208 comm -13 "$_my_connections" "$_my_tickles" |
1209 while read _src _dst ; do
1210 ctdb deltickle $_src $_dst
1213 rm -f "$_my_connections" "$_my_tickles"
1215 # Remove stale files from killed scripts
1216 find "$tickledir" -type f -mmin +10 | xargs -r rm
1219 ########################################################
1220 # load a site local config file
1221 ########################################################
1223 [ -n "$CTDB_RC_LOCAL" -a -x "$CTDB_RC_LOCAL" ] && {
1227 [ -x $CTDB_BASE/rc.local ] && {
1228 . $CTDB_BASE/rc.local
1231 [ -d $CTDB_BASE/rc.local.d ] && {
1232 for i in $CTDB_BASE/rc.local.d/* ; do
1233 [ -x "$i" ] && . "$i"
1237 script_name="${0##*/}" # basename
1238 service_fail_limit=1