1 # Hey Emacs, this is a -*- shell-script -*- !!!
3 # utility functions for ctdb event scripts
5 if [ -z "$CTDB_BASE" ] ; then
6 echo 'CTDB_BASE unset in CTDB functions file'
11 # CTDB_VARDIR is used elsewhere
12 # shellcheck disable=SC2034
13 CTDB_VARDIR="/usr/local/var/lib/ctdb"
14 ctdb_rundir="/usr/local/var/run/ctdb"
16 CTDB="${CTDB:-/usr/local/bin/ctdb}"
18 # Only (and always) override these variables in test code
20 if [ -z "$CTDB_SCRIPT_VARDIR" ] ; then
21 CTDB_SCRIPT_VARDIR="/usr/local/var/lib/ctdb/state"
24 if [ -z "$CTDB_SYS_ETCDIR" ] ; then
25 CTDB_SYS_ETCDIR="/etc"
28 if [ -z "$CTDB_HELPER_BINDIR" ] ; then
29 CTDB_HELPER_BINDIR="/usr/local/libexec/ctdb"
32 #######################################
33 # pull in a system config file, if any
35 rewrite_ctdb_options ()
39 _opts_defaults="mode=700"
40 # Get any extra options specified after colon
41 if [ "$CTDB_DBDIR" = "tmpfs" ] ; then
44 _opts="${CTDB_DBDIR#tmpfs:}"
46 # It is OK to repeat mount options - last value wins.
47 # CTDB_DBDIR_TMPFS_OPTIONS is used by ctdbd_wrapper
48 # shellcheck disable=SC2034
49 CTDB_DBDIR_TMPFS_OPTIONS="${_opts_defaults}${_opts:+,}${_opts}"
51 CTDB_DBDIR="${ctdb_rundir}/CTDB_DBDIR"
54 # shellcheck disable=SC2034
55 CTDB_DBDIR_TMPFS_OPTIONS=""
62 foo="${service_config:-${service_name}}"
63 if [ -n "$foo" ] ; then
69 if [ "$1" != "ctdb" ] ; then
77 if [ -f "${CTDB_SYS_ETCDIR}/sysconfig/$1" ]; then
78 . "${CTDB_SYS_ETCDIR}/sysconfig/$1"
79 elif [ -f "${CTDB_SYS_ETCDIR}/default/$1" ]; then
80 . "${CTDB_SYS_ETCDIR}/default/$1"
81 elif [ -f "${CTDB_BASE}/sysconfig/$1" ]; then
82 . "${CTDB_BASE}/sysconfig/$1"
85 if [ "$1" = "ctdb" ] ; then
86 _config="${CTDBD_CONF:-${CTDB_BASE}/ctdbd.conf}"
87 if [ -r "$_config" ] ; then
98 ##############################################################
100 # CTDB_SCRIPT_DEBUGLEVEL can be overwritten by setting it in a
101 # configuration file.
104 if [ "${CTDB_SCRIPT_DEBUGLEVEL:-2}" -ge 4 ] ; then
105 # If there are arguments then echo them. Otherwise expect to
106 # use stdin, which allows us to pass lots of debug using a
108 if [ -n "$1" ] ; then
111 sed -e 's@^@DEBUG: @'
114 if [ -z "$1" ] ; then
129 # Log given message or stdin to either syslog or a CTDB log file
130 # $1 is the tag passed to logger if syslog is in use.
135 case "$CTDB_LOGGING" in
137 if [ -n "$CTDB_LOGGING" ] ; then
138 _file="${CTDB_LOGGING#file:}"
140 _file="/usr/local/var/log/log.ctdb"
143 if [ -n "$*" ] ; then
151 # Handle all syslog:* variants here too. There's no tool to do
152 # the lossy things, so just use logger.
153 logger -t "ctdbd: ${_tag}" "$*"
158 # When things are run in the background in an eventscript then logging
159 # output might get lost. This is the "solution". :-)
160 background_with_logging ()
163 "$@" 2>&1 </dev/null |
164 script_log "${script_name}&"
170 ##############################################################
171 # check number of args for different events
177 echo "ERROR: must supply interface, IP and maskbits"
183 echo "ERROR: must supply old interface, new interface, IP and maskbits"
190 ##############################################################
191 # determine on what type of system (init style) we are running
194 # only do detection if not already set:
195 [ -z "$CTDB_INIT_STYLE" ] || return
197 if [ -x /sbin/startproc ]; then
198 CTDB_INIT_STYLE="suse"
199 elif [ -x /sbin/start-stop-daemon ]; then
200 CTDB_INIT_STYLE="debian"
202 CTDB_INIT_STYLE="redhat"
206 ######################################################
207 # simulate /sbin/service on platforms that don't have it
208 # _service() makes it easier to hook the service() function for
215 # do nothing, when no service was specified
216 [ -z "$_service_name" ] && return
218 if [ -x /sbin/service ]; then
219 $_nice /sbin/service "$_service_name" "$_op"
220 elif [ -x /usr/sbin/service ]; then
221 $_nice /usr/sbin/service "$_service_name" "$_op"
222 elif [ -x /bin/systemctl ]; then
223 $_nice /bin/systemctl "$_op" "$_service_name"
224 elif [ -x "${CTDB_SYS_ETCDIR}/init.d/${_service_name}" ]; then
225 $_nice "${CTDB_SYS_ETCDIR}/init.d/${_service_name}" "$_op"
226 elif [ -x "${CTDB_SYS_ETCDIR}/rc.d/init.d/${_service_name}" ]; then
227 $_nice "${CTDB_SYS_ETCDIR}/rc.d/init.d/${_service_name}" "$_op"
237 ######################################################
238 # simulate /sbin/service (niced) on platforms that don't have it
245 ######################################################
246 # Cached retrieval of PNN from local node. This never changes so why
247 # open a client connection to the server each time this is needed?
248 # This sets $pnn - this avoid an unnecessary subprocess.
251 _pnn_file="${CTDB_SCRIPT_VARDIR}/my-pnn"
252 if [ ! -f "$_pnn_file" ] ; then
253 $CTDB pnn | sed -e 's@.*:@@' >"$_pnn_file"
256 read pnn <"$_pnn_file"
259 # Cached retrieval of private IP address from local node. This never
260 # changes. Sets $ip_address to avoid an unnecessary subprocess.
261 ctdb_get_ip_address ()
263 _ip_addr_file="${CTDB_SCRIPT_VARDIR}/my-ip-address"
264 if [ ! -f "$_ip_addr_file" ] ; then
265 $CTDB -X nodestatus |
266 awk -F '|' 'NR == 2 { print $3 }' >"$_ip_addr_file"
269 # ip_address is used by caller
270 # shellcheck disable=SC2034
271 read ip_address <"$_ip_addr_file"
274 ######################################################
275 # wrapper around /proc/ settings to allow them to be hooked
277 # 1st arg is relative path under /proc/, 2nd arg is value to set
280 echo "$2" >"/proc/$1"
285 if [ -w "/proc/$1" ] ; then
290 ######################################################
291 # wrapper around getting file contents from /proc/ to allow
292 # this to be hooked for testing
293 # 1st arg is relative path under /proc/
299 ######################################################
300 # Print up to $_max kernel stack traces for processes named $_program
301 program_stack_traces ()
307 for _pid in $(pidof "$_prog") ; do
308 [ "$_count" -le "$_max" ] || break
310 # Do this first to avoid racing with process exit
311 _stack=$(get_proc "${_pid}/stack" 2>/dev/null)
312 if [ -n "$_stack" ] ; then
313 echo "Stack trace for ${_prog}[${_pid}]:"
315 _count=$((_count + 1))
320 ######################################################
321 # Ensure $service_name is set
322 assert_service_name ()
324 [ -n "$service_name" ] || die "INTERNAL ERROR: \$service_name not set"
327 ######################################################
328 # check a set of directories is available
329 # return 1 on a missing directory
330 # directories are read from stdin
331 ######################################################
332 ctdb_check_directories_probe()
334 while IFS="" read d ; do
340 [ -d "${d}/." ] || return 1
345 ######################################################
346 # check a set of directories is available
347 # directories are read from stdin
348 ######################################################
349 ctdb_check_directories()
351 ctdb_check_directories_probe || {
352 echo "ERROR: $service_name directory \"$d\" not available"
357 ######################################################
358 # check a set of tcp ports
359 # usage: ctdb_check_tcp_ports <ports...>
360 ######################################################
362 # This flag file is created when a service is initially started. It
363 # is deleted the first time TCP port checks for that service succeed.
364 # Until then ctdb_check_tcp_ports() prints a more subtle "error"
365 # message if a port check fails.
366 _ctdb_check_tcp_common ()
369 _d="${CTDB_SCRIPT_VARDIR}/failcount"
370 _ctdb_service_started_file="${_d}/${service_name}.started"
373 ctdb_check_tcp_init ()
375 _ctdb_check_tcp_common
376 mkdir -p "${_ctdb_service_started_file%/*}" # dirname
377 touch "$_ctdb_service_started_file"
380 # Check whether something is listening on all of the given TCP ports
381 # using the "ctdb checktcpport" command.
382 ctdb_check_tcp_ports()
384 if [ -z "$1" ] ; then
385 echo "INTERNAL ERROR: ctdb_check_tcp_ports - no ports specified"
389 for _p ; do # process each function argument (port)
390 _cmd="$CTDB checktcpport $_p"
395 _ctdb_check_tcp_common
396 if [ ! -f "$_ctdb_service_started_file" ] ; then
397 echo "ERROR: $service_name tcp port $_p is not responding"
398 debug "\"ctdb checktcpport $_p\" was able to bind to port"
400 echo "INFO: $service_name tcp port $_p is not responding"
406 # Couldn't bind, something already listening, next port...
410 echo "ERROR: unexpected error running \"ctdb checktcpport\""
412 $CTDB checktcpport (exited with $_ret) with output:
419 # All ports listening
420 _ctdb_check_tcp_common
421 rm -f "$_ctdb_service_started_file"
425 ######################################################
426 # check a unix socket
427 # usage: ctdb_check_unix_socket SERVICE_NAME <socket_path>
428 ######################################################
429 ctdb_check_unix_socket() {
431 [ -z "$socket_path" ] && return
433 if ! netstat --unix -a -n | grep -q "^unix.*LISTEN.*${socket_path}$"; then
434 echo "ERROR: $service_name socket $socket_path not found"
439 ######################################################
440 # check a command returns zero status
441 # usage: ctdb_check_command <command>
442 ######################################################
443 ctdb_check_command ()
445 _out=$("$@" 2>&1) || {
446 echo "ERROR: $* returned error"
452 ################################################
453 # kill off any TCP connections with the given IP
454 ################################################
455 kill_tcp_connections ()
461 if [ "$3" = "oneway" ] ; then
465 get_tcp_connections_for_ip "$_ip" | {
470 while read _dst _src; do
471 _destport="${_dst##*:}"
474 # we only do one-way killtcp for CIFS
475 139|445) __oneway=true ;;
478 echo "Killing TCP connection $_src $_dst"
479 _connections="${_connections}${_nl}${_src} ${_dst}"
480 if ! $__oneway ; then
481 _connections="${_connections}${_nl}${_dst} ${_src}"
484 _killcount=$((_killcount + 1))
487 if [ $_killcount -eq 0 ] ; then
491 echo "$_connections" | \
492 "${CTDB_HELPER_BINDIR}/ctdb_killtcp" "$_iface" || {
493 echo "Failed to kill TCP connections"
497 _remaining=$(get_tcp_connections_for_ip "$_ip" | wc -l)
499 if [ "$_remaining" -eq 0 ] ; then
500 echo "Killed $_killcount TCP connections to released IP $_ip"
504 _t="${_remaining}/${_killcount}"
505 echo "Failed to kill TCP connections for IP $_ip (${_t} remaining)"
509 ##################################################################
510 # kill off the local end for any TCP connections with the given IP
511 ##################################################################
512 kill_tcp_connections_local_only ()
514 kill_tcp_connections "$@" "oneway"
517 ##################################################################
518 # tickle any TCP connections with the given IP
519 ##################################################################
520 tickle_tcp_connections ()
524 get_tcp_connections_for_ip "$_ip" |
528 while read dest src; do
529 echo "Tickle TCP connection $src $dest"
530 $CTDB tickle "$src" "$dest" >/dev/null 2>&1 || _failed=true
531 echo "Tickle TCP connection $dest $src"
532 $CTDB tickle "$dest" "$src" >/dev/null 2>&1 || _failed=true
536 echo "Failed to send tickle control"
541 get_tcp_connections_for_ip ()
545 ss -tn state established "src [$_ip]" | awk 'NR > 1 {print $3, $4}'
548 ########################################################
556 # Ensure interface is up
557 ip link set "$_iface" up || \
558 die "Failed to bringup interface $_iface"
560 # Only need to define broadcast for IPv4
566 # Intentionally unquoted multi-word value here
567 # shellcheck disable=SC2086
568 ip addr add "$_ip/$_maskbits" $_bcast dev "$_iface" || {
569 echo "Failed to add $_ip/$_maskbits on dev $_iface"
573 # Wait 5 seconds for IPv6 addresses to stop being tentative...
574 if [ -z "$_bcast" ] ; then
575 for _x in $(seq 1 10) ; do
576 ip addr show to "${_ip}/128" | grep -q "tentative" || break
580 # If the address was a duplicate then it won't be on the
581 # interface so flag an error.
582 _t=$(ip addr show to "${_ip}/128")
585 echo "Failed to add $_ip/$_maskbits on dev $_iface"
588 *tentative*|*dadfailed*)
589 echo "Failed to add $_ip/$_maskbits on dev $_iface"
590 ip addr del "$_ip/$_maskbits" dev "$_iface"
597 delete_ip_from_iface()
603 # This could be set globally for all interfaces but it is probably
604 # better to avoid surprises, so limit it the interfaces where CTDB
605 # has public IP addresses. There isn't anywhere else convenient
606 # to do this so just set it each time. This is much cheaper than
607 # remembering and re-adding secondaries.
608 set_proc "sys/net/ipv4/conf/${_iface}/promote_secondaries" 1
610 ip addr del "$_ip/$_maskbits" dev "$_iface" || {
611 echo "Failed to del $_ip on dev $_iface"
616 # If the given IP is hosted then print 2 items: maskbits and iface
626 ip addr show to "${_addr}/${_bits}" 2>/dev/null | \
627 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
628 sub("@.*", "", iface) } \
629 $1 ~ /inet/ { mask = $2; sub(".*/", "", mask); \
635 _addr="${1%/*}" # Remove optional maskbits
637 # Intentional word splitting here
638 # shellcheck disable=SC2046
639 set -- $(ip_maskbits_iface "$_addr")
640 if [ -n "$1" ] ; then
643 echo "Removing public address $_addr/$_maskbits from device $_iface"
644 delete_ip_from_iface "$_iface" "$_addr" "$_maskbits" >/dev/null 2>&1
648 drop_all_public_ips ()
650 # _x is intentionally ignored
651 # shellcheck disable=SC2034
652 while read _ip _x ; do
654 done <"${CTDB_PUBLIC_ADDRESSES:-/dev/null}"
659 set_proc_maybe sys/net/ipv4/route/flush 1
660 set_proc_maybe sys/net/ipv6/route/flush 1
663 ########################################################
664 # Interface monitoring
666 # If the interface is a virtual one (e.g. VLAN) then get the
667 # underlying interface
668 interface_get_real ()
670 # Output of "ip link show <iface>"
673 # Extract the full interface description to see if it is a VLAN
674 _t=$(echo "$_iface_info" |
675 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
679 # VLAN: use the underlying interface, after the '@'
683 # Not a regular VLAN. For backward compatibility, assume
684 # there is some other sort of VLAN that doesn't have the
685 # '@' in the output and only use what is before a '.'. If
686 # there is no '.' then this will be the whole interface
692 # Check whether an interface is operational
697 _iface_info=$(ip link show "$_iface" 2>&1) || {
698 echo "ERROR: Monitored interface ${_iface} does not exist"
703 # If the interface is a virtual one (e.g. VLAN) then get the
704 # underlying interface.
705 _realiface=$(interface_get_real "$_iface_info")
707 if _bi=$(get_proc "net/bonding/${_realiface}" 2>/dev/null) ; then
708 # This is a bond: various monitoring strategies
709 echo "$_bi" | grep -q 'Currently Active Slave: None' && {
710 echo "ERROR: No active slaves for bond device ${_realiface}"
713 echo "$_bi" | grep -q '^MII Status: up' || {
714 echo "ERROR: public network interface ${_realiface} is down"
717 echo "$_bi" | grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' && {
718 # This works around a bug in the driver where the
719 # overall bond status can be up but none of the actual
720 # physical interfaces have a link.
721 echo "$_bi" | grep 'MII Status:' | tail -n +2 | grep -q '^MII Status: up' || {
722 echo "ERROR: No active slaves for 802.ad bond device ${_realiface}"
732 # loopback is always working
736 # we don't know how to test ib links
740 ethtool "$_iface" | grep -q 'Link detected: yes' || {
741 # On some systems, this is not successful when a
742 # cable is plugged but the interface has not been
743 # brought up previously. Bring the interface up
745 ip link set "$_iface" up
746 ethtool "$_iface" | grep -q 'Link detected: yes' || {
747 echo "ERROR: No link on the public network interface ${_iface}"
757 ########################################################
759 _ctdb_counter_common () {
760 _service_name="${1:-${service_name:-${script_name}}}"
761 _counter_file="${CTDB_SCRIPT_VARDIR}/failcount/${_service_name}"
762 mkdir -p "${_counter_file%/*}" # dirname
764 ctdb_counter_init () {
765 _ctdb_counter_common "$1"
769 ctdb_counter_incr () {
770 _ctdb_counter_common "$1"
773 echo -n 1 >> "$_counter_file"
775 ctdb_counter_get () {
776 _ctdb_counter_common "$1"
778 stat -c "%s" "$_counter_file" 2>/dev/null || echo 0
781 ########################################################
783 ctdb_setup_service_state_dir ()
785 service_state_dir="${CTDB_SCRIPT_VARDIR}/service_state/${1:-${service_name}}"
786 mkdir -p "$service_state_dir" || {
787 echo "Error creating state dir \"$service_state_dir\""
792 ########################################################
793 # Managed status history, for auto-start/stop
795 _ctdb_managed_common ()
797 _ctdb_managed_file="${CTDB_SCRIPT_VARDIR}/managed_history/${service_name}"
800 ctdb_service_managed ()
803 mkdir -p "${_ctdb_managed_file%/*}" # dirname
804 touch "$_ctdb_managed_file"
807 ctdb_service_unmanaged ()
810 rm -f "$_ctdb_managed_file"
813 is_ctdb_previously_managed_service ()
816 [ -f "$_ctdb_managed_file" ]
819 ##################################################################
820 # Reconfigure a service on demand
822 _ctdb_service_reconfigure_common ()
824 _d="${CTDB_SCRIPT_VARDIR}/service_status/${service_name}"
826 _ctdb_service_reconfigure_flag="$_d/reconfigure"
829 ctdb_service_needs_reconfigure ()
831 _ctdb_service_reconfigure_common
832 [ -e "$_ctdb_service_reconfigure_flag" ]
835 ctdb_service_set_reconfigure ()
837 _ctdb_service_reconfigure_common
838 >"$_ctdb_service_reconfigure_flag"
841 ctdb_service_unset_reconfigure ()
843 _ctdb_service_reconfigure_common
844 rm -f "$_ctdb_service_reconfigure_flag"
847 ctdb_service_reconfigure ()
849 echo "Reconfiguring service \"${service_name}\"..."
850 ctdb_service_unset_reconfigure
851 service_reconfigure || return $?
855 # Default service_reconfigure() function does nothing.
856 service_reconfigure ()
861 ctdb_reconfigure_take_lock ()
863 _ctdb_service_reconfigure_common
864 _lock="${_d}/reconfigure_lock"
865 mkdir -p "${_lock%/*}" # dirname
870 # This is overkill but will work if we need to extend this to
871 # allow certain events to run multiple times in parallel
872 # (e.g. takeip) and write multiple PIDs to the file.
874 if [ -n "$_locker_event" ] ; then
876 if [ -n "$_pid" -a "$_pid" != $$ ] && \
877 kill -0 "$_pid" 2>/dev/null ; then
883 printf "%s\n%s\n" "$event_name" $$ >"$_lock"
888 ctdb_reconfigure_release_lock ()
890 _ctdb_service_reconfigure_common
891 _lock="${_d}/reconfigure_lock"
896 ctdb_replay_monitor_status ()
898 echo "Replaying previous status for this script due to reconfigure..."
899 # Leading separator ('|') is missing in some versions...
900 _out=$($CTDB scriptstatus -X | grep -E "^\|?monitor\|${script_name}\|")
901 # Output looks like this:
902 # |monitor|60.nfs|1|ERROR|1314764004.030861|1314764004.035514|foo bar|
903 # This is the cheapest way of getting fields in the middle.
904 # Intentional word splitting here
905 # shellcheck disable=SC2046,2086
906 set -- $(IFS="|" ; echo $_out)
909 # The error output field can include colons so we'll try to
910 # preserve them. The weak checking at the beginning tries to make
911 # this work for both broken (no leading '|') and fixed output.
913 _err_out="${_out#*monitor|${script_name}|*|*|*|*|}"
915 OK) : ;; # Do nothing special.
917 # Recast this as an error, since we can't exit with the
918 # correct negative number.
920 _err_out="[Replay of TIMEDOUT scriptstatus - note incorrect return code.] ${_err_out}"
923 # Recast this as an OK, since we can't exit with the
924 # correct negative number.
926 _err_out="[Replay of DISABLED scriptstatus - note incorrect return code.] ${_err_out}"
928 *) : ;; # Must be ERROR, do nothing special.
930 if [ -n "$_err_out" ] ; then
936 ctdb_service_check_reconfigure ()
940 # We only care about some events in this function. For others we
942 case "$event_name" in
943 monitor|ipreallocated|reconfigure) : ;;
947 if ctdb_reconfigure_take_lock ; then
948 # No events covered by this function are running, so proceed
950 case "$event_name" in
952 (ctdb_service_reconfigure)
956 if ctdb_service_needs_reconfigure ; then
957 ctdb_service_reconfigure
962 ctdb_reconfigure_release_lock
964 # Somebody else is running an event we don't want to collide
965 # with. We proceed with caution.
966 case "$event_name" in
968 # Tell whoever called us to retry.
972 # Defer any scheduled reconfigure and just run the
973 # rest of the ipreallocated event, as per the
974 # eventscript. There's an assumption here that the
975 # event doesn't depend on any scheduled reconfigure.
976 # This is true in the current code.
980 # There is most likely a reconfigure in progress so
981 # the service is possibly unstable. As above, we
982 # defer any scheduled reconfigured. We also replay
983 # the previous monitor status since that's the best
984 # information we have.
985 ctdb_replay_monitor_status
991 ##################################################################
992 # Does CTDB manage this service? - and associated auto-start/stop
994 ctdb_compat_managed_service ()
996 if [ "$1" = "yes" -a "$2" = "$service_name" ] ; then
997 CTDB_MANAGED_SERVICES="$CTDB_MANAGED_SERVICES $2"
1001 is_ctdb_managed_service ()
1005 # $t is used just for readability and to allow better accurate
1006 # matching via leading/trailing spaces
1007 t=" $CTDB_MANAGED_SERVICES "
1009 # Return 0 if "<space>$service_name<space>" appears in $t
1010 if [ "${t#* ${service_name} }" != "${t}" ] ; then
1014 # If above didn't match then update $CTDB_MANAGED_SERVICES for
1015 # backward compatibility and try again.
1016 ctdb_compat_managed_service "$CTDB_MANAGES_VSFTPD" "vsftpd"
1017 ctdb_compat_managed_service "$CTDB_MANAGES_SAMBA" "samba"
1018 ctdb_compat_managed_service "$CTDB_MANAGES_WINBIND" "winbind"
1019 ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "apache2"
1020 ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "httpd"
1021 ctdb_compat_managed_service "$CTDB_MANAGES_ISCSI" "iscsi"
1022 ctdb_compat_managed_service "$CTDB_MANAGES_CLAMD" "clamd"
1023 ctdb_compat_managed_service "$CTDB_MANAGES_NFS" "nfs"
1025 t=" $CTDB_MANAGED_SERVICES "
1027 # Return 0 if "<space>$service_name<space>" appears in $t
1028 [ "${t#* ${service_name} }" != "${t}" ]
1031 ctdb_start_stop_service ()
1035 # Allow service-start/service-stop pseudo-events to start/stop
1036 # services when we're not auto-starting/stopping and we're not
1038 case "$event_name" in
1040 if is_ctdb_managed_service ; then
1041 die 'service-start event not permitted when service is managed'
1043 if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
1044 die 'service-start event not permitted with CTDB_SERVICE_AUTOSTARTSTOP=yes'
1050 if is_ctdb_managed_service ; then
1051 die 'service-stop event not permitted when service is managed'
1053 if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
1054 die 'service-stop event not permitted with CTDB_SERVICE_AUTOSTARTSTOP=yes'
1061 # Do nothing unless configured to...
1062 [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] || return 0
1064 [ "$event_name" = "monitor" ] || return 0
1066 if is_ctdb_managed_service ; then
1067 if ! is_ctdb_previously_managed_service ; then
1068 echo "Starting service \"$service_name\" - now managed"
1069 background_with_logging ctdb_service_start
1073 if is_ctdb_previously_managed_service ; then
1074 echo "Stopping service \"$service_name\" - no longer managed"
1075 background_with_logging ctdb_service_stop
1081 ctdb_service_start ()
1083 # The service is marked managed if we've ever tried to start it.
1084 ctdb_service_managed
1086 service_start || return $?
1092 ctdb_service_stop ()
1094 ctdb_service_unmanaged
1098 # Default service_start() and service_stop() functions.
1100 # These may be overridden in an eventscript.
1103 service "$service_name" start
1108 service "$service_name" stop
1111 ##################################################################
1113 # This exists only for backward compatibility with 3rd party scripts
1115 ctdb_standard_event_handler ()
1122 _family="$1" ; shift
1123 if [ "$_family" = "inet6" ] ; then
1124 _iptables_cmd="ip6tables"
1126 _iptables_cmd="iptables"
1129 # iptables doesn't like being re-entered, so flock-wrap it.
1130 flock -w 30 "${CTDB_SCRIPT_VARDIR}/iptables.flock" "$_iptables_cmd" "$@"
1133 # AIX (and perhaps others?) doesn't have mktemp
1134 if ! type mktemp >/dev/null 2>&1 ; then
1138 if [ "$1" = "-d" ] ; then
1142 _d="${TMPDIR:-/tmp}"
1143 _hex10=$(dd if=/dev/urandom count=20 2>/dev/null | \
1145 sed -e 's@\(..........\).*@\1@')
1146 _t="${_d}/tmp.${_hex10}"
1159 ######################################################################
1160 # NFS callout handling
1164 if [ -z "$CTDB_NFS_CALLOUT" ] ; then
1165 CTDB_NFS_CALLOUT="${CTDB_BASE}/nfs-linux-kernel-callout"
1167 # Always export, for statd callout
1168 export CTDB_NFS_CALLOUT
1170 # If the callout wants to use this then it must create it
1171 export CTDB_NFS_CALLOUT_STATE_DIR="${service_state_dir}/callout-state"
1173 # Export, if set, for use by clustered NFS callouts
1174 if [ -n "$CTDB_NFS_STATE_FS_TYPE" ] ; then
1175 export CTDB_NFS_STATE_FS_TYPE
1177 if [ -n "$CTDB_NFS_STATE_MNT" ] ; then
1178 export CTDB_NFS_STATE_MNT
1181 nfs_callout_cache="${service_state_dir}/nfs_callout_cache"
1182 nfs_callout_cache_callout="${nfs_callout_cache}/CTDB_NFS_CALLOUT"
1183 nfs_callout_cache_ops="${nfs_callout_cache}/ops"
1186 nfs_callout_register ()
1188 mkdir -p "$nfs_callout_cache_ops"
1189 rm -f "$nfs_callout_cache_ops"/*
1191 echo "$CTDB_NFS_CALLOUT" >"$nfs_callout_cache_callout"
1193 _t=$(eval "$CTDB_NFS_CALLOUT" "register")
1194 if [ -n "$_t" ] ; then
1196 while IFS="" read _op ; do
1197 touch "${nfs_callout_cache_ops}/${_op}"
1200 touch "${nfs_callout_cache_ops}/ALL"
1206 # Re-run registration if $CTDB_NFS_CALLOUT has changed
1208 if [ -r "$nfs_callout_cache_callout" ] ; then
1209 read _prev <"$nfs_callout_cache_callout"
1211 if [ "$CTDB_NFS_CALLOUT" != "$_prev" ] ; then
1212 nfs_callout_register
1215 # Run the operation if it is registered...
1216 if [ -e "${nfs_callout_cache_ops}/${1}" ] || \
1217 [ -e "${nfs_callout_cache_ops}/ALL" ]; then
1218 eval "$CTDB_NFS_CALLOUT" "$@"
1222 ########################################################
1224 ########################################################
1230 tickledir="${CTDB_SCRIPT_VARDIR}/tickles"
1231 mkdir -p "$tickledir"
1235 # What public IPs do I hold?
1236 _ips=$($CTDB -X ip | awk -F'|' -v pnn="$pnn" '$3 == pnn {print $2}')
1238 # IPs and port as ss filters
1240 for _ip in $_ips ; do
1241 _ip_filter="${_ip_filter}${_ip_filter:+ || }src [${_ip}]"
1243 _port_filter="sport == :${_port}"
1245 # Record connections to our public IPs in a temporary file.
1246 # This temporary file is in CTDB's private state directory and
1247 # $$ is used to avoid a very rare race involving CTDB's script
1248 # debugging. No security issue, nothing to see here...
1249 _my_connections="${tickledir}/${_port}.connections.$$"
1250 # Parentheses are needed around the filters for precedence but
1251 # the parentheses can't be empty!
1252 ss -tn state established \
1253 "${_ip_filter:+( ${_ip_filter} )}" \
1254 "${_port_filter:+( ${_port_filter} )}" |
1255 awk 'NR > 1 {print $4, $3}' |
1256 sort >"$_my_connections"
1258 # Record our current tickles in a temporary file
1259 _my_tickles="${tickledir}/${_port}.tickles.$$"
1260 for _i in $_ips ; do
1261 $CTDB -X gettickles "$_i" "$_port" |
1262 awk -F'|' 'NR > 1 { printf "%s:%s %s:%s\n", $2, $3, $4, $5 }'
1264 sort >"$_my_tickles"
1266 # Add tickles for connections that we haven't already got tickles for
1267 comm -23 "$_my_connections" "$_my_tickles" |
1268 while read _src _dst ; do
1269 $CTDB addtickle "$_src" "$_dst"
1272 # Remove tickles for connections that are no longer there
1273 comm -13 "$_my_connections" "$_my_tickles" |
1274 while read _src _dst ; do
1275 $CTDB deltickle "$_src" "$_dst"
1278 rm -f "$_my_connections" "$_my_tickles"
1280 # Remove stale files from killed scripts
1281 find "$tickledir" -type f -mmin +10 | xargs -r rm
1284 ########################################################
1285 # load a site local config file
1286 ########################################################
1288 [ -n "$CTDB_RC_LOCAL" -a -x "$CTDB_RC_LOCAL" ] && {
1292 [ -x "${CTDB_BASE}/rc.local" ] && {
1293 . "${CTDB_BASE}/rc.local"
1296 [ -d "${CTDB_BASE}/rc.local.d" ] && {
1297 for i in "${CTDB_BASE}/rc.local.d"/* ; do
1298 [ -x "$i" ] && . "$i"
1302 script_name="${0##*/}" # basename