1 # Hey Emacs, this is a -*- shell-script -*- !!!
3 # utility functions for ctdb event scripts
5 if [ -z "$CTDB_BASE" ]; then
6 echo 'CTDB_BASE unset in CTDB functions file'
11 # CTDB_VARDIR is used elsewhere
12 # shellcheck disable=SC2034
13 CTDB_VARDIR="/usr/local/var/lib/ctdb"
15 CTDB="${CTDB:-/usr/local/bin/ctdb}"
17 # Only (and always) override these variables in test code
19 if [ -z "$CTDB_SCRIPT_VARDIR" ]; then
20 CTDB_SCRIPT_VARDIR="/usr/local/var/lib/ctdb/scripts"
23 if [ -z "$CTDB_SYS_ETCDIR" ]; then
24 CTDB_SYS_ETCDIR="/etc"
27 if [ -z "$CTDB_HELPER_BINDIR" ]; then
28 CTDB_HELPER_BINDIR="/usr/local/libexec/ctdb"
31 #######################################
32 # pull in a system config file, if any
38 if [ -f "${CTDB_SYS_ETCDIR}/sysconfig/${_i}" ]; then
39 . "${CTDB_SYS_ETCDIR}/sysconfig/${_i}"
41 elif [ -f "${CTDB_SYS_ETCDIR}/default/${_i}" ]; then
42 . "${CTDB_SYS_ETCDIR}/default/${_i}"
48 # load_script_options [ component script ]
49 # script is an event script name relative to a component
50 # component is currently ignored
55 elif [ $# -eq 0 ]; then
58 die "usage: load_script_options [ component script ]"
61 _options="${CTDB_BASE}/script.options"
63 if [ -r "$_options" ]; then
67 if [ -n "$_script" ]; then
68 _s="${CTDB_BASE}/events/legacy/${_script}"
72 _options="${_s}.options"
74 if [ -r "$_options" ]; then
79 ##############################################################
90 # Log given message or stdin to either syslog or a CTDB log file
91 # $1 is the tag passed to logger if syslog is in use.
97 case "$CTDB_LOGGING" in
99 if [ -n "$CTDB_LOGGING" ]; then
100 _file="${CTDB_LOGGING#file:}"
102 _file="/usr/local/var/log/log.ctdb"
113 # Handle all syslog:* variants here too. There's no tool to do
114 # the lossy things, so just use logger.
115 logger -t "ctdbd: ${_tag}" "$@"
120 # When things are run in the background in an eventscript then logging
121 # output might get lost. This is the "solution". :-)
122 background_with_logging()
125 "$@" 2>&1 </dev/null |
126 script_log "${script_name}&"
132 ##############################################################
133 # check number of args for different events
139 echo "ERROR: must supply interface, IP and maskbits"
145 echo "ERROR: must supply old interface, new interface, IP and maskbits"
152 ##############################################################
153 # determine on what type of system (init style) we are running
156 # only do detection if not already set:
157 if [ -n "$CTDB_INIT_STYLE" ]; then
161 if [ -x /sbin/startproc ]; then
162 CTDB_INIT_STYLE="suse"
163 elif [ -x /sbin/start-stop-daemon ]; then
164 CTDB_INIT_STYLE="debian"
166 CTDB_INIT_STYLE="redhat"
170 ######################################################
171 # simulate /sbin/service on platforms that don't have it
172 # _service() makes it easier to hook the service() function for
179 # do nothing, when no service was specified
180 [ -z "$_service_name" ] && return
182 if [ -x /sbin/service ]; then
183 $_nice /sbin/service "$_service_name" "$_op"
184 elif [ -x /usr/sbin/service ]; then
185 $_nice /usr/sbin/service "$_service_name" "$_op"
186 elif [ -x /bin/systemctl ]; then
187 $_nice /bin/systemctl "$_op" "$_service_name"
188 elif [ -x "${CTDB_SYS_ETCDIR}/init.d/${_service_name}" ]; then
189 $_nice "${CTDB_SYS_ETCDIR}/init.d/${_service_name}" "$_op"
190 elif [ -x "${CTDB_SYS_ETCDIR}/rc.d/init.d/${_service_name}" ]; then
191 $_nice "${CTDB_SYS_ETCDIR}/rc.d/init.d/${_service_name}" "$_op"
201 ######################################################
202 # simulate /sbin/service (niced) on platforms that don't have it
209 ######################################################
210 # Cached retrieval of PNN from local node. This never changes so why
211 # open a client connection to the server each time this is needed?
214 _pnn_file="${CTDB_SCRIPT_VARDIR}/my-pnn"
215 if [ ! -f "$_pnn_file" ]; then
216 $CTDB pnn >"$_pnn_file"
222 # Cached retrieval of private IP address from local node. This never
224 ctdb_get_ip_address()
226 _ip_addr_file="${CTDB_SCRIPT_VARDIR}/my-ip-address"
227 if [ ! -f "$_ip_addr_file" ]; then
228 $CTDB -X nodestatus |
229 awk -F '|' 'NR == 2 { print $3 }' >"$_ip_addr_file"
235 # Cached retrieval of database options for use by event scripts.
237 # If the variables are already set then they should not be overwritten
238 # - this should only happen during event script testing.
239 ctdb_get_db_options()
241 _db_opts_file="${CTDB_SCRIPT_VARDIR}/db_options.cache"
243 if [ ! -f "$_db_opts_file" ]; then
245 ctdb_translate_option "database" \
246 "volatile database directory" \
248 ctdb_translate_option "database" \
249 "persistent database directory" \
250 "CTDB_DBDIR_PERSISTENT"
251 ctdb_translate_option "database" \
252 "state database directory" \
260 ctdb_translate_option()
266 # ctdb-config already prints an error if something goes wrong
267 _t=$("${CTDB_HELPER_BINDIR}/ctdb-config" get "$_section" "$_opt") ||
269 echo "${_variable}=\"${_t}\""
272 ######################################################
273 # wrapper around /proc/ settings to allow them to be hooked
275 # 1st arg is relative path under /proc/, 2nd arg is value to set
278 echo "$2" >"/proc/$1"
283 if [ -w "/proc/$1" ]; then
288 ######################################################
289 # wrapper around getting file contents from /proc/ to allow
290 # this to be hooked for testing
291 # 1st arg is relative path under /proc/
297 ######################################################
298 # Print up to $_max kernel stack traces for processes named $_program
299 program_stack_traces()
305 for _pid in $(pidof "$_prog"); do
306 [ "$_count" -le "$_max" ] || break
308 # Do this first to avoid racing with process exit
309 _stack=$(get_proc "${_pid}/stack" 2>/dev/null)
310 if [ -n "$_stack" ]; then
311 echo "Stack trace for ${_prog}[${_pid}]:"
313 _count=$((_count + 1))
318 ######################################################
319 # Ensure $service_name is set
320 assert_service_name()
322 # service_name is set by the event script
323 # shellcheck disable=SC2154
324 [ -n "$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 -r d; do
340 [ -d "${d}/." ] || return 1
346 ######################################################
347 # check a set of directories is available
348 # directories are read from stdin
349 ######################################################
350 ctdb_check_directories()
352 ctdb_check_directories_probe || {
353 echo "ERROR: $service_name directory \"$d\" not available"
358 ######################################################
359 # check a set of tcp ports
360 # usage: ctdb_check_tcp_ports <ports...>
361 ######################################################
363 # Check whether something is listening on all of the given TCP ports
364 # using the "ctdb checktcpport" command.
365 ctdb_check_tcp_ports()
368 echo "INTERNAL ERROR: ctdb_check_tcp_ports - no ports specified"
372 for _p; do # process each function argument (port)
373 _cmd="$CTDB checktcpport $_p"
378 echo "$service_name not listening on TCP port $_p"
382 # Couldn't bind, something already listening, next port
386 echo "unexpected error (${_ret}) running \"${_cmd}\""
387 if [ -n "$_out" ]; then
395 # All ports listening
399 ######################################################
400 # check a unix socket
401 # usage: ctdb_check_unix_socket SOCKPATH
402 ######################################################
403 ctdb_check_unix_socket()
407 if [ -z "$_sockpath" ]; then
408 echo "ERROR: ctdb_check_unix_socket() requires socket path"
412 _out=$(ss -l -x "src ${_sockpath}" | tail -n +2)
413 if [ -z "$_out" ]; then
414 echo "ERROR: ${service_name} not listening on ${_sockpath}"
419 ################################################
420 # kill off any TCP connections with the given IP
421 ################################################
422 kill_tcp_connections()
428 if [ "$3" = "oneway" ]; then
432 get_tcp_connections_for_ip "$_ip" | {
437 while read -r _dst _src; do
438 _destport="${_dst##*:}"
441 # we only do one-way killtcp for CIFS
442 139 | 445) __oneway=true ;;
445 _connections="${_connections}${_nl}${_src} ${_dst}"
447 _connections="${_connections}${_nl}${_dst} ${_src}"
450 _killcount=$((_killcount + 1))
453 if [ $_killcount -eq 0 ]; then
457 if [ -n "$CTDB_KILLTCP_DEBUGLEVEL" ]; then
458 _debuglevel="$CTDB_KILLTCP_DEBUGLEVEL"
460 _debuglevel="$CTDB_DEBUGLEVEL"
462 echo "$_connections" |
463 CTDB_DEBUGLEVEL="$_debuglevel" \
464 "${CTDB_HELPER_BINDIR}/ctdb_killtcp" "$_iface" || {
465 echo "Failed to kill TCP connections"
469 _connections=$(get_tcp_connections_for_ip "$_ip")
470 if [ -z "$_connections" ]; then
473 _remaining=$(echo "$_connections" | wc -l)
476 _actually_killed=$((_killcount - _remaining))
478 _t="${_actually_killed}/${_killcount}"
479 echo "Killed ${_t} TCP connections to released IP $_ip"
481 if [ -n "$_connections" ]; then
482 echo "Remaining connections:"
483 echo "$_connections" | sed -e 's|^| |'
488 ##################################################################
489 # kill off the local end for any TCP connections with the given IP
490 ##################################################################
491 kill_tcp_connections_local_only()
493 kill_tcp_connections "$@" "oneway"
496 ##################################################################
497 # tickle any TCP connections with the given IP
498 ##################################################################
499 tickle_tcp_connections()
503 # Get connections, both directions
504 _conns=$(get_tcp_connections_for_ip "$_ip" |
505 awk '{ print $1, $2 ; print $2, $1 }')
507 echo "$_conns" | awk '{ print "Tickle TCP connection", $1, $2 }'
508 echo "$_conns" | ctdb tickle
511 get_tcp_connections_for_ip()
515 ss -tn state established "src [$_ip]" | awk 'NR > 1 {print $3, $4}'
518 ########################################################
526 # Ensure interface is up
527 ip link set "$_iface" up ||
528 die "Failed to bringup interface $_iface"
530 # Only need to define broadcast for IPv4
536 # Intentionally unquoted multi-word value here
537 # shellcheck disable=SC2086
538 ip addr add "$_ip/$_maskbits" $_bcast dev "$_iface" || {
539 echo "Failed to add $_ip/$_maskbits on dev $_iface"
543 # Wait 5 seconds for IPv6 addresses to stop being tentative...
544 if [ -z "$_bcast" ]; then
545 for _x in $(seq 1 10); do
546 ip addr show to "${_ip}/128" | grep -q "tentative" || break
550 # If the address was a duplicate then it won't be on the
551 # interface so flag an error.
552 _t=$(ip addr show to "${_ip}/128")
555 echo "Failed to add $_ip/$_maskbits on dev $_iface"
558 *tentative* | *dadfailed*)
559 echo "Failed to add $_ip/$_maskbits on dev $_iface"
560 ip addr del "$_ip/$_maskbits" dev "$_iface"
567 delete_ip_from_iface()
573 # This could be set globally for all interfaces but it is probably
574 # better to avoid surprises, so limit it the interfaces where CTDB
575 # has public IP addresses. There isn't anywhere else convenient
576 # to do this so just set it each time. This is much cheaper than
577 # remembering and re-adding secondaries.
578 set_proc "sys/net/ipv4/conf/${_iface}/promote_secondaries" 1
580 ip addr del "$_ip/$_maskbits" dev "$_iface" || {
581 echo "Failed to del $_ip on dev $_iface"
586 # If the given IP is hosted then print 2 items: maskbits and iface
595 ip addr show to "${_addr}/${_bits}" 2>/dev/null |
596 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ;
597 sub("@.*", "", iface) }
598 $1 ~ /inet/ { mask = $2; sub(".*/", "", mask);
604 _addr="${1%/*}" # Remove optional maskbits
606 # Intentional word splitting here
607 # shellcheck disable=SC2046
608 set -- $(ip_maskbits_iface "$_addr")
612 echo "Removing public address $_addr/$_maskbits from device $_iface"
613 delete_ip_from_iface "$_iface" "$_addr" "$_maskbits" >/dev/null 2>&1
617 drop_all_public_ips()
619 # _x is intentionally ignored
620 # shellcheck disable=SC2034
621 while read -r _ip _x; do
626 done <"${CTDB_BASE}/public_addresses"
631 set_proc_maybe sys/net/ipv4/route/flush 1
632 set_proc_maybe sys/net/ipv6/route/flush 1
635 ########################################################
636 # Interface monitoring
638 # If the interface is a virtual one (e.g. VLAN) then get the
639 # underlying interface
644 # If $_iface is a VLAN (i.e. contains an '@') then strip every
645 # before the '@', otherwise print the whole interface
649 # Check whether an interface is operational
654 _iface_info=$(ip -br link show "$_iface" 2>&1) || {
655 echo "ERROR: Monitored interface ${_iface} does not exist"
659 # If the interface is a virtual one (e.g. VLAN) then get the
660 # underlying interface.
661 _realiface=$(interface_get_real "${_iface_info%% *}")
663 if _bi=$(get_proc "net/bonding/${_realiface}" 2>/dev/null); then
664 # This is a bond: various monitoring strategies
665 echo "$_bi" | grep -q 'Currently Active Slave: None' && {
666 echo "ERROR: No active slaves for bond device ${_realiface}"
669 echo "$_bi" | grep -q '^MII Status: up' || {
670 echo "ERROR: public network interface ${_realiface} is down"
673 echo "$_bi" | grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' && {
674 # This works around a bug in the driver where the
675 # overall bond status can be up but none of the actual
676 # physical interfaces have a link.
677 echo "$_bi" | grep 'MII Status:' | tail -n +2 | grep -q '^MII Status: up' || {
678 echo "ERROR: No active slaves for 802.ad bond device ${_realiface}"
688 # loopback is always working
692 # we don't know how to test ib links
696 ethtool "$_iface" | grep -q 'Link detected: yes' || {
697 # On some systems, this is not successful when a
698 # cable is plugged but the interface has not been
699 # brought up previously. Bring the interface up
701 ip link set "$_iface" up
702 ethtool "$_iface" | grep -q 'Link detected: yes' || {
703 echo "ERROR: No link on the public network interface ${_iface}"
713 ########################################################
715 _ctdb_counter_common()
717 [ $# -le 1 ] || die "usage: _ctdb_counter_common [name]"
719 if [ $# -eq 1 ]; then
720 _counter_name="${1}.failcount"
722 _counter_name="failcount"
725 if [ -z "$script_state_dir" ]; then
726 die "ctdb_counter_* functions need ctdb_setup_state_dir()"
729 _counter_file="${script_state_dir}/${_counter_name}"
731 # Some code passes an argument
732 # shellcheck disable=SC2120
735 _ctdb_counter_common "$1"
741 _ctdb_counter_common "$1"
743 # unary counting using newlines!
744 echo >>"$_counter_file"
748 _ctdb_counter_common "$1"
750 _val=$(wc -c <"$_counter_file" 2>/dev/null || echo 0)
751 # Strip leading spaces from output of wc (on freebsd)
752 # shellcheck disable=SC2086
756 ########################################################
758 # ctdb_setup_state_dir <type> <name>
759 # Sets/creates script_state_dir)
760 ctdb_setup_state_dir()
762 [ $# -eq 2 ] || die "usage: ctdb_setup_state_dir <type> <name>"
767 script_state_dir="${CTDB_SCRIPT_VARDIR}/${_type}/${_name}"
769 mkdir -p "$script_state_dir" ||
770 die "Error creating script state dir \"${script_state_dir}\""
773 ##################################################################
774 # Reconfigure a service on demand
776 _ctdb_service_reconfigure_common()
778 if [ -z "$script_state_dir" ]; then
779 die "ctdb_service_*_reconfigure() needs ctdb_setup_state_dir()"
782 _ctdb_service_reconfigure_flag="${script_state_dir}/need_reconfigure"
785 ctdb_service_needs_reconfigure()
787 _ctdb_service_reconfigure_common
788 [ -e "$_ctdb_service_reconfigure_flag" ]
791 ctdb_service_set_reconfigure()
793 _ctdb_service_reconfigure_common
794 : >"$_ctdb_service_reconfigure_flag"
797 ctdb_service_unset_reconfigure()
799 _ctdb_service_reconfigure_common
800 rm -f "$_ctdb_service_reconfigure_flag"
803 ctdb_service_reconfigure()
805 echo "Reconfiguring service \"${service_name}\"..."
806 ctdb_service_unset_reconfigure
807 service_reconfigure || return $?
808 # Intentionally have this use $service_name as default
809 # shellcheck disable=SC2119
813 # Default service_reconfigure() function does nothing.
814 service_reconfigure()
819 # Default service_start() and service_stop() functions.
821 # These may be overridden in an eventscript.
824 service "$service_name" start
829 service "$service_name" stop
832 ##################################################################
834 # This exists only for backward compatibility with 3rd party scripts
836 ctdb_standard_event_handler()
845 if [ "$_family" = "inet6" ]; then
846 _iptables_cmd="ip6tables"
848 _iptables_cmd="iptables"
851 # iptables doesn't like being re-entered, so flock-wrap it.
852 flock -w 30 "${CTDB_SCRIPT_VARDIR}/iptables.flock" "$_iptables_cmd" "$@"
855 # AIX (and perhaps others?) doesn't have mktemp
856 # type is commonly supported and more portable than which(1)
857 # shellcheck disable=SC2039
858 if ! type mktemp >/dev/null 2>&1; then
862 if [ "$1" = "-d" ]; then
867 _hex10=$(dd if=/dev/urandom count=20 2>/dev/null |
870 _t="${_d}/tmp.${_hex10}"
883 ######################################################################
884 # NFS callout handling
890 if [ -z "$CTDB_NFS_CALLOUT" ]; then
891 CTDB_NFS_CALLOUT="${CTDB_BASE}/nfs-linux-kernel-callout"
893 # Always export, for statd callout
894 export CTDB_NFS_CALLOUT
896 # If the callout wants to use this then it must create it
897 export CTDB_NFS_CALLOUT_STATE_DIR="${_state_dir}/callout-state"
899 # Export, if set, for use by clustered NFS callouts
900 if [ -n "$CTDB_NFS_STATE_FS_TYPE" ]; then
901 export CTDB_NFS_STATE_FS_TYPE
903 if [ -n "$CTDB_NFS_STATE_MNT" ]; then
904 export CTDB_NFS_STATE_MNT
907 nfs_callout_cache="${_state_dir}/nfs_callout_cache"
908 nfs_callout_cache_callout="${nfs_callout_cache}/CTDB_NFS_CALLOUT"
909 nfs_callout_cache_ops="${nfs_callout_cache}/ops"
912 nfs_callout_register()
914 mkdir -p "$nfs_callout_cache_ops"
915 rm -f "$nfs_callout_cache_ops"/*
917 echo "$CTDB_NFS_CALLOUT" >"$nfs_callout_cache_callout"
919 _t=$("$CTDB_NFS_CALLOUT" "register")
920 if [ -n "$_t" ]; then
922 while IFS="" read -r _op; do
923 touch "${nfs_callout_cache_ops}/${_op}"
926 touch "${nfs_callout_cache_ops}/ALL"
932 # Re-run registration if $CTDB_NFS_CALLOUT has changed
934 if [ -r "$nfs_callout_cache_callout" ]; then
935 read -r _prev <"$nfs_callout_cache_callout"
937 if [ "$CTDB_NFS_CALLOUT" != "$_prev" ]; then
941 # Run the operation if it is registered...
942 if [ -e "${nfs_callout_cache_ops}/${1}" ] ||
943 [ -e "${nfs_callout_cache_ops}/ALL" ]; then
944 "$CTDB_NFS_CALLOUT" "$@"
948 ########################################################
950 ########################################################
956 tickledir="${CTDB_SCRIPT_VARDIR}/tickles"
957 mkdir -p "$tickledir"
959 # What public IPs do I hold?
961 _ips=$($CTDB -X ip | awk -F'|' -v pnn="$_pnn" '$3 == pnn {print $2}')
963 # IPs and port as ss filters
966 _ip_filter="${_ip_filter}${_ip_filter:+ || }src [${_ip}]"
968 _port_filter="sport == :${_port}"
970 # Record connections to our public IPs in a temporary file.
971 # This temporary file is in CTDB's private state directory and
972 # $$ is used to avoid a very rare race involving CTDB's script
973 # debugging. No security issue, nothing to see here...
974 _my_connections="${tickledir}/${_port}.connections.$$"
975 # Parentheses are needed around the filters for precedence but
976 # the parentheses can't be empty!
978 # Recent versions of ss print square brackets around IPv6
979 # addresses. While it is desirable to update CTDB's address
980 # parsing and printing code, something needs to be done here
981 # for backward compatibility, so just delete the brackets.
982 ss -tn state established \
983 "${_ip_filter:+( ${_ip_filter} )}" \
984 "${_port_filter:+( ${_port_filter} )}" |
985 awk 'NR > 1 {print $4, $3}' |
987 sort >"$_my_connections"
989 # Record our current tickles in a temporary file
990 _my_tickles="${tickledir}/${_port}.tickles.$$"
992 $CTDB -X gettickles "$_i" "$_port" |
993 awk -F'|' 'NR > 1 { printf "%s:%s %s:%s\n", $2, $3, $4, $5 }'
997 # Add tickles for connections that we haven't already got tickles for
998 comm -23 "$_my_connections" "$_my_tickles" |
1001 # Remove tickles for connections that are no longer there
1002 comm -13 "$_my_connections" "$_my_tickles" |
1005 rm -f "$_my_connections" "$_my_tickles"
1007 # Remove stale files from killed scripts
1008 # Files can't have spaces in name, more portable than -print0/-0
1009 # shellcheck disable=SC2038
1010 (cd "$tickledir" && find . -type f -mmin +10 | xargs -r rm)
1013 ########################################################
1014 # load a site local config file
1015 ########################################################
1017 [ -x "${CTDB_BASE}/rc.local" ] && {
1018 . "${CTDB_BASE}/rc.local"
1021 [ -d "${CTDB_BASE}/rc.local.d" ] && {
1022 for i in "${CTDB_BASE}/rc.local.d"/*; do
1023 [ -x "$i" ] && . "$i"
1027 script_name="${0##*/}" # basename