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
51 load_script_options ()
53 if [ $# -eq 2 ] ; then
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 # $1 is the tag passed to logger if syslog is in use.
96 case "$CTDB_LOGGING" in
98 if [ -n "$CTDB_LOGGING" ] ; then
99 _file="${CTDB_LOGGING#file:}"
101 _file="/usr/local/var/log/log.ctdb"
104 if [ -n "$*" ] ; then
112 # Handle all syslog:* variants here too. There's no tool to do
113 # the lossy things, so just use logger.
114 logger -t "ctdbd: ${_tag}" "$@"
119 # When things are run in the background in an eventscript then logging
120 # output might get lost. This is the "solution". :-)
121 background_with_logging ()
124 "$@" 2>&1 </dev/null |
125 script_log "${script_name}&"
131 ##############################################################
132 # check number of args for different events
138 echo "ERROR: must supply interface, IP and maskbits"
144 echo "ERROR: must supply old interface, new interface, IP and maskbits"
151 ##############################################################
152 # determine on what type of system (init style) we are running
155 # only do detection if not already set:
156 if [ -n "$CTDB_INIT_STYLE" ] ; then
160 if [ -x /sbin/startproc ]; then
161 CTDB_INIT_STYLE="suse"
162 elif [ -x /sbin/start-stop-daemon ]; then
163 CTDB_INIT_STYLE="debian"
165 CTDB_INIT_STYLE="redhat"
169 ######################################################
170 # simulate /sbin/service on platforms that don't have it
171 # _service() makes it easier to hook the service() function for
178 # do nothing, when no service was specified
179 [ -z "$_service_name" ] && return
181 if [ -x /sbin/service ]; then
182 $_nice /sbin/service "$_service_name" "$_op"
183 elif [ -x /usr/sbin/service ]; then
184 $_nice /usr/sbin/service "$_service_name" "$_op"
185 elif [ -x /bin/systemctl ]; then
186 $_nice /bin/systemctl "$_op" "$_service_name"
187 elif [ -x "${CTDB_SYS_ETCDIR}/init.d/${_service_name}" ]; then
188 $_nice "${CTDB_SYS_ETCDIR}/init.d/${_service_name}" "$_op"
189 elif [ -x "${CTDB_SYS_ETCDIR}/rc.d/init.d/${_service_name}" ]; then
190 $_nice "${CTDB_SYS_ETCDIR}/rc.d/init.d/${_service_name}" "$_op"
200 ######################################################
201 # simulate /sbin/service (niced) on platforms that don't have it
208 ######################################################
209 # Cached retrieval of PNN from local node. This never changes so why
210 # open a client connection to the server each time this is needed?
213 _pnn_file="${CTDB_SCRIPT_VARDIR}/my-pnn"
214 if [ ! -f "$_pnn_file" ] ; then
215 $CTDB pnn >"$_pnn_file"
221 # Cached retrieval of private IP address from local node. This never
223 ctdb_get_ip_address ()
225 _ip_addr_file="${CTDB_SCRIPT_VARDIR}/my-ip-address"
226 if [ ! -f "$_ip_addr_file" ] ; then
227 $CTDB -X nodestatus |
228 awk -F '|' 'NR == 2 { print $3 }' >"$_ip_addr_file"
231 # ip_address is used by caller
232 # shellcheck disable=SC2034
236 # Cached retrieval of database options for use by event scripts.
238 # If the variables are already set then they should not be overwritten
239 # - this should only happen during event script testing.
240 ctdb_get_db_options ()
242 _db_opts_file="${CTDB_SCRIPT_VARDIR}/db_options.cache"
244 if [ ! -f "$_db_opts_file" ] ; then
246 ctdb_translate_option "database" \
247 "volatile database directory" \
249 ctdb_translate_option "database" \
250 "persistent database directory" \
251 "CTDB_DBDIR_PERSISTENT"
252 ctdb_translate_option "database" \
253 "state database directory" \
261 ctdb_translate_option ()
267 # ctdb-config already prints an error if something goes wrong
268 _t=$("${CTDB_HELPER_BINDIR}/ctdb-config" get "$_section" "$_opt") || \
270 echo "${_variable}=\"${_t}\""
273 ######################################################
274 # wrapper around /proc/ settings to allow them to be hooked
276 # 1st arg is relative path under /proc/, 2nd arg is value to set
279 echo "$2" >"/proc/$1"
284 if [ -w "/proc/$1" ] ; then
289 ######################################################
290 # wrapper around getting file contents from /proc/ to allow
291 # this to be hooked for testing
292 # 1st arg is relative path under /proc/
298 ######################################################
299 # Print up to $_max kernel stack traces for processes named $_program
300 program_stack_traces ()
306 for _pid in $(pidof "$_prog") ; do
307 [ "$_count" -le "$_max" ] || break
309 # Do this first to avoid racing with process exit
310 _stack=$(get_proc "${_pid}/stack" 2>/dev/null)
311 if [ -n "$_stack" ] ; then
312 echo "Stack trace for ${_prog}[${_pid}]:"
314 _count=$((_count + 1))
319 ######################################################
320 # Ensure $service_name is set
321 assert_service_name ()
323 # service_name is set by the event script
324 # shellcheck disable=SC2154
325 [ -n "$service_name" ] || die "INTERNAL ERROR: \$service_name not set"
328 ######################################################
329 # check a set of directories is available
330 # return 1 on a missing directory
331 # directories are read from stdin
332 ######################################################
333 ctdb_check_directories_probe()
335 while IFS="" read d ; do
341 [ -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()
367 if [ -z "$1" ] ; then
368 echo "INTERNAL ERROR: ctdb_check_tcp_ports - no ports specified"
372 for _p ; do # process each function argument (port)
373 _cmd="$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 _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}"
446 if ! $__oneway ; then
447 _connections="${_connections}${_nl}${_dst} ${_src}"
450 _killcount=$((_killcount + 1))
453 if [ $_killcount -eq 0 ] ; then
457 echo "$_connections" | \
458 "${CTDB_HELPER_BINDIR}/ctdb_killtcp" "$_iface" || {
459 echo "Failed to kill TCP connections"
463 _connections=$(get_tcp_connections_for_ip "$_ip")
464 if [ -z "$_connections" ] ; then
467 _remaining=$(echo "$_connections" | wc -l)
470 _actually_killed=$((_killcount - _remaining))
472 _t="${_actually_killed}/${_killcount}"
473 echo "Killed ${_t} TCP connections to released IP $_ip"
475 if [ -n "$_connections" ] ; then
476 echo "Remaining connections:"
477 echo "$_connections" | sed -e 's|^| |'
482 ##################################################################
483 # kill off the local end for any TCP connections with the given IP
484 ##################################################################
485 kill_tcp_connections_local_only ()
487 kill_tcp_connections "$@" "oneway"
490 ##################################################################
491 # tickle any TCP connections with the given IP
492 ##################################################################
493 tickle_tcp_connections ()
497 # Get connections, both directions
498 _conns=$(get_tcp_connections_for_ip "$_ip" | \
499 awk '{ print $1, $2 ; print $2, $1 }')
501 echo "$_conns" | awk '{ print "Tickle TCP connection", $1, $2 }'
502 echo "$_conns" | ctdb tickle
505 get_tcp_connections_for_ip ()
509 ss -tn state established "src [$_ip]" | awk 'NR > 1 {print $3, $4}'
512 ########################################################
520 # Ensure interface is up
521 ip link set "$_iface" up || \
522 die "Failed to bringup interface $_iface"
524 # Only need to define broadcast for IPv4
530 # Intentionally unquoted multi-word value here
531 # shellcheck disable=SC2086
532 ip addr add "$_ip/$_maskbits" $_bcast dev "$_iface" || {
533 echo "Failed to add $_ip/$_maskbits on dev $_iface"
537 # Wait 5 seconds for IPv6 addresses to stop being tentative...
538 if [ -z "$_bcast" ] ; then
539 for _x in $(seq 1 10) ; do
540 ip addr show to "${_ip}/128" | grep -q "tentative" || break
544 # If the address was a duplicate then it won't be on the
545 # interface so flag an error.
546 _t=$(ip addr show to "${_ip}/128")
549 echo "Failed to add $_ip/$_maskbits on dev $_iface"
552 *tentative*|*dadfailed*)
553 echo "Failed to add $_ip/$_maskbits on dev $_iface"
554 ip addr del "$_ip/$_maskbits" dev "$_iface"
561 delete_ip_from_iface()
567 # This could be set globally for all interfaces but it is probably
568 # better to avoid surprises, so limit it the interfaces where CTDB
569 # has public IP addresses. There isn't anywhere else convenient
570 # to do this so just set it each time. This is much cheaper than
571 # remembering and re-adding secondaries.
572 set_proc "sys/net/ipv4/conf/${_iface}/promote_secondaries" 1
574 ip addr del "$_ip/$_maskbits" dev "$_iface" || {
575 echo "Failed to del $_ip on dev $_iface"
580 # If the given IP is hosted then print 2 items: maskbits and iface
589 ip addr show to "${_addr}/${_bits}" 2>/dev/null | \
590 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ;
591 sub("@.*", "", iface) }
592 $1 ~ /inet/ { mask = $2; sub(".*/", "", mask);
598 _addr="${1%/*}" # Remove optional maskbits
600 # Intentional word splitting here
601 # shellcheck disable=SC2046
602 set -- $(ip_maskbits_iface "$_addr")
603 if [ -n "$1" ] ; then
606 echo "Removing public address $_addr/$_maskbits from device $_iface"
607 delete_ip_from_iface "$_iface" "$_addr" "$_maskbits" >/dev/null 2>&1
611 drop_all_public_ips ()
613 # _x is intentionally ignored
614 # shellcheck disable=SC2034
615 while read _ip _x ; do
617 done <"${CTDB_BASE}/public_addresses"
622 set_proc_maybe sys/net/ipv4/route/flush 1
623 set_proc_maybe sys/net/ipv6/route/flush 1
626 ########################################################
627 # Interface monitoring
629 # If the interface is a virtual one (e.g. VLAN) then get the
630 # underlying interface
631 interface_get_real ()
633 # Output of "ip link show <iface>"
636 # Extract the full interface description to see if it is a VLAN
637 _t=$(echo "$_iface_info" |
638 awk 'NR == 1 { iface = $2; sub(":$", "", iface) ;
642 # VLAN: use the underlying interface, after the '@'
646 # Not a regular VLAN. For backward compatibility, assume
647 # there is some other sort of VLAN that doesn't have the
648 # '@' in the output and only use what is before a '.'. If
649 # there is no '.' then this will be the whole interface
655 # Check whether an interface is operational
660 _iface_info=$(ip link show "$_iface" 2>&1) || {
661 echo "ERROR: Monitored interface ${_iface} does not exist"
666 # If the interface is a virtual one (e.g. VLAN) then get the
667 # underlying interface.
668 _realiface=$(interface_get_real "$_iface_info")
670 if _bi=$(get_proc "net/bonding/${_realiface}" 2>/dev/null) ; then
671 # This is a bond: various monitoring strategies
672 echo "$_bi" | grep -q 'Currently Active Slave: None' && {
673 echo "ERROR: No active slaves for bond device ${_realiface}"
676 echo "$_bi" | grep -q '^MII Status: up' || {
677 echo "ERROR: public network interface ${_realiface} is down"
680 echo "$_bi" | grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' && {
681 # This works around a bug in the driver where the
682 # overall bond status can be up but none of the actual
683 # physical interfaces have a link.
684 echo "$_bi" | grep 'MII Status:' | tail -n +2 | grep -q '^MII Status: up' || {
685 echo "ERROR: No active slaves for 802.ad bond device ${_realiface}"
695 # loopback is always working
699 # we don't know how to test ib links
703 ethtool "$_iface" | grep -q 'Link detected: yes' || {
704 # On some systems, this is not successful when a
705 # cable is plugged but the interface has not been
706 # brought up previously. Bring the interface up
708 ip link set "$_iface" up
709 ethtool "$_iface" | grep -q 'Link detected: yes' || {
710 echo "ERROR: No link on the public network interface ${_iface}"
720 ########################################################
722 _ctdb_counter_common ()
724 [ $# -le 1 ] || die "usage: _ctdb_counter_common [name]"
726 if [ $# -eq 1 ] ; then
727 _counter_name="${1}.failcount"
729 _counter_name="failcount"
732 if [ -z "$script_state_dir" ] ; then
733 die "ctdb_counter_* functions need ctdb_setup_state_dir()"
736 _counter_file="${script_state_dir}/${_counter_name}"
738 # Some code passes an argument
739 # shellcheck disable=SC2120
740 ctdb_counter_init () {
741 _ctdb_counter_common "$1"
745 ctdb_counter_incr () {
746 _ctdb_counter_common "$1"
748 # unary counting using newlines!
749 echo >>"$_counter_file"
751 ctdb_counter_get () {
752 _ctdb_counter_common "$1"
754 _val=$(wc -c < "$_counter_file" 2>/dev/null || echo 0)
755 # Strip leading spaces from ouput of wc (on freebsd)
756 # shellcheck disable=SC2086
760 ########################################################
762 # ctdb_setup_state_dir <type> <name>
763 # Sets/creates script_state_dir)
764 ctdb_setup_state_dir ()
766 [ $# -eq 2 ] || die "usage: ctdb_setup_state_dir <type> <name>"
771 script_state_dir="${CTDB_SCRIPT_VARDIR}/${_type}/${_name}"
773 mkdir -p "$script_state_dir" || \
774 die "Error creating script state dir \"${script_state_dir}\""
777 ##################################################################
778 # Reconfigure a service on demand
780 _ctdb_service_reconfigure_common ()
782 if [ -z "$script_state_dir" ] ; then
783 die "ctdb_service_*_reconfigure() needs ctdb_setup_state_dir()"
786 _ctdb_service_reconfigure_flag="${script_state_dir}/need_reconfigure"
789 ctdb_service_needs_reconfigure ()
791 _ctdb_service_reconfigure_common
792 [ -e "$_ctdb_service_reconfigure_flag" ]
795 ctdb_service_set_reconfigure ()
797 _ctdb_service_reconfigure_common
798 : >"$_ctdb_service_reconfigure_flag"
801 ctdb_service_unset_reconfigure ()
803 _ctdb_service_reconfigure_common
804 rm -f "$_ctdb_service_reconfigure_flag"
807 ctdb_service_reconfigure ()
809 echo "Reconfiguring service \"${service_name}\"..."
810 ctdb_service_unset_reconfigure
811 service_reconfigure || return $?
812 # Intentionally have this use $service_name as default
813 # shellcheck disable=SC2119
817 # Default service_reconfigure() function does nothing.
818 service_reconfigure ()
823 # Default service_start() and service_stop() functions.
825 # These may be overridden in an eventscript.
828 service "$service_name" start
833 service "$service_name" stop
836 ##################################################################
838 # This exists only for backward compatibility with 3rd party scripts
840 ctdb_standard_event_handler ()
848 if [ "$_family" = "inet6" ] ; then
849 _iptables_cmd="ip6tables"
851 _iptables_cmd="iptables"
854 # iptables doesn't like being re-entered, so flock-wrap it.
855 flock -w 30 "${CTDB_SCRIPT_VARDIR}/iptables.flock" "$_iptables_cmd" "$@"
858 # AIX (and perhaps others?) doesn't have mktemp
859 # type is commonly supported and more portable than which(1)
860 # shellcheck disable=SC2039
861 if ! type mktemp >/dev/null 2>&1 ; then
865 if [ "$1" = "-d" ] ; then
870 _hex10=$(dd if=/dev/urandom count=20 2>/dev/null | \
873 _t="${_d}/tmp.${_hex10}"
886 ######################################################################
887 # NFS callout handling
893 if [ -z "$CTDB_NFS_CALLOUT" ] ; then
894 CTDB_NFS_CALLOUT="${CTDB_BASE}/nfs-linux-kernel-callout"
896 # Always export, for statd callout
897 export CTDB_NFS_CALLOUT
899 # If the callout wants to use this then it must create it
900 export CTDB_NFS_CALLOUT_STATE_DIR="${_state_dir}/callout-state"
902 # Export, if set, for use by clustered NFS callouts
903 if [ -n "$CTDB_NFS_STATE_FS_TYPE" ] ; then
904 export CTDB_NFS_STATE_FS_TYPE
906 if [ -n "$CTDB_NFS_STATE_MNT" ] ; then
907 export CTDB_NFS_STATE_MNT
910 nfs_callout_cache="${_state_dir}/nfs_callout_cache"
911 nfs_callout_cache_callout="${nfs_callout_cache}/CTDB_NFS_CALLOUT"
912 nfs_callout_cache_ops="${nfs_callout_cache}/ops"
915 nfs_callout_register ()
917 mkdir -p "$nfs_callout_cache_ops"
918 rm -f "$nfs_callout_cache_ops"/*
920 echo "$CTDB_NFS_CALLOUT" >"$nfs_callout_cache_callout"
922 _t=$(eval "$CTDB_NFS_CALLOUT" "register")
923 if [ -n "$_t" ] ; then
925 while IFS="" read _op ; do
926 touch "${nfs_callout_cache_ops}/${_op}"
929 touch "${nfs_callout_cache_ops}/ALL"
935 # Re-run registration if $CTDB_NFS_CALLOUT has changed
937 if [ -r "$nfs_callout_cache_callout" ] ; then
938 read _prev <"$nfs_callout_cache_callout"
940 if [ "$CTDB_NFS_CALLOUT" != "$_prev" ] ; then
944 # Run the operation if it is registered...
945 if [ -e "${nfs_callout_cache_ops}/${1}" ] || \
946 [ -e "${nfs_callout_cache_ops}/ALL" ]; then
947 eval "$CTDB_NFS_CALLOUT" "$@"
951 ########################################################
953 ########################################################
959 tickledir="${CTDB_SCRIPT_VARDIR}/tickles"
960 mkdir -p "$tickledir"
962 # What public IPs do I hold?
964 _ips=$($CTDB -X ip | awk -F'|' -v pnn="$_pnn" '$3 == pnn {print $2}')
966 # IPs and port as ss filters
968 for _ip in $_ips ; do
969 _ip_filter="${_ip_filter}${_ip_filter:+ || }src [${_ip}]"
971 _port_filter="sport == :${_port}"
973 # Record connections to our public IPs in a temporary file.
974 # This temporary file is in CTDB's private state directory and
975 # $$ is used to avoid a very rare race involving CTDB's script
976 # debugging. No security issue, nothing to see here...
977 _my_connections="${tickledir}/${_port}.connections.$$"
978 # Parentheses are needed around the filters for precedence but
979 # the parentheses can't be empty!
980 ss -tn state established \
981 "${_ip_filter:+( ${_ip_filter} )}" \
982 "${_port_filter:+( ${_port_filter} )}" |
983 awk 'NR > 1 {print $4, $3}' |
984 sort >"$_my_connections"
986 # Record our current tickles in a temporary file
987 _my_tickles="${tickledir}/${_port}.tickles.$$"
989 $CTDB -X gettickles "$_i" "$_port" |
990 awk -F'|' 'NR > 1 { printf "%s:%s %s:%s\n", $2, $3, $4, $5 }'
994 # Add tickles for connections that we haven't already got tickles for
995 comm -23 "$_my_connections" "$_my_tickles" | \
998 # Remove tickles for connections that are no longer there
999 comm -13 "$_my_connections" "$_my_tickles" | \
1002 rm -f "$_my_connections" "$_my_tickles"
1004 # Remove stale files from killed scripts
1005 # Files can't have spaces in name, more portable than -print0/-0
1006 # shellcheck disable=SC2038
1007 (cd "$tickledir" && find . -type f -mmin +10 | xargs -r rm)
1010 ########################################################
1011 # load a site local config file
1012 ########################################################
1014 [ -x "${CTDB_BASE}/rc.local" ] && {
1015 . "${CTDB_BASE}/rc.local"
1018 [ -d "${CTDB_BASE}/rc.local.d" ] && {
1019 for i in "${CTDB_BASE}/rc.local.d"/* ; do
1020 [ -x "$i" ] && . "$i"
1024 script_name="${0##*/}" # basename