1
0
mirror of https://github.com/samba-team/samba.git synced 2024-12-23 17:34:34 +03:00
samba-mirror/ctdb/config/functions

1251 lines
30 KiB
Plaintext
Raw Normal View History

# Hey Emacs, this is a -*- shell-script -*- !!!
# utility functions for ctdb event scripts
if [ -z "$CTDB_BASE" ] ; then
echo 'CTDB_BASE unset in CTDB functions file'
exit 1
fi
CTDB_VARDIR="/usr/local/var/lib/ctdb"
ctdb_rundir="/usr/local/var/run/ctdb"
# Only (and always) override these variables in test code
if [ -z "$CTDB_SCRIPT_VARDIR" ] ; then
CTDB_SCRIPT_VARDIR="/usr/local/var/lib/ctdb/state"
fi
if [ -z "$CTDB_SYS_ETCDIR" ] ; then
CTDB_SYS_ETCDIR="/etc"
fi
if [ -z "$CTDB_HELPER_BINDIR" ] ; then
CTDB_HELPER_BINDIR="/usr/local/libexec/ctdb"
fi
#######################################
# pull in a system config file, if any
rewrite_ctdb_options ()
{
case "$CTDB_DBDIR" in
tmpfs|tmpfs:*)
_opts_defaults="mode=700"
# Get any extra options specified after colon
if [ "$CTDB_DBDIR" = "tmpfs" ] ; then
_opts=""
else
_opts="${CTDB_DBDIR#tmpfs:}"
fi
# This is an internal variable, only used by ctdbd_wrapper.
# It is OK to repeat mount options - last value wins
CTDB_DBDIR_TMPFS_OPTIONS="${_opts_defaults}${_opts:+,}${_opts}"
CTDB_DBDIR="${ctdb_rundir}/CTDB_DBDIR"
;;
*)
CTDB_DBDIR_TMPFS_OPTIONS=""
esac
}
_loadconfig() {
if [ -z "$1" ] ; then
foo="${service_config:-${service_name}}"
if [ -n "$foo" ] ; then
loadconfig "$foo"
return
fi
fi
if [ "$1" != "ctdb" ] ; then
loadconfig "ctdb"
fi
if [ -z "$1" ] ; then
return
fi
if [ -f $CTDB_SYS_ETCDIR/sysconfig/$1 ]; then
. $CTDB_SYS_ETCDIR/sysconfig/$1
elif [ -f $CTDB_SYS_ETCDIR/default/$1 ]; then
. $CTDB_SYS_ETCDIR/default/$1
elif [ -f $CTDB_BASE/sysconfig/$1 ]; then
. $CTDB_BASE/sysconfig/$1
fi
if [ "$1" = "ctdb" ] ; then
_config="${CTDBD_CONF:-${CTDB_BASE}/ctdbd.conf}"
if [ -r "$_config" ] ; then
. "$_config"
fi
rewrite_ctdb_options
fi
}
loadconfig () {
_loadconfig "$@"
}
##############################################################
# CTDB_SCRIPT_DEBUGLEVEL can be overwritten by setting it in a
# configuration file.
debug ()
{
if [ ${CTDB_SCRIPT_DEBUGLEVEL:-2} -ge 4 ] ; then
# If there are arguments then echo them. Otherwise expect to
# use stdin, which allows us to pass lots of debug using a
# here document.
if [ -n "$1" ] ; then
echo "DEBUG: $*"
else
sed -e 's@^@DEBUG: @'
fi
else
if [ -z "$1" ] ; then
cat >/dev/null
fi
fi
}
die ()
{
_msg="$1"
_rc="${2:-1}"
echo "$_msg" >&2
exit $_rc
}
# Log given message or stdin to either syslog or a CTDB log file
# $1 is the tag passed to logger if syslog is in use.
script_log ()
{
_tag="$1" ; shift
case "$CTDB_LOGGING" in
file:*|"")
if [ -n "$CTDB_LOGGING" ] ; then
_file="${CTDB_LOGGING#file:}"
else
_file="/usr/local/var/log/log.ctdb"
fi
{
if [ -n "$*" ] ; then
echo "$*"
else
cat
fi
} >>"$_file"
;;
*)
# Handle all syslog:* variants here too. There's no tool to do
# the lossy things, so just use logger.
logger -t "ctdbd: ${_tag}" $*
;;
esac
}
# When things are run in the background in an eventscript then logging
# output might get lost. This is the "solution". :-)
background_with_logging ()
{
(
"$@" 2>&1 </dev/null |
script_log "${script_name}&"
)&
return 0
}
##############################################################
# check number of args for different events
ctdb_check_args ()
{
case "$1" in
takeip|releaseip)
if [ $# != 4 ]; then
echo "ERROR: must supply interface, IP and maskbits"
exit 1
fi
;;
updateip)
if [ $# != 5 ]; then
echo "ERROR: must supply old interface, new interface, IP and maskbits"
exit 1
fi
;;
esac
}
##############################################################
# determine on what type of system (init style) we are running
detect_init_style()
{
# only do detection if not already set:
[ -z "$CTDB_INIT_STYLE" ] || return
if [ -x /sbin/startproc ]; then
CTDB_INIT_STYLE="suse"
elif [ -x /sbin/start-stop-daemon ]; then
CTDB_INIT_STYLE="debian"
else
CTDB_INIT_STYLE="redhat"
fi
}
######################################################
# simulate /sbin/service on platforms that don't have it
# _service() makes it easier to hook the service() function for
# testing.
_service ()
{
_service_name="$1"
_op="$2"
# do nothing, when no service was specified
[ -z "$_service_name" ] && return
if [ -x /sbin/service ]; then
$_nice /sbin/service "$_service_name" "$_op"
elif [ -x /usr/sbin/service ]; then
$_nice /usr/sbin/service "$_service_name" "$_op"
elif [ -x /bin/systemctl ]; then
$_nice /bin/systemctl "$_op" "$_service_name"
elif [ -x $CTDB_SYS_ETCDIR/init.d/$_service_name ]; then
$_nice $CTDB_SYS_ETCDIR/init.d/$_service_name "$_op"
elif [ -x $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name ]; then
$_nice $CTDB_SYS_ETCDIR/rc.d/init.d/$_service_name "$_op"
fi
}
service()
{
_nice=""
_service "$@"
}
######################################################
# simulate /sbin/service (niced) on platforms that don't have it
nice_service()
{
_nice="nice"
_service "$@"
}
######################################################
# Cached retrieval of PNN from local node. This never changes so why
# open a client connection to the server each time this is needed?
# This sets $pnn - this avoid an unnecessary subprocess.
ctdb_get_pnn ()
{
_pnn_file="${CTDB_SCRIPT_VARDIR}/my-pnn"
if [ ! -f "$_pnn_file" ] ; then
ctdb pnn | sed -e 's@.*:@@' >"$_pnn_file"
fi
read pnn <"$_pnn_file"
}
# Cached retrieval of private IP address from local node. This never
# changes. Sets $ip_address to avoid an unnecessary subprocess.
ctdb_get_ip_address ()
{
_ip_addr_file="${CTDB_SCRIPT_VARDIR}/my-ip-address"
if [ ! -f "$_ip_addr_file" ] ; then
ctdb -X nodestatus |
awk -F '|' 'NR == 2 { print $3 }' >"$_ip_addr_file"
fi
read ip_address <"$_ip_addr_file"
}
######################################################
# wrapper around /proc/ settings to allow them to be hooked
# for testing
# 1st arg is relative path under /proc/, 2nd arg is value to set
set_proc ()
{
echo "$2" >"/proc/$1"
}
set_proc_maybe ()
{
if [ -w "/proc/$1" ] ; then
set_proc "$1" "$2"
fi
}
######################################################
# wrapper around getting file contents from /proc/ to allow
# this to be hooked for testing
# 1st arg is relative path under /proc/
get_proc ()
{
cat "/proc/$1"
}
######################################################
# Print up to $_max kernel stack traces for processes named $_program
program_stack_traces ()
{
_prog="$1"
_max="${2:-1}"
_count=1
for _pid in $(pidof "$_prog") ; do
[ $_count -le $_max ] || break
# Do this first to avoid racing with process exit
_stack=$(get_proc "${_pid}/stack" 2>/dev/null)
if [ -n "$_stack" ] ; then
echo "Stack trace for ${_prog}[${_pid}]:"
echo "$_stack"
_count=$(($_count + 1))
fi
done
}
######################################################
# Ensure $service_name is set
assert_service_name ()
{
[ -n "$service_name" ] || die "INTERNAL ERROR: \$service_name not set"
}
######################################################
# check a set of directories is available
# return 1 on a missing directory
# directories are read from stdin
######################################################
ctdb_check_directories_probe()
{
while IFS="" read d ; do
case "$d" in
*%*)
continue
;;
*)
[ -d "${d}/." ] || return 1
esac
done
}
######################################################
# check a set of directories is available
# directories are read from stdin
######################################################
ctdb_check_directories()
{
ctdb_check_directories_probe || {
echo "ERROR: $service_name directory \"$d\" not available"
exit 1
}
}
######################################################
# check a set of tcp ports
# usage: ctdb_check_tcp_ports <ports...>
######################################################
# This flag file is created when a service is initially started. It
# is deleted the first time TCP port checks for that service succeed.
# Until then ctdb_check_tcp_ports() prints a more subtle "error"
# message if a port check fails.
_ctdb_check_tcp_common ()
{
assert_service_name
_d="${CTDB_SCRIPT_VARDIR}/failcount"
_ctdb_service_started_file="${_d}/${service_name}.started"
}
ctdb_check_tcp_init ()
{
_ctdb_check_tcp_common
mkdir -p "${_ctdb_service_started_file%/*}" # dirname
touch "$_ctdb_service_started_file"
}
# Check whether something is listening on all of the given TCP ports
# using the "ctdb checktcpport" command.
ctdb_check_tcp_ports()
{
if [ -z "$1" ] ; then
echo "INTERNAL ERROR: ctdb_check_tcp_ports - no ports specified"
exit 1
fi
for _p ; do # process each function argument (port)
_cmd="ctdb checktcpport $_p"
_out=$($_cmd 2>&1)
_ret=$?
case "$_ret" in
0)
_ctdb_check_tcp_common
if [ ! -f "$_ctdb_service_started_file" ] ; then
echo "ERROR: $service_name tcp port $_p is not responding"
debug "\"ctdb checktcpport $_p\" was able to bind to port"
else
echo "INFO: $service_name tcp port $_p is not responding"
fi
return 1
;;
98)
# Couldn't bind, something already listening, next port...
continue
;;
*)
echo "ERROR: unexpected error running \"ctdb checktcpport\""
debug <<EOF
ctdb checktcpport (exited with $_ret) with output:
$_out"
EOF
return $_ret
esac
done
# All ports listening
_ctdb_check_tcp_common
rm -f "$_ctdb_service_started_file"
return 0
}
######################################################
# check a unix socket
# usage: ctdb_check_unix_socket SERVICE_NAME <socket_path>
######################################################
ctdb_check_unix_socket() {
socket_path="$1"
[ -z "$socket_path" ] && return
if ! netstat --unix -a -n | grep -q "^unix.*LISTEN.*${socket_path}$"; then
echo "ERROR: $service_name socket $socket_path not found"
return 1
fi
}
######################################################
# check a command returns zero status
# usage: ctdb_check_command <command>
######################################################
ctdb_check_command ()
{
_out=$("$@" 2>&1) || {
echo "ERROR: $* returned error"
echo "$_out" | debug
exit 1
}
}
################################################
# kill off any TCP connections with the given IP
################################################
kill_tcp_connections ()
{
_iface="$1"
_ip="$2"
_oneway=false
if [ "$3" = "oneway" ] ; then
_oneway=true
fi
get_tcp_connections_for_ip "$_ip" | {
_killcount=0
_connections=""
_nl="
"
while read _dst _src; do
_destport="${_dst##*:}"
__oneway=$_oneway
case $_destport in
# we only do one-way killtcp for CIFS
139|445) __oneway=true ;;
esac
echo "Killing TCP connection $_src $_dst"
_connections="${_connections}${_nl}${_src} ${_dst}"
if ! $__oneway ; then
_connections="${_connections}${_nl}${_dst} ${_src}"
fi
_killcount=$(($_killcount + 1))
done
if [ $_killcount -eq 0 ] ; then
return
fi
echo "$_connections" | \
"${CTDB_HELPER_BINDIR}/ctdb_killtcp" "$_iface" || {
echo "Failed to kill TCP connections"
return
}
_remaining=$(get_tcp_connections_for_ip $_ip | wc -l)
if [ $_remaining -eq 0 ] ; then
echo "Killed $_killcount TCP connections to released IP $_ip"
return
fi
_t="${_remaining}/${_killcount}"
echo "Failed to kill TCP connections for IP $_ip (${_t} remaining)"
}
}
##################################################################
# kill off the local end for any TCP connections with the given IP
##################################################################
kill_tcp_connections_local_only ()
{
kill_tcp_connections "$@" "oneway"
}
##################################################################
# tickle any TCP connections with the given IP
##################################################################
tickle_tcp_connections ()
{
_ip="$1"
get_tcp_connections_for_ip "$_ip" |
{
_failed=false
while read dest src; do
echo "Tickle TCP connection $src $dest"
ctdb tickle $src $dest >/dev/null 2>&1 || _failed=true
echo "Tickle TCP connection $dest $src"
ctdb tickle $dest $src >/dev/null 2>&1 || _failed=true
done
if $_failed ; then
echo "Failed to send tickle control"
fi
}
}
get_tcp_connections_for_ip ()
{
_ip="$1"
ss -tn state established "src [$_ip]" | awk 'NR > 1 {print $3, $4}'
}
ctdb-eventscripts: Deleting IPs should use the promote_secondaries option If a primary IP address is being deleted from an interface, the secondaries are remembered and added back after the primary is deleted. This is done under a lock shared by the add/del script code. It is necessary because, by default, Linux deletes secondaries when the corresponding primary is deleted. There is a race here between ctdbd and the scripts, since ctdbd doesn't know about the lock. If ctdbd receives a release IP control and the IP address is not on an interface then it is regarded as a "Redundant release of IP" so no "releaseip" event is generated. This can occur if the IP address in question is a secondary that has been temporarily dropped. It is more likely if the number of secondaries is large. Since Linux 2.6.12 (i.e. 2005) Linux has supported a promote_secondaries option on interfaces. This option is currently undocumented but that will change in Linux 3.14. With promote_secondaries enabled the kernel will not drop secondaries but will promote a corresponding secondary instead. The kernel does all necessary locking. Use promote_secondaries to simplify the code, avoid re-adding secondaries, avoid re-adding routes and provide improved performance. This could be done conditionally, with a fallback to legacy secondary-re-adding code, but no supported Linux distribution is running a pre-2.6.12 kernel so this is unnecessary. Signed-off-by: Martin Schwenke <martin@meltin.net> Reviewed-by: Amitay Isaacs <amitay@gmail.com>
2014-01-28 07:41:25 +04:00
########################################################
add_ip_to_iface ()
{
_iface=$1
_ip=$2
_maskbits=$3
ctdb-eventscripts: Deleting IPs should use the promote_secondaries option If a primary IP address is being deleted from an interface, the secondaries are remembered and added back after the primary is deleted. This is done under a lock shared by the add/del script code. It is necessary because, by default, Linux deletes secondaries when the corresponding primary is deleted. There is a race here between ctdbd and the scripts, since ctdbd doesn't know about the lock. If ctdbd receives a release IP control and the IP address is not on an interface then it is regarded as a "Redundant release of IP" so no "releaseip" event is generated. This can occur if the IP address in question is a secondary that has been temporarily dropped. It is more likely if the number of secondaries is large. Since Linux 2.6.12 (i.e. 2005) Linux has supported a promote_secondaries option on interfaces. This option is currently undocumented but that will change in Linux 3.14. With promote_secondaries enabled the kernel will not drop secondaries but will promote a corresponding secondary instead. The kernel does all necessary locking. Use promote_secondaries to simplify the code, avoid re-adding secondaries, avoid re-adding routes and provide improved performance. This could be done conditionally, with a fallback to legacy secondary-re-adding code, but no supported Linux distribution is running a pre-2.6.12 kernel so this is unnecessary. Signed-off-by: Martin Schwenke <martin@meltin.net> Reviewed-by: Amitay Isaacs <amitay@gmail.com>
2014-01-28 07:41:25 +04:00
# Ensure interface is up
ip link set "$_iface" up || \
die "Failed to bringup interface $_iface"
# Only need to define broadcast for IPv4
case "$ip" in
*:*) _bcast="" ;;
*) _bcast="brd +" ;;
esac
ip addr add "$_ip/$_maskbits" $_bcast dev "$_iface" || {
echo "Failed to add $_ip/$_maskbits on dev $_iface"
return 1
}
# Wait 5 seconds for IPv6 addresses to stop being tentative...
if [ -z "$_bcast" ] ; then
for _x in $(seq 1 10) ; do
ip addr show to "${_ip}/128" | grep -q "tentative" || break
sleep 0.5
done
# If the address was a duplicate then it won't be on the
# interface so flag an error.
_t=$(ip addr show to "${_ip}/128")
case "$_t" in
"")
echo "Failed to add $_ip/$_maskbits on dev $_iface"
return 1
;;
*tentative*|*dadfailed*)
echo "Failed to add $_ip/$_maskbits on dev $_iface"
ip addr del "$_ip/$_maskbits" dev "$_iface"
return 1
;;
esac
fi
}
delete_ip_from_iface()
{
_iface=$1
_ip=$2
_maskbits=$3
ctdb-eventscripts: Deleting IPs should use the promote_secondaries option If a primary IP address is being deleted from an interface, the secondaries are remembered and added back after the primary is deleted. This is done under a lock shared by the add/del script code. It is necessary because, by default, Linux deletes secondaries when the corresponding primary is deleted. There is a race here between ctdbd and the scripts, since ctdbd doesn't know about the lock. If ctdbd receives a release IP control and the IP address is not on an interface then it is regarded as a "Redundant release of IP" so no "releaseip" event is generated. This can occur if the IP address in question is a secondary that has been temporarily dropped. It is more likely if the number of secondaries is large. Since Linux 2.6.12 (i.e. 2005) Linux has supported a promote_secondaries option on interfaces. This option is currently undocumented but that will change in Linux 3.14. With promote_secondaries enabled the kernel will not drop secondaries but will promote a corresponding secondary instead. The kernel does all necessary locking. Use promote_secondaries to simplify the code, avoid re-adding secondaries, avoid re-adding routes and provide improved performance. This could be done conditionally, with a fallback to legacy secondary-re-adding code, but no supported Linux distribution is running a pre-2.6.12 kernel so this is unnecessary. Signed-off-by: Martin Schwenke <martin@meltin.net> Reviewed-by: Amitay Isaacs <amitay@gmail.com>
2014-01-28 07:41:25 +04:00
# This could be set globally for all interfaces but it is probably
# better to avoid surprises, so limit it the interfaces where CTDB
# has public IP addresses. There isn't anywhere else convenient
# to do this so just set it each time. This is much cheaper than
# remembering and re-adding secondaries.
set_proc "sys/net/ipv4/conf/${_iface}/promote_secondaries" 1
ip addr del "$_ip/$_maskbits" dev "$_iface" || {
echo "Failed to del $_ip on dev $_iface"
return 1
}
}
# If the given IP is hosted then print 2 items: maskbits and iface
ip_maskbits_iface ()
{
_addr="$1"
case "$_addr" in
*:*) _family="inet6" ; _bits=128 ;;
*) _family="inet" ; _bits=32 ;;
esac
ip addr show to "${_addr}/${_bits}" 2>/dev/null | \
awk -v family="${_family}" \
'NR == 1 { iface = $2; sub(":$", "", iface) ; \
sub("@.*", "", iface) } \
$1 ~ /inet/ { mask = $2; sub(".*/", "", mask); \
print mask, iface, family }'
}
drop_ip ()
{
_addr="${1%/*}" # Remove optional maskbits
set -- $(ip_maskbits_iface $_addr)
if [ -n "$1" ] ; then
_maskbits="$1"
_iface="$2"
echo "Removing public address $_addr/$_maskbits from device $_iface"
delete_ip_from_iface $_iface $_addr $_maskbits >/dev/null 2>&1
fi
}
drop_all_public_ips ()
{
while read _ip _x ; do
drop_ip "$_ip"
done <"${CTDB_PUBLIC_ADDRESSES:-/dev/null}"
}
flush_route_cache ()
{
set_proc_maybe sys/net/ipv4/route/flush 1
set_proc_maybe sys/net/ipv6/route/flush 1
}
########################################################
# Interface monitoring
# If the interface is a virtual one (e.g. VLAN) then get the
# underlying interface
interface_get_real ()
{
# Output of "ip link show <iface>"
_iface_info="$1"
# Extract the full interface description to see if it is a VLAN
_t=$(echo "$_iface_info" |
awk 'NR == 1 { iface = $2; sub(":$", "", iface) ; \
print iface }')
case "$_t" in
*@*)
# VLAN: use the underlying interface, after the '@'
echo "${_t##*@}"
;;
*)
# Not a regular VLAN. For backward compatibility, assume
# there is some other sort of VLAN that doesn't have the
# '@' in the output and only use what is before a '.'. If
# there is no '.' then this will be the whole interface
# name.
echo "${_t%%.*}"
esac
}
# Check whether an interface is operational
interface_monitor ()
{
_iface="$1"
_iface_info=$(ip link show "$_iface" 2>&1) || {
echo "ERROR: Monitored interface ${_iface} does not exist"
return 1
}
# If the interface is a virtual one (e.g. VLAN) then get the
# underlying interface.
_realiface=$(interface_get_real "$_iface_info")
if _bi=$(get_proc "net/bonding/${_realiface}" 2>/dev/null) ; then
# This is a bond: various monitoring strategies
echo "$_bi" | grep -q 'Currently Active Slave: None' && {
echo "ERROR: No active slaves for bond device ${_realiface}"
return 1
}
echo "$_bi" | grep -q '^MII Status: up' || {
echo "ERROR: public network interface ${_realiface} is down"
return 1
}
echo "$_bi" | grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' && {
# This works around a bug in the driver where the
# overall bond status can be up but none of the actual
# physical interfaces have a link.
echo "$_bi" | grep 'MII Status:' | tail -n +2 | grep -q '^MII Status: up' || {
echo "ERROR: No active slaves for 802.ad bond device ${_realiface}"
return 1
}
}
return 0
else
# Not a bond
case "$_iface" in
lo*)
# loopback is always working
return 0
;;
ib*)
# we don't know how to test ib links
return 0
;;
*)
ethtool "$_iface" | grep -q 'Link detected: yes' || {
# On some systems, this is not successful when a
# cable is plugged but the interface has not been
# brought up previously. Bring the interface up
# and try again...
ip link set "$_iface" up
ethtool "$_iface" | grep -q 'Link detected: yes' || {
echo "ERROR: No link on the public network interface ${_iface}"
return 1
}
}
return 0
;;
esac
fi
}
########################################################
# Simple counters
_ctdb_counter_common () {
_service_name="${1:-${service_name:-${script_name}}}"
_counter_file="${CTDB_SCRIPT_VARDIR}/failcount/${_service_name}"
mkdir -p "${_counter_file%/*}" # dirname
}
ctdb_counter_init () {
_ctdb_counter_common "$1"
>"$_counter_file"
}
ctdb_counter_incr () {
_ctdb_counter_common "$1"
# unary counting!
echo -n 1 >> "$_counter_file"
}
ctdb_counter_get () {
_ctdb_counter_common "$1"
# unary counting!
stat -c "%s" "$_counter_file" 2>/dev/null || echo 0
}
ctdb_check_counter () {
_msg="${1:-error}" # "error" - anything else is silent on fail
_op="${2:--ge}" # an integer operator supported by test
_limit="${3:-${service_fail_limit}}"
shift 3
_size=$(ctdb_counter_get "$1")
_hit=false
if [ "$_op" != "%" ] ; then
if [ $_size $_op $_limit ] ; then
_hit=true
fi
else
if [ $(($_size $_op $_limit)) -eq 0 ] ; then
_hit=true
fi
fi
if $_hit ; then
if [ "$_msg" = "error" ] ; then
echo "ERROR: $_size consecutive failures for $_service_name, marking node unhealthy"
exit 1
else
return 1
fi
fi
}
########################################################
ctdb_setup_service_state_dir ()
{
service_state_dir="${CTDB_SCRIPT_VARDIR}/service_state/${1:-${service_name}}"
mkdir -p "$service_state_dir" || {
echo "Error creating state dir \"$service_state_dir\""
exit 1
}
}
########################################################
# Managed status history, for auto-start/stop
_ctdb_managed_common ()
{
_ctdb_managed_file="${CTDB_SCRIPT_VARDIR}/managed_history/${service_name}"
}
ctdb_service_managed ()
{
_ctdb_managed_common
mkdir -p "${_ctdb_managed_file%/*}" # dirname
touch "$_ctdb_managed_file"
}
ctdb_service_unmanaged ()
{
_ctdb_managed_common
rm -f "$_ctdb_managed_file"
}
is_ctdb_previously_managed_service ()
{
_ctdb_managed_common
[ -f "$_ctdb_managed_file" ]
}
##################################################################
# Reconfigure a service on demand
_ctdb_service_reconfigure_common ()
{
_d="${CTDB_SCRIPT_VARDIR}/service_status/${service_name}"
mkdir -p "$_d"
_ctdb_service_reconfigure_flag="$_d/reconfigure"
}
ctdb_service_needs_reconfigure ()
{
_ctdb_service_reconfigure_common
[ -e "$_ctdb_service_reconfigure_flag" ]
}
ctdb_service_set_reconfigure ()
{
_ctdb_service_reconfigure_common
>"$_ctdb_service_reconfigure_flag"
}
ctdb_service_unset_reconfigure ()
{
_ctdb_service_reconfigure_common
rm -f "$_ctdb_service_reconfigure_flag"
}
ctdb_service_reconfigure ()
{
echo "Reconfiguring service \"${service_name}\"..."
ctdb_service_unset_reconfigure
service_reconfigure || return $?
ctdb_counter_init
}
# Default service_reconfigure() function does nothing.
service_reconfigure ()
{
:
}
ctdb_reconfigure_take_lock ()
{
_ctdb_service_reconfigure_common
_lock="${_d}/reconfigure_lock"
mkdir -p "${_lock%/*}" # dirname
touch "$_lock"
(
flock 0
# This is overkill but will work if we need to extend this to
# allow certain events to run multiple times in parallel
# (e.g. takeip) and write multiple PIDs to the file.
read _locker_event
if [ -n "$_locker_event" ] ; then
while read _pid ; do
if [ -n "$_pid" -a "$_pid" != $$ ] && \
kill -0 "$_pid" 2>/dev/null ; then
exit 1
fi
done
fi
printf "%s\n%s\n" "$event_name" $$ >"$_lock"
exit 0
) <"$_lock"
}
ctdb_reconfigure_release_lock ()
{
_ctdb_service_reconfigure_common
_lock="${_d}/reconfigure_lock"
rm -f "$_lock"
}
ctdb_replay_monitor_status ()
{
echo "Replaying previous status for this script due to reconfigure..."
# Leading separator ('|') is missing in some versions...
_out=$(ctdb scriptstatus -X | grep -E "^\|?monitor\|${script_name}\|")
# Output looks like this:
# |monitor|60.nfs|1|ERROR|1314764004.030861|1314764004.035514|foo bar|
# This is the cheapest way of getting fields in the middle.
set -- $(IFS="|" ; echo $_out)
_code="$3"
_status="$4"
# The error output field can include colons so we'll try to
# preserve them. The weak checking at the beginning tries to make
# this work for both broken (no leading '|') and fixed output.
_out="${_out%|}"
_err_out="${_out#*monitor|${script_name}|*|*|*|*|}"
case "$_status" in
OK) : ;; # Do nothing special.
TIMEDOUT)
# Recast this as an error, since we can't exit with the
# correct negative number.
_code=1
_err_out="[Replay of TIMEDOUT scriptstatus - note incorrect return code.] ${_err_out}"
;;
DISABLED)
# Recast this as an OK, since we can't exit with the
# correct negative number.
_code=0
_err_out="[Replay of DISABLED scriptstatus - note incorrect return code.] ${_err_out}"
;;
*) : ;; # Must be ERROR, do nothing special.
esac
if [ -n "$_err_out" ] ; then
echo "$_err_out"
fi
exit $_code
}
ctdb_service_check_reconfigure ()
{
assert_service_name
# We only care about some events in this function. For others we
# return now.
case "$event_name" in
monitor|ipreallocated|reconfigure) : ;;
*) return 0 ;;
esac
if ctdb_reconfigure_take_lock ; then
# No events covered by this function are running, so proceed
# with gay abandon.
case "$event_name" in
reconfigure)
(ctdb_service_reconfigure)
exit $?
;;
ipreallocated)
if ctdb_service_needs_reconfigure ; then
ctdb_service_reconfigure
fi
;;
esac
ctdb_reconfigure_release_lock
else
# Somebody else is running an event we don't want to collide
# with. We proceed with caution.
case "$event_name" in
reconfigure)
# Tell whoever called us to retry.
exit 2
;;
ipreallocated)
# Defer any scheduled reconfigure and just run the
# rest of the ipreallocated event, as per the
# eventscript. There's an assumption here that the
# event doesn't depend on any scheduled reconfigure.
# This is true in the current code.
return 0
;;
monitor)
# There is most likely a reconfigure in progress so
# the service is possibly unstable. As above, we
# defer any scheduled reconfigured. We also replay
# the previous monitor status since that's the best
# information we have.
ctdb_replay_monitor_status
;;
esac
fi
}
##################################################################
# Does CTDB manage this service? - and associated auto-start/stop
ctdb_compat_managed_service ()
{
if [ "$1" = "yes" -a "$2" = "$service_name" ] ; then
CTDB_MANAGED_SERVICES="$CTDB_MANAGED_SERVICES $2"
fi
}
is_ctdb_managed_service ()
{
assert_service_name
# $t is used just for readability and to allow better accurate
# matching via leading/trailing spaces
t=" $CTDB_MANAGED_SERVICES "
# Return 0 if "<space>$service_name<space>" appears in $t
if [ "${t#* ${service_name} }" != "${t}" ] ; then
return 0
fi
# If above didn't match then update $CTDB_MANAGED_SERVICES for
# backward compatibility and try again.
ctdb_compat_managed_service "$CTDB_MANAGES_VSFTPD" "vsftpd"
ctdb_compat_managed_service "$CTDB_MANAGES_SAMBA" "samba"
ctdb_compat_managed_service "$CTDB_MANAGES_WINBIND" "winbind"
ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "apache2"
ctdb_compat_managed_service "$CTDB_MANAGES_HTTPD" "httpd"
ctdb_compat_managed_service "$CTDB_MANAGES_ISCSI" "iscsi"
ctdb_compat_managed_service "$CTDB_MANAGES_CLAMD" "clamd"
ctdb_compat_managed_service "$CTDB_MANAGES_NFS" "nfs"
t=" $CTDB_MANAGED_SERVICES "
# Return 0 if "<space>$service_name<space>" appears in $t
[ "${t#* ${service_name} }" != "${t}" ]
}
ctdb_start_stop_service ()
{
assert_service_name
# Allow service-start/service-stop pseudo-events to start/stop
# services when we're not auto-starting/stopping and we're not
# monitoring.
case "$event_name" in
service-start)
if is_ctdb_managed_service ; then
die 'service-start event not permitted when service is managed'
fi
if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
die 'service-start event not permitted with $CTDB_SERVICE_AUTOSTARTSTOP = yes'
fi
ctdb_service_start
exit $?
;;
service-stop)
if is_ctdb_managed_service ; then
die 'service-stop event not permitted when service is managed'
fi
if [ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] ; then
die 'service-stop event not permitted with $CTDB_SERVICE_AUTOSTARTSTOP = yes'
fi
ctdb_service_stop
exit $?
;;
esac
# Do nothing unless configured to...
[ "$CTDB_SERVICE_AUTOSTARTSTOP" = "yes" ] || return 0
[ "$event_name" = "monitor" ] || return 0
if is_ctdb_managed_service ; then
if ! is_ctdb_previously_managed_service ; then
echo "Starting service \"$service_name\" - now managed"
background_with_logging ctdb_service_start
exit $?
fi
else
if is_ctdb_previously_managed_service ; then
echo "Stopping service \"$service_name\" - no longer managed"
background_with_logging ctdb_service_stop
exit $?
fi
fi
}
ctdb_service_start ()
{
# The service is marked managed if we've ever tried to start it.
ctdb_service_managed
service_start || return $?
ctdb_counter_init
ctdb_check_tcp_init
}
ctdb_service_stop ()
{
ctdb_service_unmanaged
service_stop
}
# Default service_start() and service_stop() functions.
# These may be overridden in an eventscript.
service_start ()
{
service "$service_name" start
}
service_stop ()
{
service "$service_name" stop
}
##################################################################
ctdb_standard_event_handler ()
{
:
}
iptables_wrapper ()
{
_family="$1" ; shift
if [ "$_family" = "inet6" ] ; then
_iptables_cmd="ip6tables"
else
_iptables_cmd="iptables"
fi
# iptables doesn't like being re-entered, so flock-wrap it.
flock -w 30 "${CTDB_SCRIPT_VARDIR}/iptables.flock" "$_iptables_cmd" "$@"
}
# AIX (and perhaps others?) doesn't have mktemp
if ! type mktemp >/dev/null 2>&1 ; then
mktemp ()
{
_dir=false
if [ "$1" = "-d" ] ; then
_dir=true
shift
fi
_d="${TMPDIR:-/tmp}"
_hex10=$(dd if=/dev/urandom count=20 2>/dev/null | \
md5sum | \
sed -e 's@\(..........\).*@\1@')
_t="${_d}/tmp.${_hex10}"
(
umask 077
if $_dir ; then
mkdir "$_t"
else
>"$_t"
fi
)
echo "$_t"
}
fi
########################################################
# tickle handling
########################################################
update_tickles ()
{
_port="$1"
tickledir="${CTDB_SCRIPT_VARDIR}/tickles"
mkdir -p "$tickledir"
ctdb_get_pnn
# What public IPs do I hold?
_ips=$(ctdb -X ip | awk -F'|' -v pnn=$pnn '$3 == pnn {print $2}')
# IPs and port as ss filters
_ip_filter=""
for _ip in $_ips ; do
_ip_filter="${_ip_filter}${_ip_filter:+ || }src [${_ip}]"
done
_port_filter="sport == :${_port}"
# Record connections to our public IPs in a temporary file.
# This temporary file is in CTDB's private state directory and
# $$ is used to avoid a very rare race involving CTDB's script
# debugging. No security issue, nothing to see here...
_my_connections="${tickledir}/${_port}.connections.$$"
# Parentheses are needed around the filters for precedence but
# the parentheses can't be empty!
ss -tn state established \
"${_ip_filter:+( ${_ip_filter} )}" \
"${_port_filter:+( ${_port_filter} )}" |
awk 'NR > 1 {print $4, $3}' |
sort >"$_my_connections"
# Record our current tickles in a temporary file
_my_tickles="${tickledir}/${_port}.tickles.$$"
for _i in $_ips ; do
ctdb -X gettickles $_i $_port |
awk -F'|' 'NR > 1 { printf "%s:%s %s:%s\n", $2, $3, $4, $5 }'
done |
sort >"$_my_tickles"
# Add tickles for connections that we haven't already got tickles for
comm -23 "$_my_connections" "$_my_tickles" |
while read _src _dst ; do
ctdb addtickle $_src $_dst
done
# Remove tickles for connections that are no longer there
comm -13 "$_my_connections" "$_my_tickles" |
while read _src _dst ; do
ctdb deltickle $_src $_dst
done
rm -f "$_my_connections" "$_my_tickles"
# Remove stale files from killed scripts
find "$tickledir" -type f -mmin +10 | xargs -r rm
}
########################################################
# load a site local config file
########################################################
[ -n "$CTDB_RC_LOCAL" -a -x "$CTDB_RC_LOCAL" ] && {
. "$CTDB_RC_LOCAL"
}
[ -x $CTDB_BASE/rc.local ] && {
. $CTDB_BASE/rc.local
}
[ -d $CTDB_BASE/rc.local.d ] && {
for i in $CTDB_BASE/rc.local.d/* ; do
[ -x "$i" ] && . "$i"
done
}
script_name="${0##*/}" # basename
service_fail_limit=1
event_name="$1"