}
[ ! -f "$CTDB_PUBLIC_ADDRESSES" ] && {
- echo "No public addresses file found. Nothing to do for 10.interfaces"
exit 0
}
+monitor_interfaces()
+{
+ local INTERFACES=`cat $CTDB_PUBLIC_ADDRESSES |
+ sed -e "s/^[^\t ]*[\t ]*//" -e "s/,/ /g" -e "s/[\t ]*$//"`
+
+ [ "$CTDB_PUBLIC_INTERFACE" ] && INTERFACES="$CTDB_PUBLIC_INTERFACE $INTERFACES"
+ [ "$CTDB_NATGW_PUBLIC_IFACE" ] && INTERFACES="$CTDB_NATGW_PUBLIC_IFACE $INTERFACES"
+
+
+ # For all but the 1st line, get the 2nd last field with commas
+ # changes to spaces.
+ local IFACES=`ctdb -Y ip -v | sed -e '1d' -e 's/:[^:]*:$//' -e 's/^.*://' -e 's/,/ /g'`
+
+ local IFACE
+
+ INTERFACES=`for IFACE in $INTERFACES $IFACES ; do echo $IFACE ; done | sort | uniq`
+
+ local fail=0
+ local ok=0
+ for IFACE in $INTERFACES ; do
+
+ ip addr show $IFACE 2>/dev/null >/dev/null || {
+ echo Interface $IFACE does not exist but it is used by public addresses.
+ continue
+ }
+
+ # These interfaces are sometimes bond devices
+ # When we use VLANs for bond interfaces, there will only
+ # be an entry in /proc for the underlying real interface
+ local REALIFACE=`echo $IFACE |sed -e 's/\..*$//'`
+ [ -f /proc/net/bonding/$REALIFACE ] && {
+ grep -q 'Currently Active Slave: None' /proc/net/bonding/$REALIFACE && {
+ echo "ERROR: No active slaves for bond device $REALIFACE"
+ fail=1
+ ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
+ continue;
+ }
+ grep -q '^MII Status: up' /proc/net/bonding/$REALIFACE || {
+ echo "ERROR: public network interface $REALIFACE is down"
+ fail=1
+ ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
+ continue;
+ }
+ grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' /proc/net/bonding/$REALIFACE && {
+ grep 'MII Status:' /proc/net/bonding/$REALIFACE | tail -n +2 | grep -q '^MII Status: up' || {
+ echo No active slaves for 802.ad bond device $REALIFACE
+ ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
+ fail=1
+ continue
+ }
+ }
+ ok=1 # we only set ok for interfaces known to ctdbd
+ ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
+ continue;
+ }
+
+ case $IFACE in
+ lo*)
+ # loopback is always working
+ ok=1 # we only set ok for interfaces known to ctdbd
+ ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
+ ;;
+ ib*)
+ # we dont know how to test ib links
+ ok=1 # we only set ok for interfaces known to ctdbd
+ ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
+ ;;
+ *)
+ [ -z "$IFACE" ] || {
+ [ "$(basename $(readlink /sys/class/net/$IFACE/device/driver) 2>/dev/null)" = virtio_net ] ||
+ 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...
+ /sbin/ip link set $IFACE up
+ ethtool $IFACE | grep -q 'Link detected: yes' || {
+ echo "ERROR: No link on the public network interface $IFACE"
+ fail=1
+ ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
+ continue
+ }
+ }
+ ok=1 # we only set ok for interfaces known to ctdbd
+ ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
+ }
+ ;;
+ esac
+
+ done
+
+ test x"$fail" = x"0" && {
+ return 0;
+ }
+
+ test x"$ok" = x"1" && {
+ return 2;
+ }
+
+ return 1;
+}
+
case "$1" in
#############################
# called when ctdbd starts up
[ -f /proc/sys/net/ipv4/conf/all/arp_filter ] && {
echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter
}
- cat "$CTDB_PUBLIC_ADDRESSES" | cut -d/ -f1 | while read _IP; do
- _IP_HELD=`/sbin/ip addr show | grep "inet $_IP/"`
- [ -z "$_IP_HELD" ] || {
- _IFACE=`echo $_IP_HELD | sed -e "s/.*\s//"`
- _NM=`echo $_IP_HELD | sed -e "s/.*$_IP\///" -e "s/\s.*//"`
- echo "Removing public address $_IP/$_NM from device $_IFACE"
- /sbin/ip addr del $_IP/$_NM dev $_IFACE
- }
+ ;;
+
+ #############################
+ # called after ctdbd has done its initial recovery
+ # and we start the services to become healthy
+ startup)
+ # Assume all links are good initially
+ INTERFACES=`for IFACE in $INTERFACES ; do echo $IFACE ; done | sort | uniq`
+
+ for IFACE in $INTERFACES ; do
+ ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
done
+
+ monitor_interfaces
+
;;
echo 1 > /proc/sys/net/ipv4/route/flush
;;
+ ##################################################
+ # called when ctdbd wants to update an IP address
+ updateip)
+ if [ $# != 5 ]; then
+ echo "must supply old interface, new interface, IP and maskbits"
+ exit 1
+ fi
+
+ # moving an IP is a bit more complex than it seems.
+ # First we drop all traffic on the old interface.
+ # Then we try to add the ip to the new interface and before
+ # we finally remove it from the old interface.
+ #
+ # 1) firewall this IP, so no new external packets arrive for it
+ # 2) add the IP to the new interface
+ # 3) remove the IP from the old interface
+ # 4) remove the firewall rule
+ # 5) use ctdb gratiousarp to propagate the new mac address
+ # 6) use netstat -tn to find existing connections, and tickle them
+ oiface=$2
+ niface=$3
+ ip=$4
+ maskbits=$5
+
+ failed=0
+ # we do an extra delete to cope with the script being killed
+ iptables -D INPUT -i $oiface -d $ip -j DROP 2> /dev/null
+ iptables -I INPUT -i $oiface -d $ip -j DROP
+
+ delete_ip_from_iface $oiface $ip $maskbits 2>/dev/null
+ delete_ip_from_iface $niface $ip $maskbits 2>/dev/null
+
+ add_ip_to_iface $niface $ip $maskbits || {
+ iptables -D INPUT -i $oiface -d $ip -j DROP 2> /dev/null
+ exit 1;
+ }
+
+ # cope with the script being killed while we have the interface blocked
+ iptables -D INPUT -i $oiface -d $ip -j DROP 2> /dev/null
+
+ # flush our route cache
+ echo 1 > /proc/sys/net/ipv4/route/flush
+
+ # propagate the new mac address
+ ctdb gratiousarp $ip $niface
+
+ # tickle all existing connections, so that dropped packets
+ # are retransmited and the tcp streams work
+
+ tickle_tcp_connections $ip
+
+ ;;
+
###########################################
# called when ctdbd has finished a recovery
;;
monitor)
- INTERFACES=`cat $CTDB_PUBLIC_ADDRESSES |
- sed -e "s/^[^\t ]*[\t ]*//" -e "s/[\t ]*$//"`
-
- [ "$CTDB_PUBLIC_INTERFACE" ] && INTERFACES="$CTDB_PUBLIC_INTERFACE $INTERFACES"
- [ "$CTDB_NATGW_PUBLIC_IFACE" ] && INTERFACES="$CTDB_NATGW_PUBLIC_IFACE $INTERFACES"
+ monitor_interfaces
+ ret=$?
- INTERFACES=`for IFACE in $INTERFACES ; do echo $IFACE ; done | sort | uniq`
-
- for IFACE in $INTERFACES ; do
- # These interfaces are sometimes bond devices
- # When we use VLANs for bond interfaces, there will only
- # be an entry in /proc for the underlying real interface
- REALIFACE=`echo $IFACE |sed -e 's/\..*$//'`
- [ -f /proc/net/bonding/$REALIFACE ] && {
- grep -q 'Currently Active Slave: None' /proc/net/bonding/$REALIFACE && {
- echo "ERROR: No active slaves for bond device $REALIFACE"
- exit 1
+ test x"$ret" = x"2" && {
+ test x"$CTDB_PARTIALLY_ONLINE_INTERFACES" != x"yes" && {
+ exit 1;
}
- grep -q '^MII Status: up' /proc/net/bonding/$REALIFACE || {
- echo "ERROR: public network interface $REALIFACE is down"
- exit 1
- }
- exit 0;
- }
+ # as long as we have one interface available don't become
+ # unhealthy
+ ret=0
+ }
- case $IFACE in
- ib*)
- # we dont know how to test ib links
- ;;
- *)
- [ -z "$IFACE" ] || {
- /usr/sbin/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...
- /sbin/ip link set $IFACE up
- /usr/sbin/ethtool $IFACE | grep -q 'Link detected: yes' || {
- echo "ERROR: No link on the public network interface $IFACE"
- exit 1
- }
- }
- }
- ;;
- esac
- done
+ test x"$ret" != x"0" && {
+ exit 1;
+ }
;;
*)
ctdb_standard_event_handler "$@"