How to connect two network interfaces on the same subnet?
Environment
- Red Hat Enterprise Linux 5
- Red Hat Enterprise Linux 6
- Red Hat Enterprise Linux 7
- Red Hat Enterprise Linux 8
- Multiple network interfaces, each with an IP address in the same subnet
Issue
- How to connect two network interfaces on the same subnet?
-
In our environment, there are three bonding devices connected with the same segment.
+---------------------+ | Linux | | .168 .169 | +-----+--------+------+ │ │ +-----+--------+------+ | Switch | +---------+-----------+ │ +---------+-----------+ | .1 | | Gateway | +---------------------+
-
We have captured packet and found that packet should be transmitted from bond0 was actually transmitted from bond1
- Also, it was confirmed that the transmit port staggered even if there was no bonding setting.
Resolution
It is usually not a good idea to connect two interfaces using the same subnet on the system, as you may run into the issues presented in the following two pages:
- Why is my server responding to arp requests from the wrong interface on Red Hat Enterprise Linux ?
- When RHEL has multiple IPs configured, only one is reachable from a remote network. Or why does RHEL ignore packets when the route for outbound traffic differs from the route of incoming traffic?
The recommended solution is to use a single interface or a single bond/team interface, then add the other IP addresses to that interface in the form of an alias IP address. If fault tolerance and/or load balancing is required at the network interface level, then an alias on a bond or team should be used. Please note that teaming is not supported in Red Hat Enterprise Linux 6 or earlier. If fault tolerance and/or load balancing is not needed at the interface level, then you can add the alias IP address to a single interface.
The following page provides steps on how to add an alias IP address: How to assign alias IP addresses to a network card (NIC)?.
The following page provides steps on how to set up a bond interface: How do I configure a bonding device on Red Hat Enterprise Linux (RHEL)?
The following pages provide documentation on teaming:
- Red Hat Enterprise Linux 7: Red Hat Enterprise Linux 7 Networking Guide Chapter 8. Configure Network Teaming
- Red Hat Enterprise Linux 8: Red Hat Enterprise Linux 8 Configuring and managing networking Chapter 17. Configuring network teaming
If your environment necessitates a multi-interface same-subnet solution and the above alias interface approach does not work for your environment, you may perform the steps in the Resolution section of the following two pages to avoid the most common issues when setting up multiple interfaces on the same subnet:
Root Cause
- When there are 2 interfaces on the same subnet there is no assurance as to which interface will be used to transmit traffic and the machine will accept traffic for either IP on either interface.
- This is because in Linux the IP address belongs to the host and is not associated with the interface.
- If you ping with
-I DEV
, attempting to use a given interface, there is no guarantee the reply packet (if there even is one) will come back to the same interface, so pings done with-I DEV
may not work.
Diagnostic Steps
- Setup system with 2 interfaces on the same subnet.
- Ping a target and capture packets with
tcpdump
.
This solution is part of Red Hat’s fast-track publication program, providing a huge library of solutions that Red Hat engineers have created while supporting our customers. To give you the knowledge you need the instant it becomes available, these articles may be presented in a raw and unedited form.
14 Comments
Is there a typo in step 2) ??
Should these lines be identical?
or shouldn't the second 'ip route add' look like this...
2. Add routing table to t1, t2 :
.
this article still have wrong device name for default routing in t2 table
There is also anotehr typo...
As for 4):
should say
Thanks for your comments. I've updated the article with the corrections.
This script automatically implements the changes and creation of static routes described above, in case anyone wants to use it.
All you need to provide is the subnet that you want to effect.
#!/bin/sh
# fixFlux.sh - Assigns fixed routes for each NIC
# residing on the same subnet to resolve
# the ARP FlUX problem
# See https://access.redhat.com/kb/docs/DOC-59096
#
# Date: 7/21/2011
SCRIPT_REVISION=1.0.1
SCRIPT_LOG=/dev/nul
# Author: Ken Banks
# Company: Kontron America.
#
# CHANGE LOG:
# 7/21/2011 1.0.1 Original code.
# 8/23/2011 1.0.2 Fix a bug with 'undo' to re-enble NetworkManager
# and another bug that allowed multiple 'no's to be
# added to NM_CONTROLLED= if the script was run more than once.
SYSCTL=/etc/sysctl.conf
RT_TABLES=/etc/iproute2/rt_tables
function usage() {
echo "usage: $0 subnet <undo>"
echo " version $SCRIPT_REVISION"
echo " : This script will solve the Arp Flux for multiple NICs"
echo " : residing on the same subnet, as described in this document:"
echo " : https://access.redhat.com/kb/docs/DOC-59096"
echo
echo " e.g. $0 192.168.1 - Configures a static route for ever IC on 192.168.1.0"
echo " $0 192.168.1 undo - Reverts the changes caused above and re-instates NetworkManager"
}
function log_message() {
echo -e "$*"
echo -e "$*" >> $SCRIPT_LOG
}
function compareValues() {
awk "BEGIN { if ( $1 $2 $3 ) print \"TRUE\" }"
}
function addRoute() {
NICNAME=$1
TABLENUM=$2
TABLENAME=$3
SUBNET=$4
DEFAULTGW=$5
NICIP=$6
FirstOctect=`echo $SUBNET | awk -F. '{ print $1 }'`
ip route show | grep $6 > /dev/nul
if [ "$?" != "0" ] ; then
echo ip route add $SUBNET dev $NICNAME src $NICIP table $TABLENAME
ip route add $SUBNET dev $NICNAME src $NICIP table $TABLENAME
echo ip route add table $TABLENAME default via $DEFAULTGW dev $NICNAME
ip route add table $TABLENAME default via $DEFAULTGW dev $NICNAME
#echo ip route show table $TABLENAME
#ip route show table $TABLENAME
fi
ETHROUTE=/etc/sysconfig/network-scripts/route-$NICNAME
echo $FirstOctect.0.0.0/8 dev $NICNAME src $NICIP table $TABLENAME > $ETHROUTE
echo default via $DEFAULTGW dev $NICNAME table $TABLENAME >> $ETHROUTE
echo "Modified: $ETHROUTE"
ETHRULE=/etc/sysconfig/network-scripts/rule-$NICNAME
echo table $TABLENAME from $NICIP > $ETHRULE
echo "Modified: $ETHRULE"
}
function disableNetworkManager() {
service NetworkManager status | grep stopped >/dev/nul
if [ "$?" != "0" ] ; then
service NetworkManager stop > /dev/nul
fi
echo "NetworkManager service is stopped."
chkconfig NetworkManager --list | grep :on >/dev/nul
if [ "$?" == "0" ] ; then
chkconfig NetworkManager off > /dev/nul
fi
echo "NetworkManager service is disabled."
ETHSCRIPT=/etc/sysconfig/network-scripts/ifcfg-$1
TMPSCRIPT=/tmp/tmpcfg-$1
grep -i NM_CONTROLLED=[\"yes\"] $ETHSCRIPT | awk -F. '{ print $1 }' > /dev/nul
if [ "$?" == "0" ] ; then
cp $ETHSCRIPT $TMPSCRIPT
cat $TMPSCRIPT | sed 's/NM_CONTROLLED=["yes"].*/NM_CONTROLLED=no/' > $ETHSCRIPT
echo "Modified: $ETHSCRIPT"
fi
}
function processNIC() {
# inet 192.168.6.99/24 brd 192.168.6.255 scope global eth0show table
# inet 192.168.6.60/24 brd 192.168.6.255 scope global eth1
export NICNAME=$1
export TABLENUM=$2
export TABLENAME=$3
export SUBNET=$4
export DEFAULTGW=`ip route | grep default | awk '{ print $3 }'`
export NICDATA=`ifconfig -a $NICNAME | grep Bcast`
NICIP=`echo $NICDATA | awk '{ print $2 }' | awk -F: '{ print $2 }'`
if [ "$NICIP" == "" ] ; then
echo "ERROR: No IP for this NIC: $NICNAME"
exit
fi
echo "====PROCESS NIC==== $NICDATA"
echo "==== $NICNAME ===="
grep $TABLENAME $RT_TABLES > /dev/nul
if [ "$?" != "0" ] ; then
echo "Creating route table: $TABLENUM $TABLENAME"
echo "$TABLENUM $TABLENAME" >> $RT_TABLES
else
echo "Route table already exists: $TABLENUM $TABLENAME"
fi
addRoute $NICNAME $TABLENUM $TABLENAME $SUBNET $DEFAULTGW $NICIP
disableNetworkManager $NICNAME
echo
}
function delete_route() {
ip route delete $1 dev $2
}
function undo_routes() {
SUBNET=$1
echo 0 > /proc/sys/net/ipv4/conf/default/arp_filter
echo 0 > /proc/sys/net/ipv4/conf/all/arp_filter
cp $RT_TABLES /tmp/rt_tables
grep -v "^.00 t[0-9]*" /tmp/rt_tables > $RT_TABLES
rm -f /etc/sysconfig/network-scripts/r[ou][ul][te]*
for i in `ip route show | grep src | awk '{ print $1"-"$3 }'` ; do
delete_route `echo $i | sed 's/-/ /'`
done
grep net.ipv4.conf.all.arp_filter $SYSCTL > /dev/nul
if [ "$?" == "0" ] ; then
grep "arp_filter[ =]*0" $SYSCTL > /dev/nul
if [ "$?" != "0" ] ; then
cp $SYSCTL /tmp/tmpctl.conf
cat /tmp/tmpctl.conf | sed 's/arp_filter[ =]*./arp_filter = 0/' >> $SYSCTL
echo "Modified: $SYSCTL"
fi
fi
for device in `ip addr | grep $SUBNET | awk '{ print $7 }'` ; do
ETHSCRIPT=/etc/sysconfig/network-scripts/ifcfg-$device
TMPSCRIPT=/tmp/tmpcfg-$device
grep -i NM_CONTROLLED=[\"no\"] $ETHSCRIPT | awk -F. '{ print $1 }' > /dev/nul
if [ "$?" == "0" ] ; then
cp $ETHSCRIPT $TMPSCRIPT
cat $TMPSCRIPT | sed 's/NM_CONTROLLED=[\"no\"].*/NM_CONTROLLED=yes/' > $ETHSCRIPT
echo "Modified: $ETHSCRIPT"
fi
done
chkconfig NetworkManager on
service NetworkManager start
}
function main() {
[ "$1" == "" ] && usage && exit
SUBNET=$1
[ "$2" == "undo" ] && undo_routes $SUBNET && exit
echo "# $0 $*"
log_message "=========================================================================="
log_message "=============================BEGIN ROUTE CREATION========================="
log_message "=========================================================================="
table=1
for device in `ip addr | grep $SUBNET | awk '{ print $7 }'` ; do
processNIC $device $((100 * $table)) t$table $SUBNET
table=$((1 + $table))
done
echo 1 > /proc/sys/net/ipv4/conf/default/arp_filter
echo "set /proc/sys/net/ipv4/conf/default/arp_filter = 1"
echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter
echo "set /proc/sys/net/ipv4/conf/all/arp_filter = 1"
grep net.ipv4.conf.all.arp_filter $SYSCTL > /dev/nul
if [ "$?" != "0" ] ; then
echo "net.ipv4.conf.all.arp_filter = 1" >> $SYSCTL
echo "Modified: $SYSCTL"
else
grep "arp_filter[ =]*1" $SYSCTL > /dev/nul
if [ "$?" != "0" ] ; then
cp $SYSCTL /tmp/tmpctl.conf
cat /tmp/tmpctl.conf | sed 's/arp_filter[ =]*./arp_filter = 1/' >> $SYSCTL
echo "Modified: $SYSCTL"
fi
fi
service network restart
log_message "=========================================================================="
log_message "==========================ROUTES HAVE BEEN CREATED========================"
log_message "=========================================================================="
echo "Contents of file: $RT_TABLES"
cat $RT_TABLES
echo "Command output from: ip route show"
ip route show
}
main $*
the prefix was 24 at the beginning, but became 8 in the next step,way ? and I need to run "ip route add default via 10.64.208.254 dev eth1 metric 101" to make this routes persistent.
the prefix was 24 at the beginning, but became 8 in the next step,way ? and I need to run "ip route add default via 10.64.208.254 dev eth1 metric 101" to make this routes persistent.
The section with dot point "Add rules to apply traffic to the routing tables". Shows how to add an "ip rule" but does not have the correct command for displaying the newly added rule.
eg//
After adding a "rule", it would be better to run a "ip rule show" rather than "ip route show" eg//
What would be the equivalent nmcli commands on el8?
Hello,
I have re-written the Resolution section substantially. As a result, many of the previous comments may no longer make sense as they applied to the steps that were in the Resolution section before the change. The new Resolution section provides links to existing content in the knowledge base that provide more implementation details and example commands.
The root cause should be more specific: "When there are 2 interfaces on the same subnet..."
Clarify same subnet. Does it mean that the IPs have the same class C (e.g x.x.2.1 and x.x.2.100) octet, or is it when both IPs have the same gateway, or something else?
It means when two IP addresses have a subnet mask such that the addresses share the same network address and broadcast address. For example:
With classless VLSM being so widespread these days, it's uncommon to refer to subnets in terms of classful ranges like "Class C" anymore.
I do need two connections with same subnet, because I use two separated Infiniband connections.
Here are my solution.
Background:
Step1: Set IP Route and Rule
This method fails after reboot, but you can use "rc.local".
Replacing rc.local in systemd Linux systems
I cannot use nmcli to achieve the same result, can anyone tell me? I try
nmcli connection modify ib1 ipv4.routes "192.168.0.0/16 src=192.168.12.122 table=2"
and edit /etc/sysconfig/network-scripts/route-ib* file directly.Step2: Set ARP
Why is my server responding to arp requests from the wrong interface on Red Hat Enterprise Linux ?
When RHEL has multiple IPs configured, only one is reachable from a remote network.
there is no verified solution