2 Linux Ethernet Bonding Driver HOWTO
4 Latest update: 12 November 2007
6 Initial release : Thomas Davis <tadavis at lbl.gov>
7 Corrections, HA extensions : 2000/10/03-15 :
8 - Willy Tarreau <willy at meta-x.org>
9 - Constantine Gavrilov <const-g at xpert.com>
10 - Chad N. Tindel <ctindel at ieee dot org>
11 - Janice Girouard <girouard at us dot ibm dot com>
12 - Jay Vosburgh <fubar at us dot ibm dot com>
14 Reorganized and updated Feb 2005 by Jay Vosburgh
15 Added Sysfs information: 2006/04/24
16 - Mitch Williams <mitch.a.williams at intel.com>
21 The Linux bonding driver provides a method for aggregating
22 multiple network interfaces into a single logical "bonded" interface.
23 The behavior of the bonded interfaces depends upon the mode; generally
24 speaking, modes provide either hot standby or load balancing services.
25 Additionally, link integrity monitoring may be performed.
27 The bonding driver originally came from Donald Becker's
28 beowulf patches for kernel 2.0. It has changed quite a bit since, and
29 the original tools from extreme-linux and beowulf sites will not work
30 with this version of the driver.
32 For new versions of the driver, updated userspace tools, and
33 who to ask for help, please follow the links at the end of this file.
38 1. Bonding Driver Installation
40 2. Bonding Driver Options
42 3. Configuring Bonding Devices
43 3.1 Configuration with Sysconfig Support
44 3.1.1 Using DHCP with Sysconfig
45 3.1.2 Configuring Multiple Bonds with Sysconfig
46 3.2 Configuration with Initscripts Support
47 3.2.1 Using DHCP with Initscripts
48 3.2.2 Configuring Multiple Bonds with Initscripts
49 3.3 Configuring Bonding Manually with Ifenslave
50 3.3.1 Configuring Multiple Bonds Manually
51 3.4 Configuring Bonding Manually via Sysfs
53 4. Querying Bonding Configuration
54 4.1 Bonding Configuration
55 4.2 Network Configuration
57 5. Switch Configuration
59 6. 802.1q VLAN Support
62 7.1 ARP Monitor Operation
63 7.2 Configuring Multiple ARP Targets
64 7.3 MII Monitor Operation
66 8. Potential Trouble Sources
67 8.1 Adventures in Routing
68 8.2 Ethernet Device Renaming
69 8.3 Painfully Slow Or No Failed Link Detection By Miimon
75 11. Configuring Bonding for High Availability
76 11.1 High Availability in a Single Switch Topology
77 11.2 High Availability in a Multiple Switch Topology
78 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
79 11.2.2 HA Link Monitoring for Multiple Switch Topology
81 12. Configuring Bonding for Maximum Throughput
82 12.1 Maximum Throughput in a Single Switch Topology
83 12.1.1 MT Bonding Mode Selection for Single Switch Topology
84 12.1.2 MT Link Monitoring for Single Switch Topology
85 12.2 Maximum Throughput in a Multiple Switch Topology
86 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
87 12.2.2 MT Link Monitoring for Multiple Switch Topology
89 13. Switch Behavior Issues
90 13.1 Link Establishment and Failover Delays
91 13.2 Duplicated Incoming Packets
93 14. Hardware Specific Considerations
96 15. Frequently Asked Questions
98 16. Resources and Links
101 1. Bonding Driver Installation
102 ==============================
104 Most popular distro kernels ship with the bonding driver
105 already available as a module and the ifenslave user level control
106 program installed and ready for use. If your distro does not, or you
107 have need to compile bonding from source (e.g., configuring and
108 installing a mainline kernel from kernel.org), you'll need to perform
111 1.1 Configure and build the kernel with bonding
112 -----------------------------------------------
114 The current version of the bonding driver is available in the
115 drivers/net/bonding subdirectory of the most recent kernel source
116 (which is available on http://kernel.org). Most users "rolling their
117 own" will want to use the most recent kernel from kernel.org.
119 Configure kernel with "make menuconfig" (or "make xconfig" or
120 "make config"), then select "Bonding driver support" in the "Network
121 device support" section. It is recommended that you configure the
122 driver as module since it is currently the only way to pass parameters
123 to the driver or configure more than one bonding device.
125 Build and install the new kernel and modules, then continue
126 below to install ifenslave.
128 1.2 Install ifenslave Control Utility
129 -------------------------------------
131 The ifenslave user level control program is included in the
132 kernel source tree, in the file Documentation/networking/ifenslave.c.
133 It is generally recommended that you use the ifenslave that
134 corresponds to the kernel that you are using (either from the same
135 source tree or supplied with the distro), however, ifenslave
136 executables from older kernels should function (but features newer
137 than the ifenslave release are not supported). Running an ifenslave
138 that is newer than the kernel is not supported, and may or may not
141 To install ifenslave, do the following:
143 # gcc -Wall -O -I/usr/src/linux/include ifenslave.c -o ifenslave
144 # cp ifenslave /sbin/ifenslave
146 If your kernel source is not in "/usr/src/linux," then replace
147 "/usr/src/linux/include" in the above with the location of your kernel
148 source include directory.
150 You may wish to back up any existing /sbin/ifenslave, or, for
151 testing or informal use, tag the ifenslave to the kernel version
152 (e.g., name the ifenslave executable /sbin/ifenslave-2.6.10).
156 If you omit the "-I" or specify an incorrect directory, you
157 may end up with an ifenslave that is incompatible with the kernel
158 you're trying to build it for. Some distros (e.g., Red Hat from 7.1
159 onwards) do not have /usr/include/linux symbolically linked to the
160 default kernel source include directory.
162 SECOND IMPORTANT NOTE:
163 If you plan to configure bonding using sysfs, you do not need
166 2. Bonding Driver Options
167 =========================
169 Options for the bonding driver are supplied as parameters to the
170 bonding module at load time, or are specified via sysfs.
172 Module options may be given as command line arguments to the
173 insmod or modprobe command, but are usually specified in either the
174 /etc/modules.conf or /etc/modprobe.conf configuration file, or in a
175 distro-specific configuration file (some of which are detailed in the next
178 Details on bonding support for sysfs is provided in the
179 "Configuring Bonding Manually via Sysfs" section, below.
181 The available bonding driver parameters are listed below. If a
182 parameter is not specified the default value is used. When initially
183 configuring a bond, it is recommended "tail -f /var/log/messages" be
184 run in a separate window to watch for bonding driver error messages.
186 It is critical that either the miimon or arp_interval and
187 arp_ip_target parameters be specified, otherwise serious network
188 degradation will occur during link failures. Very few devices do not
189 support at least miimon, so there is really no reason not to use it.
191 Options with textual values will accept either the text name
192 or, for backwards compatibility, the option value. E.g.,
193 "mode=802.3ad" and "mode=4" set the same mode.
195 The parameters are as follows:
199 Specifies the ARP link monitoring frequency in milliseconds.
201 The ARP monitor works by periodically checking the slave
202 devices to determine whether they have sent or received
203 traffic recently (the precise criteria depends upon the
204 bonding mode, and the state of the slave). Regular traffic is
205 generated via ARP probes issued for the addresses specified by
206 the arp_ip_target option.
208 This behavior can be modified by the arp_validate option,
211 If ARP monitoring is used in an etherchannel compatible mode
212 (modes 0 and 2), the switch should be configured in a mode
213 that evenly distributes packets across all links. If the
214 switch is configured to distribute the packets in an XOR
215 fashion, all replies from the ARP targets will be received on
216 the same link which could cause the other team members to
217 fail. ARP monitoring should not be used in conjunction with
218 miimon. A value of 0 disables ARP monitoring. The default
223 Specifies the IP addresses to use as ARP monitoring peers when
224 arp_interval is > 0. These are the targets of the ARP request
225 sent to determine the health of the link to the targets.
226 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
227 addresses must be separated by a comma. At least one IP
228 address must be given for ARP monitoring to function. The
229 maximum number of targets that can be specified is 16. The
230 default value is no IP addresses.
234 Specifies whether or not ARP probes and replies should be
235 validated in the active-backup mode. This causes the ARP
236 monitor to examine the incoming ARP requests and replies, and
237 only consider a slave to be up if it is receiving the
238 appropriate ARP traffic.
244 No validation is performed. This is the default.
248 Validation is performed only for the active slave.
252 Validation is performed only for backup slaves.
256 Validation is performed for all slaves.
258 For the active slave, the validation checks ARP replies to
259 confirm that they were generated by an arp_ip_target. Since
260 backup slaves do not typically receive these replies, the
261 validation performed for backup slaves is on the ARP request
262 sent out via the active slave. It is possible that some
263 switch or network configurations may result in situations
264 wherein the backup slaves do not receive the ARP requests; in
265 such a situation, validation of backup slaves must be
268 This option is useful in network configurations in which
269 multiple bonding hosts are concurrently issuing ARPs to one or
270 more targets beyond a common switch. Should the link between
271 the switch and target fail (but not the switch itself), the
272 probe traffic generated by the multiple bonding instances will
273 fool the standard ARP monitor into considering the links as
274 still up. Use of the arp_validate option can resolve this, as
275 the ARP monitor will only consider ARP requests and replies
276 associated with its own instance of bonding.
278 This option was added in bonding version 3.1.0.
282 Specifies the time, in milliseconds, to wait before disabling
283 a slave after a link failure has been detected. This option
284 is only valid for the miimon link monitor. The downdelay
285 value should be a multiple of the miimon value; if not, it
286 will be rounded down to the nearest multiple. The default
291 Specifies whether active-backup mode should set all slaves to
292 the same MAC address at enslavement (the traditional
293 behavior), or, when enabled, perform special handling of the
294 bond's MAC address in accordance with the selected policy.
300 This setting disables fail_over_mac, and causes
301 bonding to set all slaves of an active-backup bond to
302 the same MAC address at enslavement time. This is the
307 The "active" fail_over_mac policy indicates that the
308 MAC address of the bond should always be the MAC
309 address of the currently active slave. The MAC
310 address of the slaves is not changed; instead, the MAC
311 address of the bond changes during a failover.
313 This policy is useful for devices that cannot ever
314 alter their MAC address, or for devices that refuse
315 incoming broadcasts with their own source MAC (which
316 interferes with the ARP monitor).
318 The down side of this policy is that every device on
319 the network must be updated via gratuitous ARP,
320 vs. just updating a switch or set of switches (which
321 often takes place for any traffic, not just ARP
322 traffic, if the switch snoops incoming traffic to
323 update its tables) for the traditional method. If the
324 gratuitous ARP is lost, communication may be
327 When this policy is used in conjuction with the mii
328 monitor, devices which assert link up prior to being
329 able to actually transmit and receive are particularly
330 susecptible to loss of the gratuitous ARP, and an
331 appropriate updelay setting may be required.
335 The "follow" fail_over_mac policy causes the MAC
336 address of the bond to be selected normally (normally
337 the MAC address of the first slave added to the bond).
338 However, the second and subsequent slaves are not set
339 to this MAC address while they are in a backup role; a
340 slave is programmed with the bond's MAC address at
341 failover time (and the formerly active slave receives
342 the newly active slave's MAC address).
344 This policy is useful for multiport devices that
345 either become confused or incur a performance penalty
346 when multiple ports are programmed with the same MAC
350 The default policy is none, unless the first slave cannot
351 change its MAC address, in which case the active policy is
354 This option may be modified via sysfs only when no slaves are
357 This option was added in bonding version 3.2.0. The "follow"
358 policy was added in bonding version 3.3.0.
362 Option specifying the rate in which we'll ask our link partner
363 to transmit LACPDU packets in 802.3ad mode. Possible values
367 Request partner to transmit LACPDUs every 30 seconds
370 Request partner to transmit LACPDUs every 1 second
376 Specifies the number of bonding devices to create for this
377 instance of the bonding driver. E.g., if max_bonds is 3, and
378 the bonding driver is not already loaded, then bond0, bond1
379 and bond2 will be created. The default value is 1. Specifying
380 a value of 0 will load bonding, but will not create any devices.
384 Specifies the MII link monitoring frequency in milliseconds.
385 This determines how often the link state of each slave is
386 inspected for link failures. A value of zero disables MII
387 link monitoring. A value of 100 is a good starting point.
388 The use_carrier option, below, affects how the link state is
389 determined. See the High Availability section for additional
390 information. The default value is 0.
394 Specifies one of the bonding policies. The default is
395 balance-rr (round robin). Possible values are:
399 Round-robin policy: Transmit packets in sequential
400 order from the first available slave through the
401 last. This mode provides load balancing and fault
406 Active-backup policy: Only one slave in the bond is
407 active. A different slave becomes active if, and only
408 if, the active slave fails. The bond's MAC address is
409 externally visible on only one port (network adapter)
410 to avoid confusing the switch.
412 In bonding version 2.6.2 or later, when a failover
413 occurs in active-backup mode, bonding will issue one
414 or more gratuitous ARPs on the newly active slave.
415 One gratuitous ARP is issued for the bonding master
416 interface and each VLAN interfaces configured above
417 it, provided that the interface has at least one IP
418 address configured. Gratuitous ARPs issued for VLAN
419 interfaces are tagged with the appropriate VLAN id.
421 This mode provides fault tolerance. The primary
422 option, documented below, affects the behavior of this
427 XOR policy: Transmit based on the selected transmit
428 hash policy. The default policy is a simple [(source
429 MAC address XOR'd with destination MAC address) modulo
430 slave count]. Alternate transmit policies may be
431 selected via the xmit_hash_policy option, described
434 This mode provides load balancing and fault tolerance.
438 Broadcast policy: transmits everything on all slave
439 interfaces. This mode provides fault tolerance.
443 IEEE 802.3ad Dynamic link aggregation. Creates
444 aggregation groups that share the same speed and
445 duplex settings. Utilizes all slaves in the active
446 aggregator according to the 802.3ad specification.
448 Slave selection for outgoing traffic is done according
449 to the transmit hash policy, which may be changed from
450 the default simple XOR policy via the xmit_hash_policy
451 option, documented below. Note that not all transmit
452 policies may be 802.3ad compliant, particularly in
453 regards to the packet mis-ordering requirements of
454 section 43.2.4 of the 802.3ad standard. Differing
455 peer implementations will have varying tolerances for
460 1. Ethtool support in the base drivers for retrieving
461 the speed and duplex of each slave.
463 2. A switch that supports IEEE 802.3ad Dynamic link
466 Most switches will require some type of configuration
467 to enable 802.3ad mode.
471 Adaptive transmit load balancing: channel bonding that
472 does not require any special switch support. The
473 outgoing traffic is distributed according to the
474 current load (computed relative to the speed) on each
475 slave. Incoming traffic is received by the current
476 slave. If the receiving slave fails, another slave
477 takes over the MAC address of the failed receiving
482 Ethtool support in the base drivers for retrieving the
487 Adaptive load balancing: includes balance-tlb plus
488 receive load balancing (rlb) for IPV4 traffic, and
489 does not require any special switch support. The
490 receive load balancing is achieved by ARP negotiation.
491 The bonding driver intercepts the ARP Replies sent by
492 the local system on their way out and overwrites the
493 source hardware address with the unique hardware
494 address of one of the slaves in the bond such that
495 different peers use different hardware addresses for
498 Receive traffic from connections created by the server
499 is also balanced. When the local system sends an ARP
500 Request the bonding driver copies and saves the peer's
501 IP information from the ARP packet. When the ARP
502 Reply arrives from the peer, its hardware address is
503 retrieved and the bonding driver initiates an ARP
504 reply to this peer assigning it to one of the slaves
505 in the bond. A problematic outcome of using ARP
506 negotiation for balancing is that each time that an
507 ARP request is broadcast it uses the hardware address
508 of the bond. Hence, peers learn the hardware address
509 of the bond and the balancing of receive traffic
510 collapses to the current slave. This is handled by
511 sending updates (ARP Replies) to all the peers with
512 their individually assigned hardware address such that
513 the traffic is redistributed. Receive traffic is also
514 redistributed when a new slave is added to the bond
515 and when an inactive slave is re-activated. The
516 receive load is distributed sequentially (round robin)
517 among the group of highest speed slaves in the bond.
519 When a link is reconnected or a new slave joins the
520 bond the receive traffic is redistributed among all
521 active slaves in the bond by initiating ARP Replies
522 with the selected MAC address to each of the
523 clients. The updelay parameter (detailed below) must
524 be set to a value equal or greater than the switch's
525 forwarding delay so that the ARP Replies sent to the
526 peers will not be blocked by the switch.
530 1. Ethtool support in the base drivers for retrieving
531 the speed of each slave.
533 2. Base driver support for setting the hardware
534 address of a device while it is open. This is
535 required so that there will always be one slave in the
536 team using the bond hardware address (the
537 curr_active_slave) while having a unique hardware
538 address for each slave in the bond. If the
539 curr_active_slave fails its hardware address is
540 swapped with the new curr_active_slave that was
545 Specifies the number of gratuitous ARPs to be issued after a
546 failover event. One gratuitous ARP is issued immediately after
547 the failover, subsequent ARPs are sent at a rate of one per link
548 monitor interval (arp_interval or miimon, whichever is active).
550 The valid range is 0 - 255; the default value is 1. This option
551 affects only the active-backup mode. This option was added for
552 bonding version 3.3.0.
556 Specifies the number of unsolicited IPv6 Neighbor Advertisements
557 to be issued after a failover event. One unsolicited NA is issued
558 immediately after the failover.
560 The valid range is 0 - 255; the default value is 1. This option
561 affects only the active-backup mode. This option was added for
562 bonding version 3.4.0.
566 A string (eth0, eth2, etc) specifying which slave is the
567 primary device. The specified device will always be the
568 active slave while it is available. Only when the primary is
569 off-line will alternate devices be used. This is useful when
570 one slave is preferred over another, e.g., when one slave has
571 higher throughput than another.
573 The primary option is only valid for active-backup mode.
577 Specifies the time, in milliseconds, to wait before enabling a
578 slave after a link recovery has been detected. This option is
579 only valid for the miimon link monitor. The updelay value
580 should be a multiple of the miimon value; if not, it will be
581 rounded down to the nearest multiple. The default value is 0.
585 Specifies whether or not miimon should use MII or ETHTOOL
586 ioctls vs. netif_carrier_ok() to determine the link
587 status. The MII or ETHTOOL ioctls are less efficient and
588 utilize a deprecated calling sequence within the kernel. The
589 netif_carrier_ok() relies on the device driver to maintain its
590 state with netif_carrier_on/off; at this writing, most, but
591 not all, device drivers support this facility.
593 If bonding insists that the link is up when it should not be,
594 it may be that your network device driver does not support
595 netif_carrier_on/off. The default state for netif_carrier is
596 "carrier on," so if a driver does not support netif_carrier,
597 it will appear as if the link is always up. In this case,
598 setting use_carrier to 0 will cause bonding to revert to the
599 MII / ETHTOOL ioctl method to determine the link state.
601 A value of 1 enables the use of netif_carrier_ok(), a value of
602 0 will use the deprecated MII / ETHTOOL ioctls. The default
607 Selects the transmit hash policy to use for slave selection in
608 balance-xor and 802.3ad modes. Possible values are:
612 Uses XOR of hardware MAC addresses to generate the
615 (source MAC XOR destination MAC) modulo slave count
617 This algorithm will place all traffic to a particular
618 network peer on the same slave.
620 This algorithm is 802.3ad compliant.
624 This policy uses a combination of layer2 and layer3
625 protocol information to generate the hash.
627 Uses XOR of hardware MAC addresses and IP addresses to
628 generate the hash. The formula is
630 (((source IP XOR dest IP) AND 0xffff) XOR
631 ( source MAC XOR destination MAC ))
634 This algorithm will place all traffic to a particular
635 network peer on the same slave. For non-IP traffic,
636 the formula is the same as for the layer2 transmit
639 This policy is intended to provide a more balanced
640 distribution of traffic than layer2 alone, especially
641 in environments where a layer3 gateway device is
642 required to reach most destinations.
644 This algorithm is 802.3ad compliant.
648 This policy uses upper layer protocol information,
649 when available, to generate the hash. This allows for
650 traffic to a particular network peer to span multiple
651 slaves, although a single connection will not span
654 The formula for unfragmented TCP and UDP packets is
656 ((source port XOR dest port) XOR
657 ((source IP XOR dest IP) AND 0xffff)
660 For fragmented TCP or UDP packets and all other IP
661 protocol traffic, the source and destination port
662 information is omitted. For non-IP traffic, the
663 formula is the same as for the layer2 transmit hash
666 This policy is intended to mimic the behavior of
667 certain switches, notably Cisco switches with PFC2 as
668 well as some Foundry and IBM products.
670 This algorithm is not fully 802.3ad compliant. A
671 single TCP or UDP conversation containing both
672 fragmented and unfragmented packets will see packets
673 striped across two interfaces. This may result in out
674 of order delivery. Most traffic types will not meet
675 this criteria, as TCP rarely fragments traffic, and
676 most UDP traffic is not involved in extended
677 conversations. Other implementations of 802.3ad may
678 or may not tolerate this noncompliance.
680 The default value is layer2. This option was added in bonding
681 version 2.6.3. In earlier versions of bonding, this parameter
682 does not exist, and the layer2 policy is the only policy. The
683 layer2+3 value was added for bonding version 3.2.2.
686 3. Configuring Bonding Devices
687 ==============================
689 You can configure bonding using either your distro's network
690 initialization scripts, or manually using either ifenslave or the
691 sysfs interface. Distros generally use one of two packages for the
692 network initialization scripts: initscripts or sysconfig. Recent
693 versions of these packages have support for bonding, while older
696 We will first describe the options for configuring bonding for
697 distros using versions of initscripts and sysconfig with full or
698 partial support for bonding, then provide information on enabling
699 bonding without support from the network initialization scripts (i.e.,
700 older versions of initscripts or sysconfig).
702 If you're unsure whether your distro uses sysconfig or
703 initscripts, or don't know if it's new enough, have no fear.
704 Determining this is fairly straightforward.
706 First, issue the command:
710 It will respond with a line of text starting with either
711 "initscripts" or "sysconfig," followed by some numbers. This is the
712 package that provides your network initialization scripts.
714 Next, to determine if your installation supports bonding,
717 $ grep ifenslave /sbin/ifup
719 If this returns any matches, then your initscripts or
720 sysconfig has support for bonding.
722 3.1 Configuration with Sysconfig Support
723 ----------------------------------------
725 This section applies to distros using a version of sysconfig
726 with bonding support, for example, SuSE Linux Enterprise Server 9.
728 SuSE SLES 9's networking configuration system does support
729 bonding, however, at this writing, the YaST system configuration
730 front end does not provide any means to work with bonding devices.
731 Bonding devices can be managed by hand, however, as follows.
733 First, if they have not already been configured, configure the
734 slave devices. On SLES 9, this is most easily done by running the
735 yast2 sysconfig configuration utility. The goal is for to create an
736 ifcfg-id file for each slave device. The simplest way to accomplish
737 this is to configure the devices for DHCP (this is only to get the
738 file ifcfg-id file created; see below for some issues with DHCP). The
739 name of the configuration file for each device will be of the form:
741 ifcfg-id-xx:xx:xx:xx:xx:xx
743 Where the "xx" portion will be replaced with the digits from
744 the device's permanent MAC address.
746 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
747 created, it is necessary to edit the configuration files for the slave
748 devices (the MAC addresses correspond to those of the slave devices).
749 Before editing, the file will contain multiple lines, and will look
755 UNIQUE='XNzu.WeZGOGF+4wE'
756 _nm_name='bus-pci-0001:61:01.0'
758 Change the BOOTPROTO and STARTMODE lines to the following:
763 Do not alter the UNIQUE or _nm_name lines. Remove any other
764 lines (USERCTL, etc).
766 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
767 it's time to create the configuration file for the bonding device
768 itself. This file is named ifcfg-bondX, where X is the number of the
769 bonding device to create, starting at 0. The first such file is
770 ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
771 network configuration system will correctly start multiple instances
774 The contents of the ifcfg-bondX file is as follows:
777 BROADCAST="10.0.2.255"
779 NETMASK="255.255.0.0"
784 BONDING_MODULE_OPTS="mode=active-backup miimon=100"
785 BONDING_SLAVE0="eth0"
786 BONDING_SLAVE1="bus-pci-0000:06:08.1"
788 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
789 values with the appropriate values for your network.
791 The STARTMODE specifies when the device is brought online.
792 The possible values are:
794 onboot: The device is started at boot time. If you're not
795 sure, this is probably what you want.
797 manual: The device is started only when ifup is called
798 manually. Bonding devices may be configured this
799 way if you do not wish them to start automatically
800 at boot for some reason.
802 hotplug: The device is started by a hotplug event. This is not
803 a valid choice for a bonding device.
805 off or ignore: The device configuration is ignored.
807 The line BONDING_MASTER='yes' indicates that the device is a
808 bonding master device. The only useful value is "yes."
810 The contents of BONDING_MODULE_OPTS are supplied to the
811 instance of the bonding module for this device. Specify the options
812 for the bonding mode, link monitoring, and so on here. Do not include
813 the max_bonds bonding parameter; this will confuse the configuration
814 system if you have multiple bonding devices.
816 Finally, supply one BONDING_SLAVEn="slave device" for each
817 slave. where "n" is an increasing value, one for each slave. The
818 "slave device" is either an interface name, e.g., "eth0", or a device
819 specifier for the network device. The interface name is easier to
820 find, but the ethN names are subject to change at boot time if, e.g.,
821 a device early in the sequence has failed. The device specifiers
822 (bus-pci-0000:06:08.1 in the example above) specify the physical
823 network device, and will not change unless the device's bus location
824 changes (for example, it is moved from one PCI slot to another). The
825 example above uses one of each type for demonstration purposes; most
826 configurations will choose one or the other for all slave devices.
828 When all configuration files have been modified or created,
829 networking must be restarted for the configuration changes to take
830 effect. This can be accomplished via the following:
832 # /etc/init.d/network restart
834 Note that the network control script (/sbin/ifdown) will
835 remove the bonding module as part of the network shutdown processing,
836 so it is not necessary to remove the module by hand if, e.g., the
837 module parameters have changed.
839 Also, at this writing, YaST/YaST2 will not manage bonding
840 devices (they do not show bonding interfaces on its list of network
841 devices). It is necessary to edit the configuration file by hand to
842 change the bonding configuration.
844 Additional general options and details of the ifcfg file
845 format can be found in an example ifcfg template file:
847 /etc/sysconfig/network/ifcfg.template
849 Note that the template does not document the various BONDING_
850 settings described above, but does describe many of the other options.
852 3.1.1 Using DHCP with Sysconfig
853 -------------------------------
855 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
856 will cause it to query DHCP for its IP address information. At this
857 writing, this does not function for bonding devices; the scripts
858 attempt to obtain the device address from DHCP prior to adding any of
859 the slave devices. Without active slaves, the DHCP requests are not
862 3.1.2 Configuring Multiple Bonds with Sysconfig
863 -----------------------------------------------
865 The sysconfig network initialization system is capable of
866 handling multiple bonding devices. All that is necessary is for each
867 bonding instance to have an appropriately configured ifcfg-bondX file
868 (as described above). Do not specify the "max_bonds" parameter to any
869 instance of bonding, as this will confuse sysconfig. If you require
870 multiple bonding devices with identical parameters, create multiple
873 Because the sysconfig scripts supply the bonding module
874 options in the ifcfg-bondX file, it is not necessary to add them to
875 the system /etc/modules.conf or /etc/modprobe.conf configuration file.
877 3.2 Configuration with Initscripts Support
878 ------------------------------------------
880 This section applies to distros using a recent version of
881 initscripts with bonding support, for example, Red Hat Enterprise Linux
882 version 3 or later, Fedora, etc. On these systems, the network
883 initialization scripts have knowledge of bonding, and can be configured to
884 control bonding devices. Note that older versions of the initscripts
885 package have lower levels of support for bonding; this will be noted where
888 These distros will not automatically load the network adapter
889 driver unless the ethX device is configured with an IP address.
890 Because of this constraint, users must manually configure a
891 network-script file for all physical adapters that will be members of
892 a bondX link. Network script files are located in the directory:
894 /etc/sysconfig/network-scripts
896 The file name must be prefixed with "ifcfg-eth" and suffixed
897 with the adapter's physical adapter number. For example, the script
898 for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
899 Place the following text in the file:
908 The DEVICE= line will be different for every ethX device and
909 must correspond with the name of the file, i.e., ifcfg-eth1 must have
910 a device line of DEVICE=eth1. The setting of the MASTER= line will
911 also depend on the final bonding interface name chosen for your bond.
912 As with other network devices, these typically start at 0, and go up
913 one for each device, i.e., the first bonding instance is bond0, the
914 second is bond1, and so on.
916 Next, create a bond network script. The file name for this
917 script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
918 the number of the bond. For bond0 the file is named "ifcfg-bond0",
919 for bond1 it is named "ifcfg-bond1", and so on. Within that file,
920 place the following text:
924 NETMASK=255.255.255.0
926 BROADCAST=192.168.1.255
931 Be sure to change the networking specific lines (IPADDR,
932 NETMASK, NETWORK and BROADCAST) to match your network configuration.
934 For later versions of initscripts, such as that found with Fedora
935 7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
936 and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
937 file, e.g. a line of the format:
939 BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
941 will configure the bond with the specified options. The options
942 specified in BONDING_OPTS are identical to the bonding module parameters
943 except for the arp_ip_target field when using versions of initscripts older
944 than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
945 using older versions each target should be included as a separate option and
946 should be preceded by a '+' to indicate it should be added to the list of
947 queried targets, e.g.,
949 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
951 is the proper syntax to specify multiple targets. When specifying
952 options via BONDING_OPTS, it is not necessary to edit /etc/modules.conf or
955 For even older versions of initscripts that do not support
956 BONDING_OPTS, it is necessary to edit /etc/modules.conf (or
957 /etc/modprobe.conf, depending upon your distro) to load the bonding module
958 with your desired options when the bond0 interface is brought up. The
959 following lines in /etc/modules.conf (or modprobe.conf) will load the
960 bonding module, and select its options:
963 options bond0 mode=balance-alb miimon=100
965 Replace the sample parameters with the appropriate set of
966 options for your configuration.
968 Finally run "/etc/rc.d/init.d/network restart" as root. This
969 will restart the networking subsystem and your bond link should be now
972 3.2.1 Using DHCP with Initscripts
973 ---------------------------------
975 Recent versions of initscripts (the versions supplied with Fedora
976 Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
977 work) have support for assigning IP information to bonding devices via
980 To configure bonding for DHCP, configure it as described
981 above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
982 and add a line consisting of "TYPE=Bonding". Note that the TYPE value
985 3.2.2 Configuring Multiple Bonds with Initscripts
986 -------------------------------------------------
988 Initscripts packages that are included with Fedora 7 and Red Hat
989 Enterprise Linux 5 support multiple bonding interfaces by simply
990 specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
991 number of the bond. This support requires sysfs support in the kernel,
992 and a bonding driver of version 3.0.0 or later. Other configurations may
993 not support this method for specifying multiple bonding interfaces; for
994 those instances, see the "Configuring Multiple Bonds Manually" section,
997 3.3 Configuring Bonding Manually with Ifenslave
998 -----------------------------------------------
1000 This section applies to distros whose network initialization
1001 scripts (the sysconfig or initscripts package) do not have specific
1002 knowledge of bonding. One such distro is SuSE Linux Enterprise Server
1005 The general method for these systems is to place the bonding
1006 module parameters into /etc/modules.conf or /etc/modprobe.conf (as
1007 appropriate for the installed distro), then add modprobe and/or
1008 ifenslave commands to the system's global init script. The name of
1009 the global init script differs; for sysconfig, it is
1010 /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1012 For example, if you wanted to make a simple bond of two e100
1013 devices (presumed to be eth0 and eth1), and have it persist across
1014 reboots, edit the appropriate file (/etc/init.d/boot.local or
1015 /etc/rc.d/rc.local), and add the following:
1017 modprobe bonding mode=balance-alb miimon=100
1019 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1020 ifenslave bond0 eth0
1021 ifenslave bond0 eth1
1023 Replace the example bonding module parameters and bond0
1024 network configuration (IP address, netmask, etc) with the appropriate
1025 values for your configuration.
1027 Unfortunately, this method will not provide support for the
1028 ifup and ifdown scripts on the bond devices. To reload the bonding
1029 configuration, it is necessary to run the initialization script, e.g.,
1031 # /etc/init.d/boot.local
1035 # /etc/rc.d/rc.local
1037 It may be desirable in such a case to create a separate script
1038 which only initializes the bonding configuration, then call that
1039 separate script from within boot.local. This allows for bonding to be
1040 enabled without re-running the entire global init script.
1042 To shut down the bonding devices, it is necessary to first
1043 mark the bonding device itself as being down, then remove the
1044 appropriate device driver modules. For our example above, you can do
1047 # ifconfig bond0 down
1051 Again, for convenience, it may be desirable to create a script
1052 with these commands.
1055 3.3.1 Configuring Multiple Bonds Manually
1056 -----------------------------------------
1058 This section contains information on configuring multiple
1059 bonding devices with differing options for those systems whose network
1060 initialization scripts lack support for configuring multiple bonds.
1062 If you require multiple bonding devices, but all with the same
1063 options, you may wish to use the "max_bonds" module parameter,
1066 To create multiple bonding devices with differing options, it is
1067 preferrable to use bonding parameters exported by sysfs, documented in the
1070 For versions of bonding without sysfs support, the only means to
1071 provide multiple instances of bonding with differing options is to load
1072 the bonding driver multiple times. Note that current versions of the
1073 sysconfig network initialization scripts handle this automatically; if
1074 your distro uses these scripts, no special action is needed. See the
1075 section Configuring Bonding Devices, above, if you're not sure about your
1076 network initialization scripts.
1078 To load multiple instances of the module, it is necessary to
1079 specify a different name for each instance (the module loading system
1080 requires that every loaded module, even multiple instances of the same
1081 module, have a unique name). This is accomplished by supplying multiple
1082 sets of bonding options in /etc/modprobe.conf, for example:
1085 options bond0 -o bond0 mode=balance-rr miimon=100
1088 options bond1 -o bond1 mode=balance-alb miimon=50
1090 will load the bonding module two times. The first instance is
1091 named "bond0" and creates the bond0 device in balance-rr mode with an
1092 miimon of 100. The second instance is named "bond1" and creates the
1093 bond1 device in balance-alb mode with an miimon of 50.
1095 In some circumstances (typically with older distributions),
1096 the above does not work, and the second bonding instance never sees
1097 its options. In that case, the second options line can be substituted
1100 install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1101 mode=balance-alb miimon=50
1103 This may be repeated any number of times, specifying a new and
1104 unique name in place of bond1 for each subsequent instance.
1106 It has been observed that some Red Hat supplied kernels are unable
1107 to rename modules at load time (the "-o bond1" part). Attempts to pass
1108 that option to modprobe will produce an "Operation not permitted" error.
1109 This has been reported on some Fedora Core kernels, and has been seen on
1110 RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1111 to configure multiple bonds with differing parameters (as they are older
1112 kernels, and also lack sysfs support).
1114 3.4 Configuring Bonding Manually via Sysfs
1115 ------------------------------------------
1117 Starting with version 3.0.0, Channel Bonding may be configured
1118 via the sysfs interface. This interface allows dynamic configuration
1119 of all bonds in the system without unloading the module. It also
1120 allows for adding and removing bonds at runtime. Ifenslave is no
1121 longer required, though it is still supported.
1123 Use of the sysfs interface allows you to use multiple bonds
1124 with different configurations without having to reload the module.
1125 It also allows you to use multiple, differently configured bonds when
1126 bonding is compiled into the kernel.
1128 You must have the sysfs filesystem mounted to configure
1129 bonding this way. The examples in this document assume that you
1130 are using the standard mount point for sysfs, e.g. /sys. If your
1131 sysfs filesystem is mounted elsewhere, you will need to adjust the
1132 example paths accordingly.
1134 Creating and Destroying Bonds
1135 -----------------------------
1136 To add a new bond foo:
1137 # echo +foo > /sys/class/net/bonding_masters
1139 To remove an existing bond bar:
1140 # echo -bar > /sys/class/net/bonding_masters
1142 To show all existing bonds:
1143 # cat /sys/class/net/bonding_masters
1145 NOTE: due to 4K size limitation of sysfs files, this list may be
1146 truncated if you have more than a few hundred bonds. This is unlikely
1147 to occur under normal operating conditions.
1149 Adding and Removing Slaves
1150 --------------------------
1151 Interfaces may be enslaved to a bond using the file
1152 /sys/class/net/<bond>/bonding/slaves. The semantics for this file
1153 are the same as for the bonding_masters file.
1155 To enslave interface eth0 to bond bond0:
1157 # echo +eth0 > /sys/class/net/bond0/bonding/slaves
1159 To free slave eth0 from bond bond0:
1160 # echo -eth0 > /sys/class/net/bond0/bonding/slaves
1162 When an interface is enslaved to a bond, symlinks between the
1163 two are created in the sysfs filesystem. In this case, you would get
1164 /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1165 /sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1167 This means that you can tell quickly whether or not an
1168 interface is enslaved by looking for the master symlink. Thus:
1169 # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1170 will free eth0 from whatever bond it is enslaved to, regardless of
1171 the name of the bond interface.
1173 Changing a Bond's Configuration
1174 -------------------------------
1175 Each bond may be configured individually by manipulating the
1176 files located in /sys/class/net/<bond name>/bonding
1178 The names of these files correspond directly with the command-
1179 line parameters described elsewhere in this file, and, with the
1180 exception of arp_ip_target, they accept the same values. To see the
1181 current setting, simply cat the appropriate file.
1183 A few examples will be given here; for specific usage
1184 guidelines for each parameter, see the appropriate section in this
1187 To configure bond0 for balance-alb mode:
1188 # ifconfig bond0 down
1189 # echo 6 > /sys/class/net/bond0/bonding/mode
1191 # echo balance-alb > /sys/class/net/bond0/bonding/mode
1192 NOTE: The bond interface must be down before the mode can be
1195 To enable MII monitoring on bond0 with a 1 second interval:
1196 # echo 1000 > /sys/class/net/bond0/bonding/miimon
1197 NOTE: If ARP monitoring is enabled, it will disabled when MII
1198 monitoring is enabled, and vice-versa.
1201 # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1202 # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1203 NOTE: up to 10 target addresses may be specified.
1205 To remove an ARP target:
1206 # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1208 Example Configuration
1209 ---------------------
1210 We begin with the same example that is shown in section 3.3,
1211 executed with sysfs, and without using ifenslave.
1213 To make a simple bond of two e100 devices (presumed to be eth0
1214 and eth1), and have it persist across reboots, edit the appropriate
1215 file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1220 echo balance-alb > /sys/class/net/bond0/bonding/mode
1221 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1222 echo 100 > /sys/class/net/bond0/bonding/miimon
1223 echo +eth0 > /sys/class/net/bond0/bonding/slaves
1224 echo +eth1 > /sys/class/net/bond0/bonding/slaves
1226 To add a second bond, with two e1000 interfaces in
1227 active-backup mode, using ARP monitoring, add the following lines to
1231 echo +bond1 > /sys/class/net/bonding_masters
1232 echo active-backup > /sys/class/net/bond1/bonding/mode
1233 ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1234 echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1235 echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1236 echo +eth2 > /sys/class/net/bond1/bonding/slaves
1237 echo +eth3 > /sys/class/net/bond1/bonding/slaves
1240 4. Querying Bonding Configuration
1241 =================================
1243 4.1 Bonding Configuration
1244 -------------------------
1246 Each bonding device has a read-only file residing in the
1247 /proc/net/bonding directory. The file contents include information
1248 about the bonding configuration, options and state of each slave.
1250 For example, the contents of /proc/net/bonding/bond0 after the
1251 driver is loaded with parameters of mode=0 and miimon=1000 is
1252 generally as follows:
1254 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1255 Bonding Mode: load balancing (round-robin)
1256 Currently Active Slave: eth0
1258 MII Polling Interval (ms): 1000
1262 Slave Interface: eth1
1264 Link Failure Count: 1
1266 Slave Interface: eth0
1268 Link Failure Count: 1
1270 The precise format and contents will change depending upon the
1271 bonding configuration, state, and version of the bonding driver.
1273 4.2 Network configuration
1274 -------------------------
1276 The network configuration can be inspected using the ifconfig
1277 command. Bonding devices will have the MASTER flag set; Bonding slave
1278 devices will have the SLAVE flag set. The ifconfig output does not
1279 contain information on which slaves are associated with which masters.
1281 In the example below, the bond0 interface is the master
1282 (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1283 bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1284 TLB and ALB that require a unique MAC address for each slave.
1287 bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1288 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1289 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1290 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1291 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1292 collisions:0 txqueuelen:0
1294 eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1295 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1296 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1297 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1298 collisions:0 txqueuelen:100
1299 Interrupt:10 Base address:0x1080
1301 eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1302 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1303 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1304 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1305 collisions:0 txqueuelen:100
1306 Interrupt:9 Base address:0x1400
1308 5. Switch Configuration
1309 =======================
1311 For this section, "switch" refers to whatever system the
1312 bonded devices are directly connected to (i.e., where the other end of
1313 the cable plugs into). This may be an actual dedicated switch device,
1314 or it may be another regular system (e.g., another computer running
1317 The active-backup, balance-tlb and balance-alb modes do not
1318 require any specific configuration of the switch.
1320 The 802.3ad mode requires that the switch have the appropriate
1321 ports configured as an 802.3ad aggregation. The precise method used
1322 to configure this varies from switch to switch, but, for example, a
1323 Cisco 3550 series switch requires that the appropriate ports first be
1324 grouped together in a single etherchannel instance, then that
1325 etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1326 standard EtherChannel).
1328 The balance-rr, balance-xor and broadcast modes generally
1329 require that the switch have the appropriate ports grouped together.
1330 The nomenclature for such a group differs between switches, it may be
1331 called an "etherchannel" (as in the Cisco example, above), a "trunk
1332 group" or some other similar variation. For these modes, each switch
1333 will also have its own configuration options for the switch's transmit
1334 policy to the bond. Typical choices include XOR of either the MAC or
1335 IP addresses. The transmit policy of the two peers does not need to
1336 match. For these three modes, the bonding mode really selects a
1337 transmit policy for an EtherChannel group; all three will interoperate
1338 with another EtherChannel group.
1341 6. 802.1q VLAN Support
1342 ======================
1344 It is possible to configure VLAN devices over a bond interface
1345 using the 8021q driver. However, only packets coming from the 8021q
1346 driver and passing through bonding will be tagged by default. Self
1347 generated packets, for example, bonding's learning packets or ARP
1348 packets generated by either ALB mode or the ARP monitor mechanism, are
1349 tagged internally by bonding itself. As a result, bonding must
1350 "learn" the VLAN IDs configured above it, and use those IDs to tag
1351 self generated packets.
1353 For reasons of simplicity, and to support the use of adapters
1354 that can do VLAN hardware acceleration offloading, the bonding
1355 interface declares itself as fully hardware offloading capable, it gets
1356 the add_vid/kill_vid notifications to gather the necessary
1357 information, and it propagates those actions to the slaves. In case
1358 of mixed adapter types, hardware accelerated tagged packets that
1359 should go through an adapter that is not offloading capable are
1360 "un-accelerated" by the bonding driver so the VLAN tag sits in the
1363 VLAN interfaces *must* be added on top of a bonding interface
1364 only after enslaving at least one slave. The bonding interface has a
1365 hardware address of 00:00:00:00:00:00 until the first slave is added.
1366 If the VLAN interface is created prior to the first enslavement, it
1367 would pick up the all-zeroes hardware address. Once the first slave
1368 is attached to the bond, the bond device itself will pick up the
1369 slave's hardware address, which is then available for the VLAN device.
1371 Also, be aware that a similar problem can occur if all slaves
1372 are released from a bond that still has one or more VLAN interfaces on
1373 top of it. When a new slave is added, the bonding interface will
1374 obtain its hardware address from the first slave, which might not
1375 match the hardware address of the VLAN interfaces (which was
1376 ultimately copied from an earlier slave).
1378 There are two methods to insure that the VLAN device operates
1379 with the correct hardware address if all slaves are removed from a
1382 1. Remove all VLAN interfaces then recreate them
1384 2. Set the bonding interface's hardware address so that it
1385 matches the hardware address of the VLAN interfaces.
1387 Note that changing a VLAN interface's HW address would set the
1388 underlying device -- i.e. the bonding interface -- to promiscuous
1389 mode, which might not be what you want.
1395 The bonding driver at present supports two schemes for
1396 monitoring a slave device's link state: the ARP monitor and the MII
1399 At the present time, due to implementation restrictions in the
1400 bonding driver itself, it is not possible to enable both ARP and MII
1401 monitoring simultaneously.
1403 7.1 ARP Monitor Operation
1404 -------------------------
1406 The ARP monitor operates as its name suggests: it sends ARP
1407 queries to one or more designated peer systems on the network, and
1408 uses the response as an indication that the link is operating. This
1409 gives some assurance that traffic is actually flowing to and from one
1410 or more peers on the local network.
1412 The ARP monitor relies on the device driver itself to verify
1413 that traffic is flowing. In particular, the driver must keep up to
1414 date the last receive time, dev->last_rx, and transmit start time,
1415 dev->trans_start. If these are not updated by the driver, then the
1416 ARP monitor will immediately fail any slaves using that driver, and
1417 those slaves will stay down. If networking monitoring (tcpdump, etc)
1418 shows the ARP requests and replies on the network, then it may be that
1419 your device driver is not updating last_rx and trans_start.
1421 7.2 Configuring Multiple ARP Targets
1422 ------------------------------------
1424 While ARP monitoring can be done with just one target, it can
1425 be useful in a High Availability setup to have several targets to
1426 monitor. In the case of just one target, the target itself may go
1427 down or have a problem making it unresponsive to ARP requests. Having
1428 an additional target (or several) increases the reliability of the ARP
1431 Multiple ARP targets must be separated by commas as follows:
1433 # example options for ARP monitoring with three targets
1435 options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1437 For just a single target the options would resemble:
1439 # example options for ARP monitoring with one target
1441 options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1444 7.3 MII Monitor Operation
1445 -------------------------
1447 The MII monitor monitors only the carrier state of the local
1448 network interface. It accomplishes this in one of three ways: by
1449 depending upon the device driver to maintain its carrier state, by
1450 querying the device's MII registers, or by making an ethtool query to
1453 If the use_carrier module parameter is 1 (the default value),
1454 then the MII monitor will rely on the driver for carrier state
1455 information (via the netif_carrier subsystem). As explained in the
1456 use_carrier parameter information, above, if the MII monitor fails to
1457 detect carrier loss on the device (e.g., when the cable is physically
1458 disconnected), it may be that the driver does not support
1461 If use_carrier is 0, then the MII monitor will first query the
1462 device's (via ioctl) MII registers and check the link state. If that
1463 request fails (not just that it returns carrier down), then the MII
1464 monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1465 the same information. If both methods fail (i.e., the driver either
1466 does not support or had some error in processing both the MII register
1467 and ethtool requests), then the MII monitor will assume the link is
1470 8. Potential Sources of Trouble
1471 ===============================
1473 8.1 Adventures in Routing
1474 -------------------------
1476 When bonding is configured, it is important that the slave
1477 devices not have routes that supersede routes of the master (or,
1478 generally, not have routes at all). For example, suppose the bonding
1479 device bond0 has two slaves, eth0 and eth1, and the routing table is
1482 Kernel IP routing table
1483 Destination Gateway Genmask Flags MSS Window irtt Iface
1484 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
1485 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
1486 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
1487 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
1489 This routing configuration will likely still update the
1490 receive/transmit times in the driver (needed by the ARP monitor), but
1491 may bypass the bonding driver (because outgoing traffic to, in this
1492 case, another host on network 10 would use eth0 or eth1 before bond0).
1494 The ARP monitor (and ARP itself) may become confused by this
1495 configuration, because ARP requests (generated by the ARP monitor)
1496 will be sent on one interface (bond0), but the corresponding reply
1497 will arrive on a different interface (eth0). This reply looks to ARP
1498 as an unsolicited ARP reply (because ARP matches replies on an
1499 interface basis), and is discarded. The MII monitor is not affected
1500 by the state of the routing table.
1502 The solution here is simply to insure that slaves do not have
1503 routes of their own, and if for some reason they must, those routes do
1504 not supersede routes of their master. This should generally be the
1505 case, but unusual configurations or errant manual or automatic static
1506 route additions may cause trouble.
1508 8.2 Ethernet Device Renaming
1509 ----------------------------
1511 On systems with network configuration scripts that do not
1512 associate physical devices directly with network interface names (so
1513 that the same physical device always has the same "ethX" name), it may
1514 be necessary to add some special logic to either /etc/modules.conf or
1515 /etc/modprobe.conf (depending upon which is installed on the system).
1517 For example, given a modules.conf containing the following:
1520 options bond0 mode=some-mode miimon=50
1526 If neither eth0 and eth1 are slaves to bond0, then when the
1527 bond0 interface comes up, the devices may end up reordered. This
1528 happens because bonding is loaded first, then its slave device's
1529 drivers are loaded next. Since no other drivers have been loaded,
1530 when the e1000 driver loads, it will receive eth0 and eth1 for its
1531 devices, but the bonding configuration tries to enslave eth2 and eth3
1532 (which may later be assigned to the tg3 devices).
1534 Adding the following:
1536 add above bonding e1000 tg3
1538 causes modprobe to load e1000 then tg3, in that order, when
1539 bonding is loaded. This command is fully documented in the
1540 modules.conf manual page.
1542 On systems utilizing modprobe.conf (or modprobe.conf.local),
1543 an equivalent problem can occur. In this case, the following can be
1544 added to modprobe.conf (or modprobe.conf.local, as appropriate), as
1545 follows (all on one line; it has been split here for clarity):
1547 install bonding /sbin/modprobe tg3; /sbin/modprobe e1000;
1548 /sbin/modprobe --ignore-install bonding
1550 This will, when loading the bonding module, rather than
1551 performing the normal action, instead execute the provided command.
1552 This command loads the device drivers in the order needed, then calls
1553 modprobe with --ignore-install to cause the normal action to then take
1554 place. Full documentation on this can be found in the modprobe.conf
1555 and modprobe manual pages.
1557 8.3. Painfully Slow Or No Failed Link Detection By Miimon
1558 ---------------------------------------------------------
1560 By default, bonding enables the use_carrier option, which
1561 instructs bonding to trust the driver to maintain carrier state.
1563 As discussed in the options section, above, some drivers do
1564 not support the netif_carrier_on/_off link state tracking system.
1565 With use_carrier enabled, bonding will always see these links as up,
1566 regardless of their actual state.
1568 Additionally, other drivers do support netif_carrier, but do
1569 not maintain it in real time, e.g., only polling the link state at
1570 some fixed interval. In this case, miimon will detect failures, but
1571 only after some long period of time has expired. If it appears that
1572 miimon is very slow in detecting link failures, try specifying
1573 use_carrier=0 to see if that improves the failure detection time. If
1574 it does, then it may be that the driver checks the carrier state at a
1575 fixed interval, but does not cache the MII register values (so the
1576 use_carrier=0 method of querying the registers directly works). If
1577 use_carrier=0 does not improve the failover, then the driver may cache
1578 the registers, or the problem may be elsewhere.
1580 Also, remember that miimon only checks for the device's
1581 carrier state. It has no way to determine the state of devices on or
1582 beyond other ports of a switch, or if a switch is refusing to pass
1583 traffic while still maintaining carrier on.
1588 If running SNMP agents, the bonding driver should be loaded
1589 before any network drivers participating in a bond. This requirement
1590 is due to the interface index (ipAdEntIfIndex) being associated to
1591 the first interface found with a given IP address. That is, there is
1592 only one ipAdEntIfIndex for each IP address. For example, if eth0 and
1593 eth1 are slaves of bond0 and the driver for eth0 is loaded before the
1594 bonding driver, the interface for the IP address will be associated
1595 with the eth0 interface. This configuration is shown below, the IP
1596 address 192.168.1.1 has an interface index of 2 which indexes to eth0
1597 in the ifDescr table (ifDescr.2).
1599 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1600 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
1601 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
1602 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
1603 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
1604 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
1605 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
1606 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1607 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
1608 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1610 This problem is avoided by loading the bonding driver before
1611 any network drivers participating in a bond. Below is an example of
1612 loading the bonding driver first, the IP address 192.168.1.1 is
1613 correctly associated with ifDescr.2.
1615 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1616 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
1617 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
1618 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
1619 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
1620 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
1621 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
1622 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1623 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
1624 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1626 While some distributions may not report the interface name in
1627 ifDescr, the association between the IP address and IfIndex remains
1628 and SNMP functions such as Interface_Scan_Next will report that
1631 10. Promiscuous mode
1632 ====================
1634 When running network monitoring tools, e.g., tcpdump, it is
1635 common to enable promiscuous mode on the device, so that all traffic
1636 is seen (instead of seeing only traffic destined for the local host).
1637 The bonding driver handles promiscuous mode changes to the bonding
1638 master device (e.g., bond0), and propagates the setting to the slave
1641 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
1642 the promiscuous mode setting is propagated to all slaves.
1644 For the active-backup, balance-tlb and balance-alb modes, the
1645 promiscuous mode setting is propagated only to the active slave.
1647 For balance-tlb mode, the active slave is the slave currently
1648 receiving inbound traffic.
1650 For balance-alb mode, the active slave is the slave used as a
1651 "primary." This slave is used for mode-specific control traffic, for
1652 sending to peers that are unassigned or if the load is unbalanced.
1654 For the active-backup, balance-tlb and balance-alb modes, when
1655 the active slave changes (e.g., due to a link failure), the
1656 promiscuous setting will be propagated to the new active slave.
1658 11. Configuring Bonding for High Availability
1659 =============================================
1661 High Availability refers to configurations that provide
1662 maximum network availability by having redundant or backup devices,
1663 links or switches between the host and the rest of the world. The
1664 goal is to provide the maximum availability of network connectivity
1665 (i.e., the network always works), even though other configurations
1666 could provide higher throughput.
1668 11.1 High Availability in a Single Switch Topology
1669 --------------------------------------------------
1671 If two hosts (or a host and a single switch) are directly
1672 connected via multiple physical links, then there is no availability
1673 penalty to optimizing for maximum bandwidth. In this case, there is
1674 only one switch (or peer), so if it fails, there is no alternative
1675 access to fail over to. Additionally, the bonding load balance modes
1676 support link monitoring of their members, so if individual links fail,
1677 the load will be rebalanced across the remaining devices.
1679 See Section 13, "Configuring Bonding for Maximum Throughput"
1680 for information on configuring bonding with one peer device.
1682 11.2 High Availability in a Multiple Switch Topology
1683 ----------------------------------------------------
1685 With multiple switches, the configuration of bonding and the
1686 network changes dramatically. In multiple switch topologies, there is
1687 a trade off between network availability and usable bandwidth.
1689 Below is a sample network, configured to maximize the
1690 availability of the network:
1694 +-----+----+ +-----+----+
1695 | |port2 ISL port2| |
1696 | switch A +--------------------------+ switch B |
1698 +-----+----+ +-----++---+
1701 +-------------+ host1 +---------------+
1704 In this configuration, there is a link between the two
1705 switches (ISL, or inter switch link), and multiple ports connecting to
1706 the outside world ("port3" on each switch). There is no technical
1707 reason that this could not be extended to a third switch.
1709 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
1710 -------------------------------------------------------------
1712 In a topology such as the example above, the active-backup and
1713 broadcast modes are the only useful bonding modes when optimizing for
1714 availability; the other modes require all links to terminate on the
1715 same peer for them to behave rationally.
1717 active-backup: This is generally the preferred mode, particularly if
1718 the switches have an ISL and play together well. If the
1719 network configuration is such that one switch is specifically
1720 a backup switch (e.g., has lower capacity, higher cost, etc),
1721 then the primary option can be used to insure that the
1722 preferred link is always used when it is available.
1724 broadcast: This mode is really a special purpose mode, and is suitable
1725 only for very specific needs. For example, if the two
1726 switches are not connected (no ISL), and the networks beyond
1727 them are totally independent. In this case, if it is
1728 necessary for some specific one-way traffic to reach both
1729 independent networks, then the broadcast mode may be suitable.
1731 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
1732 ----------------------------------------------------------------
1734 The choice of link monitoring ultimately depends upon your
1735 switch. If the switch can reliably fail ports in response to other
1736 failures, then either the MII or ARP monitors should work. For
1737 example, in the above example, if the "port3" link fails at the remote
1738 end, the MII monitor has no direct means to detect this. The ARP
1739 monitor could be configured with a target at the remote end of port3,
1740 thus detecting that failure without switch support.
1742 In general, however, in a multiple switch topology, the ARP
1743 monitor can provide a higher level of reliability in detecting end to
1744 end connectivity failures (which may be caused by the failure of any
1745 individual component to pass traffic for any reason). Additionally,
1746 the ARP monitor should be configured with multiple targets (at least
1747 one for each switch in the network). This will insure that,
1748 regardless of which switch is active, the ARP monitor has a suitable
1751 Note, also, that of late many switches now support a functionality
1752 generally referred to as "trunk failover." This is a feature of the
1753 switch that causes the link state of a particular switch port to be set
1754 down (or up) when the state of another switch port goes down (or up).
1755 It's purpose is to propogate link failures from logically "exterior" ports
1756 to the logically "interior" ports that bonding is able to monitor via
1757 miimon. Availability and configuration for trunk failover varies by
1758 switch, but this can be a viable alternative to the ARP monitor when using
1761 12. Configuring Bonding for Maximum Throughput
1762 ==============================================
1764 12.1 Maximizing Throughput in a Single Switch Topology
1765 ------------------------------------------------------
1767 In a single switch configuration, the best method to maximize
1768 throughput depends upon the application and network environment. The
1769 various load balancing modes each have strengths and weaknesses in
1770 different environments, as detailed below.
1772 For this discussion, we will break down the topologies into
1773 two categories. Depending upon the destination of most traffic, we
1774 categorize them into either "gatewayed" or "local" configurations.
1776 In a gatewayed configuration, the "switch" is acting primarily
1777 as a router, and the majority of traffic passes through this router to
1778 other networks. An example would be the following:
1781 +----------+ +----------+
1782 | |eth0 port1| | to other networks
1783 | Host A +---------------------+ router +------------------->
1784 | +---------------------+ | Hosts B and C are out
1785 | |eth1 port2| | here somewhere
1786 +----------+ +----------+
1788 The router may be a dedicated router device, or another host
1789 acting as a gateway. For our discussion, the important point is that
1790 the majority of traffic from Host A will pass through the router to
1791 some other network before reaching its final destination.
1793 In a gatewayed network configuration, although Host A may
1794 communicate with many other systems, all of its traffic will be sent
1795 and received via one other peer on the local network, the router.
1797 Note that the case of two systems connected directly via
1798 multiple physical links is, for purposes of configuring bonding, the
1799 same as a gatewayed configuration. In that case, it happens that all
1800 traffic is destined for the "gateway" itself, not some other network
1803 In a local configuration, the "switch" is acting primarily as
1804 a switch, and the majority of traffic passes through this switch to
1805 reach other stations on the same network. An example would be the
1808 +----------+ +----------+ +--------+
1809 | |eth0 port1| +-------+ Host B |
1810 | Host A +------------+ switch |port3 +--------+
1811 | +------------+ | +--------+
1812 | |eth1 port2| +------------------+ Host C |
1813 +----------+ +----------+port4 +--------+
1816 Again, the switch may be a dedicated switch device, or another
1817 host acting as a gateway. For our discussion, the important point is
1818 that the majority of traffic from Host A is destined for other hosts
1819 on the same local network (Hosts B and C in the above example).
1821 In summary, in a gatewayed configuration, traffic to and from
1822 the bonded device will be to the same MAC level peer on the network
1823 (the gateway itself, i.e., the router), regardless of its final
1824 destination. In a local configuration, traffic flows directly to and
1825 from the final destinations, thus, each destination (Host B, Host C)
1826 will be addressed directly by their individual MAC addresses.
1828 This distinction between a gatewayed and a local network
1829 configuration is important because many of the load balancing modes
1830 available use the MAC addresses of the local network source and
1831 destination to make load balancing decisions. The behavior of each
1832 mode is described below.
1835 12.1.1 MT Bonding Mode Selection for Single Switch Topology
1836 -----------------------------------------------------------
1838 This configuration is the easiest to set up and to understand,
1839 although you will have to decide which bonding mode best suits your
1840 needs. The trade offs for each mode are detailed below:
1842 balance-rr: This mode is the only mode that will permit a single
1843 TCP/IP connection to stripe traffic across multiple
1844 interfaces. It is therefore the only mode that will allow a
1845 single TCP/IP stream to utilize more than one interface's
1846 worth of throughput. This comes at a cost, however: the
1847 striping generally results in peer systems receiving packets out
1848 of order, causing TCP/IP's congestion control system to kick
1849 in, often by retransmitting segments.
1851 It is possible to adjust TCP/IP's congestion limits by
1852 altering the net.ipv4.tcp_reordering sysctl parameter. The
1853 usual default value is 3, and the maximum useful value is 127.
1854 For a four interface balance-rr bond, expect that a single
1855 TCP/IP stream will utilize no more than approximately 2.3
1856 interface's worth of throughput, even after adjusting
1859 Note that the fraction of packets that will be delivered out of
1860 order is highly variable, and is unlikely to be zero. The level
1861 of reordering depends upon a variety of factors, including the
1862 networking interfaces, the switch, and the topology of the
1863 configuration. Speaking in general terms, higher speed network
1864 cards produce more reordering (due to factors such as packet
1865 coalescing), and a "many to many" topology will reorder at a
1866 higher rate than a "many slow to one fast" configuration.
1868 Many switches do not support any modes that stripe traffic
1869 (instead choosing a port based upon IP or MAC level addresses);
1870 for those devices, traffic for a particular connection flowing
1871 through the switch to a balance-rr bond will not utilize greater
1872 than one interface's worth of bandwidth.
1874 If you are utilizing protocols other than TCP/IP, UDP for
1875 example, and your application can tolerate out of order
1876 delivery, then this mode can allow for single stream datagram
1877 performance that scales near linearly as interfaces are added
1880 This mode requires the switch to have the appropriate ports
1881 configured for "etherchannel" or "trunking."
1883 active-backup: There is not much advantage in this network topology to
1884 the active-backup mode, as the inactive backup devices are all
1885 connected to the same peer as the primary. In this case, a
1886 load balancing mode (with link monitoring) will provide the
1887 same level of network availability, but with increased
1888 available bandwidth. On the plus side, active-backup mode
1889 does not require any configuration of the switch, so it may
1890 have value if the hardware available does not support any of
1891 the load balance modes.
1893 balance-xor: This mode will limit traffic such that packets destined
1894 for specific peers will always be sent over the same
1895 interface. Since the destination is determined by the MAC
1896 addresses involved, this mode works best in a "local" network
1897 configuration (as described above), with destinations all on
1898 the same local network. This mode is likely to be suboptimal
1899 if all your traffic is passed through a single router (i.e., a
1900 "gatewayed" network configuration, as described above).
1902 As with balance-rr, the switch ports need to be configured for
1903 "etherchannel" or "trunking."
1905 broadcast: Like active-backup, there is not much advantage to this
1906 mode in this type of network topology.
1908 802.3ad: This mode can be a good choice for this type of network
1909 topology. The 802.3ad mode is an IEEE standard, so all peers
1910 that implement 802.3ad should interoperate well. The 802.3ad
1911 protocol includes automatic configuration of the aggregates,
1912 so minimal manual configuration of the switch is needed
1913 (typically only to designate that some set of devices is
1914 available for 802.3ad). The 802.3ad standard also mandates
1915 that frames be delivered in order (within certain limits), so
1916 in general single connections will not see misordering of
1917 packets. The 802.3ad mode does have some drawbacks: the
1918 standard mandates that all devices in the aggregate operate at
1919 the same speed and duplex. Also, as with all bonding load
1920 balance modes other than balance-rr, no single connection will
1921 be able to utilize more than a single interface's worth of
1924 Additionally, the linux bonding 802.3ad implementation
1925 distributes traffic by peer (using an XOR of MAC addresses),
1926 so in a "gatewayed" configuration, all outgoing traffic will
1927 generally use the same device. Incoming traffic may also end
1928 up on a single device, but that is dependent upon the
1929 balancing policy of the peer's 8023.ad implementation. In a
1930 "local" configuration, traffic will be distributed across the
1931 devices in the bond.
1933 Finally, the 802.3ad mode mandates the use of the MII monitor,
1934 therefore, the ARP monitor is not available in this mode.
1936 balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
1937 Since the balancing is done according to MAC address, in a
1938 "gatewayed" configuration (as described above), this mode will
1939 send all traffic across a single device. However, in a
1940 "local" network configuration, this mode balances multiple
1941 local network peers across devices in a vaguely intelligent
1942 manner (not a simple XOR as in balance-xor or 802.3ad mode),
1943 so that mathematically unlucky MAC addresses (i.e., ones that
1944 XOR to the same value) will not all "bunch up" on a single
1947 Unlike 802.3ad, interfaces may be of differing speeds, and no
1948 special switch configuration is required. On the down side,
1949 in this mode all incoming traffic arrives over a single
1950 interface, this mode requires certain ethtool support in the
1951 network device driver of the slave interfaces, and the ARP
1952 monitor is not available.
1954 balance-alb: This mode is everything that balance-tlb is, and more.
1955 It has all of the features (and restrictions) of balance-tlb,
1956 and will also balance incoming traffic from local network
1957 peers (as described in the Bonding Module Options section,
1960 The only additional down side to this mode is that the network
1961 device driver must support changing the hardware address while
1964 12.1.2 MT Link Monitoring for Single Switch Topology
1965 ----------------------------------------------------
1967 The choice of link monitoring may largely depend upon which
1968 mode you choose to use. The more advanced load balancing modes do not
1969 support the use of the ARP monitor, and are thus restricted to using
1970 the MII monitor (which does not provide as high a level of end to end
1971 assurance as the ARP monitor).
1973 12.2 Maximum Throughput in a Multiple Switch Topology
1974 -----------------------------------------------------
1976 Multiple switches may be utilized to optimize for throughput
1977 when they are configured in parallel as part of an isolated network
1978 between two or more systems, for example:
1984 +--------+ | +---------+
1986 +------+---+ +-----+----+ +-----+----+
1987 | Switch A | | Switch B | | Switch C |
1988 +------+---+ +-----+----+ +-----+----+
1990 +--------+ | +---------+
1996 In this configuration, the switches are isolated from one
1997 another. One reason to employ a topology such as this is for an
1998 isolated network with many hosts (a cluster configured for high
1999 performance, for example), using multiple smaller switches can be more
2000 cost effective than a single larger switch, e.g., on a network with 24
2001 hosts, three 24 port switches can be significantly less expensive than
2002 a single 72 port switch.
2004 If access beyond the network is required, an individual host
2005 can be equipped with an additional network device connected to an
2006 external network; this host then additionally acts as a gateway.
2008 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2009 -------------------------------------------------------------
2011 In actual practice, the bonding mode typically employed in
2012 configurations of this type is balance-rr. Historically, in this
2013 network configuration, the usual caveats about out of order packet
2014 delivery are mitigated by the use of network adapters that do not do
2015 any kind of packet coalescing (via the use of NAPI, or because the
2016 device itself does not generate interrupts until some number of
2017 packets has arrived). When employed in this fashion, the balance-rr
2018 mode allows individual connections between two hosts to effectively
2019 utilize greater than one interface's bandwidth.
2021 12.2.2 MT Link Monitoring for Multiple Switch Topology
2022 ------------------------------------------------------
2024 Again, in actual practice, the MII monitor is most often used
2025 in this configuration, as performance is given preference over
2026 availability. The ARP monitor will function in this topology, but its
2027 advantages over the MII monitor are mitigated by the volume of probes
2028 needed as the number of systems involved grows (remember that each
2029 host in the network is configured with bonding).
2031 13. Switch Behavior Issues
2032 ==========================
2034 13.1 Link Establishment and Failover Delays
2035 -------------------------------------------
2037 Some switches exhibit undesirable behavior with regard to the
2038 timing of link up and down reporting by the switch.
2040 First, when a link comes up, some switches may indicate that
2041 the link is up (carrier available), but not pass traffic over the
2042 interface for some period of time. This delay is typically due to
2043 some type of autonegotiation or routing protocol, but may also occur
2044 during switch initialization (e.g., during recovery after a switch
2045 failure). If you find this to be a problem, specify an appropriate
2046 value to the updelay bonding module option to delay the use of the
2047 relevant interface(s).
2049 Second, some switches may "bounce" the link state one or more
2050 times while a link is changing state. This occurs most commonly while
2051 the switch is initializing. Again, an appropriate updelay value may
2054 Note that when a bonding interface has no active links, the
2055 driver will immediately reuse the first link that goes up, even if the
2056 updelay parameter has been specified (the updelay is ignored in this
2057 case). If there are slave interfaces waiting for the updelay timeout
2058 to expire, the interface that first went into that state will be
2059 immediately reused. This reduces down time of the network if the
2060 value of updelay has been overestimated, and since this occurs only in
2061 cases with no connectivity, there is no additional penalty for
2062 ignoring the updelay.
2064 In addition to the concerns about switch timings, if your
2065 switches take a long time to go into backup mode, it may be desirable
2066 to not activate a backup interface immediately after a link goes down.
2067 Failover may be delayed via the downdelay bonding module option.
2069 13.2 Duplicated Incoming Packets
2070 --------------------------------
2072 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2073 suppress duplicate packets, which should largely eliminate this problem.
2074 The following description is kept for reference.
2076 It is not uncommon to observe a short burst of duplicated
2077 traffic when the bonding device is first used, or after it has been
2078 idle for some period of time. This is most easily observed by issuing
2079 a "ping" to some other host on the network, and noticing that the
2080 output from ping flags duplicates (typically one per slave).
2082 For example, on a bond in active-backup mode with five slaves
2083 all connected to one switch, the output may appear as follows:
2086 PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
2087 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
2088 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2089 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2090 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2091 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2092 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
2093 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
2094 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2096 This is not due to an error in the bonding driver, rather, it
2097 is a side effect of how many switches update their MAC forwarding
2098 tables. Initially, the switch does not associate the MAC address in
2099 the packet with a particular switch port, and so it may send the
2100 traffic to all ports until its MAC forwarding table is updated. Since
2101 the interfaces attached to the bond may occupy multiple ports on a
2102 single switch, when the switch (temporarily) floods the traffic to all
2103 ports, the bond device receives multiple copies of the same packet
2104 (one per slave device).
2106 The duplicated packet behavior is switch dependent, some
2107 switches exhibit this, and some do not. On switches that display this
2108 behavior, it can be induced by clearing the MAC forwarding table (on
2109 most Cisco switches, the privileged command "clear mac address-table
2110 dynamic" will accomplish this).
2112 14. Hardware Specific Considerations
2113 ====================================
2115 This section contains additional information for configuring
2116 bonding on specific hardware platforms, or for interfacing bonding
2117 with particular switches or other devices.
2119 14.1 IBM BladeCenter
2120 --------------------
2122 This applies to the JS20 and similar systems.
2124 On the JS20 blades, the bonding driver supports only
2125 balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2126 largely due to the network topology inside the BladeCenter, detailed
2129 JS20 network adapter information
2130 --------------------------------
2132 All JS20s come with two Broadcom Gigabit Ethernet ports
2133 integrated on the planar (that's "motherboard" in IBM-speak). In the
2134 BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2135 I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2136 An add-on Broadcom daughter card can be installed on a JS20 to provide
2137 two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2138 wired to I/O Modules 3 and 4, respectively.
2140 Each I/O Module may contain either a switch or a passthrough
2141 module (which allows ports to be directly connected to an external
2142 switch). Some bonding modes require a specific BladeCenter internal
2143 network topology in order to function; these are detailed below.
2145 Additional BladeCenter-specific networking information can be
2146 found in two IBM Redbooks (www.ibm.com/redbooks):
2148 "IBM eServer BladeCenter Networking Options"
2149 "IBM eServer BladeCenter Layer 2-7 Network Switching"
2151 BladeCenter networking configuration
2152 ------------------------------------
2154 Because a BladeCenter can be configured in a very large number
2155 of ways, this discussion will be confined to describing basic
2158 Normally, Ethernet Switch Modules (ESMs) are used in I/O
2159 modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2160 JS20 will be connected to different internal switches (in the
2161 respective I/O modules).
2163 A passthrough module (OPM or CPM, optical or copper,
2164 passthrough module) connects the I/O module directly to an external
2165 switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2166 interfaces of a JS20 can be redirected to the outside world and
2167 connected to a common external switch.
2169 Depending upon the mix of ESMs and PMs, the network will
2170 appear to bonding as either a single switch topology (all PMs) or as a
2171 multiple switch topology (one or more ESMs, zero or more PMs). It is
2172 also possible to connect ESMs together, resulting in a configuration
2173 much like the example in "High Availability in a Multiple Switch
2176 Requirements for specific modes
2177 -------------------------------
2179 The balance-rr mode requires the use of passthrough modules
2180 for devices in the bond, all connected to an common external switch.
2181 That switch must be configured for "etherchannel" or "trunking" on the
2182 appropriate ports, as is usual for balance-rr.
2184 The balance-alb and balance-tlb modes will function with
2185 either switch modules or passthrough modules (or a mix). The only
2186 specific requirement for these modes is that all network interfaces
2187 must be able to reach all destinations for traffic sent over the
2188 bonding device (i.e., the network must converge at some point outside
2191 The active-backup mode has no additional requirements.
2193 Link monitoring issues
2194 ----------------------
2196 When an Ethernet Switch Module is in place, only the ARP
2197 monitor will reliably detect link loss to an external switch. This is
2198 nothing unusual, but examination of the BladeCenter cabinet would
2199 suggest that the "external" network ports are the ethernet ports for
2200 the system, when it fact there is a switch between these "external"
2201 ports and the devices on the JS20 system itself. The MII monitor is
2202 only able to detect link failures between the ESM and the JS20 system.
2204 When a passthrough module is in place, the MII monitor does
2205 detect failures to the "external" port, which is then directly
2206 connected to the JS20 system.
2211 The Serial Over LAN (SoL) link is established over the primary
2212 ethernet (eth0) only, therefore, any loss of link to eth0 will result
2213 in losing your SoL connection. It will not fail over with other
2214 network traffic, as the SoL system is beyond the control of the
2217 It may be desirable to disable spanning tree on the switch
2218 (either the internal Ethernet Switch Module, or an external switch) to
2219 avoid fail-over delay issues when using bonding.
2222 15. Frequently Asked Questions
2223 ==============================
2227 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2228 The new driver was designed to be SMP safe from the start.
2230 2. What type of cards will work with it?
2232 Any Ethernet type cards (you can even mix cards - a Intel
2233 EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2234 devices need not be of the same speed.
2236 Starting with version 3.2.1, bonding also supports Infiniband
2237 slaves in active-backup mode.
2239 3. How many bonding devices can I have?
2243 4. How many slaves can a bonding device have?
2245 This is limited only by the number of network interfaces Linux
2246 supports and/or the number of network cards you can place in your
2249 5. What happens when a slave link dies?
2251 If link monitoring is enabled, then the failing device will be
2252 disabled. The active-backup mode will fail over to a backup link, and
2253 other modes will ignore the failed link. The link will continue to be
2254 monitored, and should it recover, it will rejoin the bond (in whatever
2255 manner is appropriate for the mode). See the sections on High
2256 Availability and the documentation for each mode for additional
2259 Link monitoring can be enabled via either the miimon or
2260 arp_interval parameters (described in the module parameters section,
2261 above). In general, miimon monitors the carrier state as sensed by
2262 the underlying network device, and the arp monitor (arp_interval)
2263 monitors connectivity to another host on the local network.
2265 If no link monitoring is configured, the bonding driver will
2266 be unable to detect link failures, and will assume that all links are
2267 always available. This will likely result in lost packets, and a
2268 resulting degradation of performance. The precise performance loss
2269 depends upon the bonding mode and network configuration.
2271 6. Can bonding be used for High Availability?
2273 Yes. See the section on High Availability for details.
2275 7. Which switches/systems does it work with?
2277 The full answer to this depends upon the desired mode.
2279 In the basic balance modes (balance-rr and balance-xor), it
2280 works with any system that supports etherchannel (also called
2281 trunking). Most managed switches currently available have such
2282 support, and many unmanaged switches as well.
2284 The advanced balance modes (balance-tlb and balance-alb) do
2285 not have special switch requirements, but do need device drivers that
2286 support specific features (described in the appropriate section under
2287 module parameters, above).
2289 In 802.3ad mode, it works with systems that support IEEE
2290 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2291 switches currently available support 802.3ad.
2293 The active-backup mode should work with any Layer-II switch.
2295 8. Where does a bonding device get its MAC address from?
2297 When using slave devices that have fixed MAC addresses, or when
2298 the fail_over_mac option is enabled, the bonding device's MAC address is
2299 the MAC address of the active slave.
2301 For other configurations, if not explicitly configured (with
2302 ifconfig or ip link), the MAC address of the bonding device is taken from
2303 its first slave device. This MAC address is then passed to all following
2304 slaves and remains persistent (even if the first slave is removed) until
2305 the bonding device is brought down or reconfigured.
2307 If you wish to change the MAC address, you can set it with
2308 ifconfig or ip link:
2310 # ifconfig bond0 hw ether 00:11:22:33:44:55
2312 # ip link set bond0 address 66:77:88:99:aa:bb
2314 The MAC address can be also changed by bringing down/up the
2315 device and then changing its slaves (or their order):
2317 # ifconfig bond0 down ; modprobe -r bonding
2318 # ifconfig bond0 .... up
2319 # ifenslave bond0 eth...
2321 This method will automatically take the address from the next
2322 slave that is added.
2324 To restore your slaves' MAC addresses, you need to detach them
2325 from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2326 then restore the MAC addresses that the slaves had before they were
2329 16. Resources and Links
2330 =======================
2332 The latest version of the bonding driver can be found in the latest
2333 version of the linux kernel, found on http://kernel.org
2335 The latest version of this document can be found in either the latest
2336 kernel source (named Documentation/networking/bonding.txt), or on the
2337 bonding sourceforge site:
2339 http://www.sourceforge.net/projects/bonding
2341 Discussions regarding the bonding driver take place primarily on the
2342 bonding-devel mailing list, hosted at sourceforge.net. If you have
2343 questions or problems, post them to the list. The list address is:
2345 bonding-devel@lists.sourceforge.net
2347 The administrative interface (to subscribe or unsubscribe) can
2350 https://lists.sourceforge.net/lists/listinfo/bonding-devel
2352 Donald Becker's Ethernet Drivers and diag programs may be found at :
2353 - http://www.scyld.com/network/
2355 You will also find a lot of information regarding Ethernet, NWay, MII,
2356 etc. at www.scyld.com.