1 .. SPDX-License-Identifier: GPL-2.0
3 ===================================
4 Linux Ethernet Bonding Driver HOWTO
5 ===================================
7 Latest update: 27 April 2011
9 Initial release: Thomas Davis <tadavis at lbl.gov>
11 Corrections, HA extensions: 2000/10/03-15:
13 - Willy Tarreau <willy at meta-x.org>
14 - Constantine Gavrilov <const-g at xpert.com>
15 - Chad N. Tindel <ctindel at ieee dot org>
16 - Janice Girouard <girouard at us dot ibm dot com>
17 - Jay Vosburgh <fubar at us dot ibm dot com>
19 Reorganized and updated Feb 2005 by Jay Vosburgh
20 Added Sysfs information: 2006/04/24
22 - Mitch Williams <mitch.a.williams at intel.com>
27 The Linux bonding driver provides a method for aggregating
28 multiple network interfaces into a single logical "bonded" interface.
29 The behavior of the bonded interfaces depends upon the mode; generally
30 speaking, modes provide either hot standby or load balancing services.
31 Additionally, link integrity monitoring may be performed.
33 The bonding driver originally came from Donald Becker's
34 beowulf patches for kernel 2.0. It has changed quite a bit since, and
35 the original tools from extreme-linux and beowulf sites will not work
36 with this version of the driver.
38 For new versions of the driver, updated userspace tools, and
39 who to ask for help, please follow the links at the end of this file.
43 1. Bonding Driver Installation
45 2. Bonding Driver Options
47 3. Configuring Bonding Devices
48 3.1 Configuration with Sysconfig Support
49 3.1.1 Using DHCP with Sysconfig
50 3.1.2 Configuring Multiple Bonds with Sysconfig
51 3.2 Configuration with Initscripts Support
52 3.2.1 Using DHCP with Initscripts
53 3.2.2 Configuring Multiple Bonds with Initscripts
54 3.3 Configuring Bonding Manually with Ifenslave
55 3.3.1 Configuring Multiple Bonds Manually
56 3.4 Configuring Bonding Manually via Sysfs
57 3.5 Configuration with Interfaces Support
58 3.6 Overriding Configuration for Special Cases
59 3.7 Configuring LACP for 802.3ad mode in a more secure way
61 4. Querying Bonding Configuration
62 4.1 Bonding Configuration
63 4.2 Network Configuration
65 5. Switch Configuration
67 6. 802.1q VLAN Support
70 7.1 ARP Monitor Operation
71 7.2 Configuring Multiple ARP Targets
72 7.3 MII Monitor Operation
74 8. Potential Trouble Sources
75 8.1 Adventures in Routing
76 8.2 Ethernet Device Renaming
77 8.3 Painfully Slow Or No Failed Link Detection By Miimon
83 11. Configuring Bonding for High Availability
84 11.1 High Availability in a Single Switch Topology
85 11.2 High Availability in a Multiple Switch Topology
86 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
87 11.2.2 HA Link Monitoring for Multiple Switch Topology
89 12. Configuring Bonding for Maximum Throughput
90 12.1 Maximum Throughput in a Single Switch Topology
91 12.1.1 MT Bonding Mode Selection for Single Switch Topology
92 12.1.2 MT Link Monitoring for Single Switch Topology
93 12.2 Maximum Throughput in a Multiple Switch Topology
94 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
95 12.2.2 MT Link Monitoring for Multiple Switch Topology
97 13. Switch Behavior Issues
98 13.1 Link Establishment and Failover Delays
99 13.2 Duplicated Incoming Packets
101 14. Hardware Specific Considerations
104 15. Frequently Asked Questions
106 16. Resources and Links
109 1. Bonding Driver Installation
110 ==============================
112 Most popular distro kernels ship with the bonding driver
113 already available as a module. If your distro does not, or you
114 have need to compile bonding from source (e.g., configuring and
115 installing a mainline kernel from kernel.org), you'll need to perform
118 1.1 Configure and build the kernel with bonding
119 -----------------------------------------------
121 The current version of the bonding driver is available in the
122 drivers/net/bonding subdirectory of the most recent kernel source
123 (which is available on http://kernel.org). Most users "rolling their
124 own" will want to use the most recent kernel from kernel.org.
126 Configure kernel with "make menuconfig" (or "make xconfig" or
127 "make config"), then select "Bonding driver support" in the "Network
128 device support" section. It is recommended that you configure the
129 driver as module since it is currently the only way to pass parameters
130 to the driver or configure more than one bonding device.
132 Build and install the new kernel and modules.
134 1.2 Bonding Control Utility
135 ---------------------------
137 It is recommended to configure bonding via iproute2 (netlink)
138 or sysfs, the old ifenslave control utility is obsolete.
140 2. Bonding Driver Options
141 =========================
143 Options for the bonding driver are supplied as parameters to the
144 bonding module at load time, or are specified via sysfs.
146 Module options may be given as command line arguments to the
147 insmod or modprobe command, but are usually specified in either the
148 ``/etc/modprobe.d/*.conf`` configuration files, or in a distro-specific
149 configuration file (some of which are detailed in the next section).
151 Details on bonding support for sysfs is provided in the
152 "Configuring Bonding Manually via Sysfs" section, below.
154 The available bonding driver parameters are listed below. If a
155 parameter is not specified the default value is used. When initially
156 configuring a bond, it is recommended "tail -f /var/log/messages" be
157 run in a separate window to watch for bonding driver error messages.
159 It is critical that either the miimon or arp_interval and
160 arp_ip_target parameters be specified, otherwise serious network
161 degradation will occur during link failures. Very few devices do not
162 support at least miimon, so there is really no reason not to use it.
164 Options with textual values will accept either the text name
165 or, for backwards compatibility, the option value. E.g.,
166 "mode=802.3ad" and "mode=4" set the same mode.
168 The parameters are as follows:
172 Specifies the new active slave for modes that support it
173 (active-backup, balance-alb and balance-tlb). Possible values
174 are the name of any currently enslaved interface, or an empty
175 string. If a name is given, the slave and its link must be up in order
176 to be selected as the new active slave. If an empty string is
177 specified, the current active slave is cleared, and a new active
178 slave is selected automatically.
180 Note that this is only available through the sysfs interface. No module
181 parameter by this name exists.
183 The normal value of this option is the name of the currently
184 active slave, or the empty string if there is no active slave or
185 the current mode does not use an active slave.
189 In an AD system, this specifies the system priority. The allowed range
190 is 1 - 65535. If the value is not specified, it takes 65535 as the
193 This parameter has effect only in 802.3ad mode and is available through
198 In an AD system, this specifies the mac-address for the actor in
199 protocol packet exchanges (LACPDUs). The value cannot be a multicast
200 address. If the all-zeroes MAC is specified, bonding will internally
201 use the MAC of the bond itself. It is preferred to have the
202 local-admin bit set for this mac but driver does not enforce it. If
203 the value is not given then system defaults to using the masters'
204 mac address as actors' system address.
206 This parameter has effect only in 802.3ad mode and is available through
211 Specifies the 802.3ad aggregation selection logic to use. The
212 possible values and their effects are:
216 The active aggregator is chosen by largest aggregate
219 Reselection of the active aggregator occurs only when all
220 slaves of the active aggregator are down or the active
221 aggregator has no slaves.
223 This is the default value.
227 The active aggregator is chosen by largest aggregate
228 bandwidth. Reselection occurs if:
230 - A slave is added to or removed from the bond
232 - Any slave's link state changes
234 - Any slave's 802.3ad association state changes
236 - The bond's administrative state changes to up
240 The active aggregator is chosen by the largest number of
241 ports (slaves). Reselection occurs as described under the
242 "bandwidth" setting, above.
244 The bandwidth and count selection policies permit failover of
245 802.3ad aggregations when partial failure of the active aggregator
246 occurs. This keeps the aggregator with the highest availability
247 (either in bandwidth or in number of ports) active at all times.
249 This option was added in bonding version 3.4.0.
253 In an AD system, the port-key has three parts as shown below -
263 This defines the upper 10 bits of the port key. The values can be
264 from 0 - 1023. If not given, the system defaults to 0.
266 This parameter has effect only in 802.3ad mode and is available through
271 Specifies that duplicate frames (received on inactive ports) should be
272 dropped (0) or delivered (1).
274 Normally, bonding will drop duplicate frames (received on inactive
275 ports), which is desirable for most users. But there are some times
276 it is nice to allow duplicate frames to be delivered.
278 The default value is 0 (drop duplicate frames received on inactive
283 Specifies the ARP link monitoring frequency in milliseconds.
285 The ARP monitor works by periodically checking the slave
286 devices to determine whether they have sent or received
287 traffic recently (the precise criteria depends upon the
288 bonding mode, and the state of the slave). Regular traffic is
289 generated via ARP probes issued for the addresses specified by
290 the arp_ip_target option.
292 This behavior can be modified by the arp_validate option,
295 If ARP monitoring is used in an etherchannel compatible mode
296 (modes 0 and 2), the switch should be configured in a mode
297 that evenly distributes packets across all links. If the
298 switch is configured to distribute the packets in an XOR
299 fashion, all replies from the ARP targets will be received on
300 the same link which could cause the other team members to
301 fail. ARP monitoring should not be used in conjunction with
302 miimon. A value of 0 disables ARP monitoring. The default
307 Specifies the IP addresses to use as ARP monitoring peers when
308 arp_interval is > 0. These are the targets of the ARP request
309 sent to determine the health of the link to the targets.
310 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
311 addresses must be separated by a comma. At least one IP
312 address must be given for ARP monitoring to function. The
313 maximum number of targets that can be specified is 16. The
314 default value is no IP addresses.
318 Specifies whether or not ARP probes and replies should be
319 validated in any mode that supports arp monitoring, or whether
320 non-ARP traffic should be filtered (disregarded) for link
327 No validation or filtering is performed.
331 Validation is performed only for the active slave.
335 Validation is performed only for backup slaves.
339 Validation is performed for all slaves.
343 Filtering is applied to all slaves. No validation is
348 Filtering is applied to all slaves, validation is performed
349 only for the active slave.
353 Filtering is applied to all slaves, validation is performed
354 only for backup slaves.
358 Enabling validation causes the ARP monitor to examine the incoming
359 ARP requests and replies, and only consider a slave to be up if it
360 is receiving the appropriate ARP traffic.
362 For an active slave, the validation checks ARP replies to confirm
363 that they were generated by an arp_ip_target. Since backup slaves
364 do not typically receive these replies, the validation performed
365 for backup slaves is on the broadcast ARP request sent out via the
366 active slave. It is possible that some switch or network
367 configurations may result in situations wherein the backup slaves
368 do not receive the ARP requests; in such a situation, validation
369 of backup slaves must be disabled.
371 The validation of ARP requests on backup slaves is mainly helping
372 bonding to decide which slaves are more likely to work in case of
373 the active slave failure, it doesn't really guarantee that the
374 backup slave will work if it's selected as the next active slave.
376 Validation is useful in network configurations in which multiple
377 bonding hosts are concurrently issuing ARPs to one or more targets
378 beyond a common switch. Should the link between the switch and
379 target fail (but not the switch itself), the probe traffic
380 generated by the multiple bonding instances will fool the standard
381 ARP monitor into considering the links as still up. Use of
382 validation can resolve this, as the ARP monitor will only consider
383 ARP requests and replies associated with its own instance of
388 Enabling filtering causes the ARP monitor to only use incoming ARP
389 packets for link availability purposes. Arriving packets that are
390 not ARPs are delivered normally, but do not count when determining
391 if a slave is available.
393 Filtering operates by only considering the reception of ARP
394 packets (any ARP packet, regardless of source or destination) when
395 determining if a slave has received traffic for link availability
398 Filtering is useful in network configurations in which significant
399 levels of third party broadcast traffic would fool the standard
400 ARP monitor into considering the links as still up. Use of
401 filtering can resolve this, as only ARP traffic is considered for
402 link availability purposes.
404 This option was added in bonding version 3.1.0.
408 Specifies the quantity of arp_ip_targets that must be reachable
409 in order for the ARP monitor to consider a slave as being up.
410 This option affects only active-backup mode for slaves with
411 arp_validation enabled.
417 consider the slave up only when any of the arp_ip_targets
422 consider the slave up only when all of the arp_ip_targets
427 Specifies the number of arp_interval monitor checks that must
428 fail in order for an interface to be marked down by the ARP monitor.
430 In order to provide orderly failover semantics, backup interfaces
431 are permitted an extra monitor check (i.e., they must fail
432 arp_missed_max + 1 times before being marked down).
434 The default value is 2, and the allowable range is 1 - 255.
438 Specifies the time, in milliseconds, to wait before disabling
439 a slave after a link failure has been detected. This option
440 is only valid for the miimon link monitor. The downdelay
441 value should be a multiple of the miimon value; if not, it
442 will be rounded down to the nearest multiple. The default
447 Specifies whether active-backup mode should set all slaves to
448 the same MAC address at enslavement (the traditional
449 behavior), or, when enabled, perform special handling of the
450 bond's MAC address in accordance with the selected policy.
456 This setting disables fail_over_mac, and causes
457 bonding to set all slaves of an active-backup bond to
458 the same MAC address at enslavement time. This is the
463 The "active" fail_over_mac policy indicates that the
464 MAC address of the bond should always be the MAC
465 address of the currently active slave. The MAC
466 address of the slaves is not changed; instead, the MAC
467 address of the bond changes during a failover.
469 This policy is useful for devices that cannot ever
470 alter their MAC address, or for devices that refuse
471 incoming broadcasts with their own source MAC (which
472 interferes with the ARP monitor).
474 The down side of this policy is that every device on
475 the network must be updated via gratuitous ARP,
476 vs. just updating a switch or set of switches (which
477 often takes place for any traffic, not just ARP
478 traffic, if the switch snoops incoming traffic to
479 update its tables) for the traditional method. If the
480 gratuitous ARP is lost, communication may be
483 When this policy is used in conjunction with the mii
484 monitor, devices which assert link up prior to being
485 able to actually transmit and receive are particularly
486 susceptible to loss of the gratuitous ARP, and an
487 appropriate updelay setting may be required.
491 The "follow" fail_over_mac policy causes the MAC
492 address of the bond to be selected normally (normally
493 the MAC address of the first slave added to the bond).
494 However, the second and subsequent slaves are not set
495 to this MAC address while they are in a backup role; a
496 slave is programmed with the bond's MAC address at
497 failover time (and the formerly active slave receives
498 the newly active slave's MAC address).
500 This policy is useful for multiport devices that
501 either become confused or incur a performance penalty
502 when multiple ports are programmed with the same MAC
506 The default policy is none, unless the first slave cannot
507 change its MAC address, in which case the active policy is
510 This option may be modified via sysfs only when no slaves are
513 This option was added in bonding version 3.2.0. The "follow"
514 policy was added in bonding version 3.3.0.
517 Option specifying whether to send LACPDU frames periodically.
520 LACPDU frames acts as "speak when spoken to".
523 LACPDU frames are sent along the configured links
524 periodically. See lacp_rate for more details.
530 Option specifying the rate in which we'll ask our link partner
531 to transmit LACPDU packets in 802.3ad mode. Possible values
535 Request partner to transmit LACPDUs every 30 seconds
538 Request partner to transmit LACPDUs every 1 second
544 Specifies the number of bonding devices to create for this
545 instance of the bonding driver. E.g., if max_bonds is 3, and
546 the bonding driver is not already loaded, then bond0, bond1
547 and bond2 will be created. The default value is 1. Specifying
548 a value of 0 will load bonding, but will not create any devices.
552 Specifies the MII link monitoring frequency in milliseconds.
553 This determines how often the link state of each slave is
554 inspected for link failures. A value of zero disables MII
555 link monitoring. A value of 100 is a good starting point.
556 The use_carrier option, below, affects how the link state is
557 determined. See the High Availability section for additional
558 information. The default value is 0.
562 Specifies the minimum number of links that must be active before
563 asserting carrier. It is similar to the Cisco EtherChannel min-links
564 feature. This allows setting the minimum number of member ports that
565 must be up (link-up state) before marking the bond device as up
566 (carrier on). This is useful for situations where higher level services
567 such as clustering want to ensure a minimum number of low bandwidth
568 links are active before switchover. This option only affect 802.3ad
571 The default value is 0. This will cause carrier to be asserted (for
572 802.3ad mode) whenever there is an active aggregator, regardless of the
573 number of available links in that aggregator. Note that, because an
574 aggregator cannot be active without at least one available link,
575 setting this option to 0 or to 1 has the exact same effect.
579 Specifies one of the bonding policies. The default is
580 balance-rr (round robin). Possible values are:
584 Round-robin policy: Transmit packets in sequential
585 order from the first available slave through the
586 last. This mode provides load balancing and fault
591 Active-backup policy: Only one slave in the bond is
592 active. A different slave becomes active if, and only
593 if, the active slave fails. The bond's MAC address is
594 externally visible on only one port (network adapter)
595 to avoid confusing the switch.
597 In bonding version 2.6.2 or later, when a failover
598 occurs in active-backup mode, bonding will issue one
599 or more gratuitous ARPs on the newly active slave.
600 One gratuitous ARP is issued for the bonding master
601 interface and each VLAN interfaces configured above
602 it, provided that the interface has at least one IP
603 address configured. Gratuitous ARPs issued for VLAN
604 interfaces are tagged with the appropriate VLAN id.
606 This mode provides fault tolerance. The primary
607 option, documented below, affects the behavior of this
612 XOR policy: Transmit based on the selected transmit
613 hash policy. The default policy is a simple [(source
614 MAC address XOR'd with destination MAC address XOR
615 packet type ID) modulo slave count]. Alternate transmit
616 policies may be selected via the xmit_hash_policy option,
619 This mode provides load balancing and fault tolerance.
623 Broadcast policy: transmits everything on all slave
624 interfaces. This mode provides fault tolerance.
628 IEEE 802.3ad Dynamic link aggregation. Creates
629 aggregation groups that share the same speed and
630 duplex settings. Utilizes all slaves in the active
631 aggregator according to the 802.3ad specification.
633 Slave selection for outgoing traffic is done according
634 to the transmit hash policy, which may be changed from
635 the default simple XOR policy via the xmit_hash_policy
636 option, documented below. Note that not all transmit
637 policies may be 802.3ad compliant, particularly in
638 regards to the packet mis-ordering requirements of
639 section 43.2.4 of the 802.3ad standard. Differing
640 peer implementations will have varying tolerances for
645 1. Ethtool support in the base drivers for retrieving
646 the speed and duplex of each slave.
648 2. A switch that supports IEEE 802.3ad Dynamic link
651 Most switches will require some type of configuration
652 to enable 802.3ad mode.
656 Adaptive transmit load balancing: channel bonding that
657 does not require any special switch support.
659 In tlb_dynamic_lb=1 mode; the outgoing traffic is
660 distributed according to the current load (computed
661 relative to the speed) on each slave.
663 In tlb_dynamic_lb=0 mode; the load balancing based on
664 current load is disabled and the load is distributed
665 only using the hash distribution.
667 Incoming traffic is received by the current slave.
668 If the receiving slave fails, another slave takes over
669 the MAC address of the failed receiving slave.
673 Ethtool support in the base drivers for retrieving the
678 Adaptive load balancing: includes balance-tlb plus
679 receive load balancing (rlb) for IPV4 traffic, and
680 does not require any special switch support. The
681 receive load balancing is achieved by ARP negotiation.
682 The bonding driver intercepts the ARP Replies sent by
683 the local system on their way out and overwrites the
684 source hardware address with the unique hardware
685 address of one of the slaves in the bond such that
686 different peers use different hardware addresses for
689 Receive traffic from connections created by the server
690 is also balanced. When the local system sends an ARP
691 Request the bonding driver copies and saves the peer's
692 IP information from the ARP packet. When the ARP
693 Reply arrives from the peer, its hardware address is
694 retrieved and the bonding driver initiates an ARP
695 reply to this peer assigning it to one of the slaves
696 in the bond. A problematic outcome of using ARP
697 negotiation for balancing is that each time that an
698 ARP request is broadcast it uses the hardware address
699 of the bond. Hence, peers learn the hardware address
700 of the bond and the balancing of receive traffic
701 collapses to the current slave. This is handled by
702 sending updates (ARP Replies) to all the peers with
703 their individually assigned hardware address such that
704 the traffic is redistributed. Receive traffic is also
705 redistributed when a new slave is added to the bond
706 and when an inactive slave is re-activated. The
707 receive load is distributed sequentially (round robin)
708 among the group of highest speed slaves in the bond.
710 When a link is reconnected or a new slave joins the
711 bond the receive traffic is redistributed among all
712 active slaves in the bond by initiating ARP Replies
713 with the selected MAC address to each of the
714 clients. The updelay parameter (detailed below) must
715 be set to a value equal or greater than the switch's
716 forwarding delay so that the ARP Replies sent to the
717 peers will not be blocked by the switch.
721 1. Ethtool support in the base drivers for retrieving
722 the speed of each slave.
724 2. Base driver support for setting the hardware
725 address of a device while it is open. This is
726 required so that there will always be one slave in the
727 team using the bond hardware address (the
728 curr_active_slave) while having a unique hardware
729 address for each slave in the bond. If the
730 curr_active_slave fails its hardware address is
731 swapped with the new curr_active_slave that was
737 Specify the number of peer notifications (gratuitous ARPs and
738 unsolicited IPv6 Neighbor Advertisements) to be issued after a
739 failover event. As soon as the link is up on the new slave
740 (possibly immediately) a peer notification is sent on the
741 bonding device and each VLAN sub-device. This is repeated at
742 the rate specified by peer_notif_delay if the number is
745 The valid range is 0 - 255; the default value is 1. These options
746 affect only the active-backup mode. These options were added for
747 bonding versions 3.3.0 and 3.4.0 respectively.
749 From Linux 3.0 and bonding version 3.7.1, these notifications
750 are generated by the ipv4 and ipv6 code and the numbers of
751 repetitions cannot be set independently.
755 Specify the number of packets to transmit through a slave before
756 moving to the next one. When set to 0 then a slave is chosen at
759 The valid range is 0 - 65535; the default value is 1. This option
760 has effect only in balance-rr mode.
764 Specify the delay, in milliseconds, between each peer
765 notification (gratuitous ARP and unsolicited IPv6 Neighbor
766 Advertisement) when they are issued after a failover event.
767 This delay should be a multiple of the link monitor interval
768 (arp_interval or miimon, whichever is active). The default
769 value is 0 which means to match the value of the link monitor
774 A string (eth0, eth2, etc) specifying which slave is the
775 primary device. The specified device will always be the
776 active slave while it is available. Only when the primary is
777 off-line will alternate devices be used. This is useful when
778 one slave is preferred over another, e.g., when one slave has
779 higher throughput than another.
781 The primary option is only valid for active-backup(1),
782 balance-tlb (5) and balance-alb (6) mode.
786 Specifies the reselection policy for the primary slave. This
787 affects how the primary slave is chosen to become the active slave
788 when failure of the active slave or recovery of the primary slave
789 occurs. This option is designed to prevent flip-flopping between
790 the primary slave and other slaves. Possible values are:
792 always or 0 (default)
794 The primary slave becomes the active slave whenever it
799 The primary slave becomes the active slave when it comes
800 back up, if the speed and duplex of the primary slave is
801 better than the speed and duplex of the current active
806 The primary slave becomes the active slave only if the
807 current active slave fails and the primary slave is up.
809 The primary_reselect setting is ignored in two cases:
811 If no slaves are active, the first slave to recover is
812 made the active slave.
814 When initially enslaved, the primary slave is always made
817 Changing the primary_reselect policy via sysfs will cause an
818 immediate selection of the best active slave according to the new
819 policy. This may or may not result in a change of the active
820 slave, depending upon the circumstances.
822 This option was added for bonding version 3.6.0.
826 Specifies if dynamic shuffling of flows is enabled in tlb
827 mode. The value has no effect on any other modes.
829 The default behavior of tlb mode is to shuffle active flows across
830 slaves based on the load in that interval. This gives nice lb
831 characteristics but can cause packet reordering. If re-ordering is
832 a concern use this variable to disable flow shuffling and rely on
833 load balancing provided solely by the hash distribution.
834 xmit-hash-policy can be used to select the appropriate hashing for
837 The sysfs entry can be used to change the setting per bond device
838 and the initial value is derived from the module parameter. The
839 sysfs entry is allowed to be changed only if the bond device is
842 The default value is "1" that enables flow shuffling while value "0"
843 disables it. This option was added in bonding driver 3.7.1
848 Specifies the time, in milliseconds, to wait before enabling a
849 slave after a link recovery has been detected. This option is
850 only valid for the miimon link monitor. The updelay value
851 should be a multiple of the miimon value; if not, it will be
852 rounded down to the nearest multiple. The default value is 0.
856 Specifies whether or not miimon should use MII or ETHTOOL
857 ioctls vs. netif_carrier_ok() to determine the link
858 status. The MII or ETHTOOL ioctls are less efficient and
859 utilize a deprecated calling sequence within the kernel. The
860 netif_carrier_ok() relies on the device driver to maintain its
861 state with netif_carrier_on/off; at this writing, most, but
862 not all, device drivers support this facility.
864 If bonding insists that the link is up when it should not be,
865 it may be that your network device driver does not support
866 netif_carrier_on/off. The default state for netif_carrier is
867 "carrier on," so if a driver does not support netif_carrier,
868 it will appear as if the link is always up. In this case,
869 setting use_carrier to 0 will cause bonding to revert to the
870 MII / ETHTOOL ioctl method to determine the link state.
872 A value of 1 enables the use of netif_carrier_ok(), a value of
873 0 will use the deprecated MII / ETHTOOL ioctls. The default
878 Selects the transmit hash policy to use for slave selection in
879 balance-xor, 802.3ad, and tlb modes. Possible values are:
883 Uses XOR of hardware MAC addresses and packet type ID
884 field to generate the hash. The formula is
886 hash = source MAC XOR destination MAC XOR packet type ID
887 slave number = hash modulo slave count
889 This algorithm will place all traffic to a particular
890 network peer on the same slave.
892 This algorithm is 802.3ad compliant.
896 This policy uses a combination of layer2 and layer3
897 protocol information to generate the hash.
899 Uses XOR of hardware MAC addresses and IP addresses to
900 generate the hash. The formula is
902 hash = source MAC XOR destination MAC XOR packet type ID
903 hash = hash XOR source IP XOR destination IP
904 hash = hash XOR (hash RSHIFT 16)
905 hash = hash XOR (hash RSHIFT 8)
906 And then hash is reduced modulo slave count.
908 If the protocol is IPv6 then the source and destination
909 addresses are first hashed using ipv6_addr_hash.
911 This algorithm will place all traffic to a particular
912 network peer on the same slave. For non-IP traffic,
913 the formula is the same as for the layer2 transmit
916 This policy is intended to provide a more balanced
917 distribution of traffic than layer2 alone, especially
918 in environments where a layer3 gateway device is
919 required to reach most destinations.
921 This algorithm is 802.3ad compliant.
925 This policy uses upper layer protocol information,
926 when available, to generate the hash. This allows for
927 traffic to a particular network peer to span multiple
928 slaves, although a single connection will not span
931 The formula for unfragmented TCP and UDP packets is
933 hash = source port, destination port (as in the header)
934 hash = hash XOR source IP XOR destination IP
935 hash = hash XOR (hash RSHIFT 16)
936 hash = hash XOR (hash RSHIFT 8)
937 And then hash is reduced modulo slave count.
939 If the protocol is IPv6 then the source and destination
940 addresses are first hashed using ipv6_addr_hash.
942 For fragmented TCP or UDP packets and all other IPv4 and
943 IPv6 protocol traffic, the source and destination port
944 information is omitted. For non-IP traffic, the
945 formula is the same as for the layer2 transmit hash
948 This algorithm is not fully 802.3ad compliant. A
949 single TCP or UDP conversation containing both
950 fragmented and unfragmented packets will see packets
951 striped across two interfaces. This may result in out
952 of order delivery. Most traffic types will not meet
953 this criteria, as TCP rarely fragments traffic, and
954 most UDP traffic is not involved in extended
955 conversations. Other implementations of 802.3ad may
956 or may not tolerate this noncompliance.
960 This policy uses the same formula as layer2+3 but it
961 relies on skb_flow_dissect to obtain the header fields
962 which might result in the use of inner headers if an
963 encapsulation protocol is used. For example this will
964 improve the performance for tunnel users because the
965 packets will be distributed according to the encapsulated
970 This policy uses the same formula as layer3+4 but it
971 relies on skb_flow_dissect to obtain the header fields
972 which might result in the use of inner headers if an
973 encapsulation protocol is used. For example this will
974 improve the performance for tunnel users because the
975 packets will be distributed according to the encapsulated
980 This policy uses a very rudimentary vlan ID and source mac
981 hash to load-balance traffic per-vlan, with failover
982 should one leg fail. The intended use case is for a bond
983 shared by multiple virtual machines, all configured to
984 use their own vlan, to give lacp-like functionality
985 without requiring lacp-capable switching hardware.
987 The formula for the hash is simply
989 hash = (vlan ID) XOR (source MAC vendor) XOR (source MAC dev)
991 The default value is layer2. This option was added in bonding
992 version 2.6.3. In earlier versions of bonding, this parameter
993 does not exist, and the layer2 policy is the only policy. The
994 layer2+3 value was added for bonding version 3.2.2.
998 Specifies the number of IGMP membership reports to be issued after
999 a failover event. One membership report is issued immediately after
1000 the failover, subsequent packets are sent in each 200ms interval.
1002 The valid range is 0 - 255; the default value is 1. A value of 0
1003 prevents the IGMP membership report from being issued in response
1004 to the failover event.
1006 This option is useful for bonding modes balance-rr (0), active-backup
1007 (1), balance-tlb (5) and balance-alb (6), in which a failover can
1008 switch the IGMP traffic from one slave to another. Therefore a fresh
1009 IGMP report must be issued to cause the switch to forward the incoming
1010 IGMP traffic over the newly selected slave.
1012 This option was added for bonding version 3.7.0.
1016 Specifies the number of seconds between instances where the bonding
1017 driver sends learning packets to each slaves peer switch.
1019 The valid range is 1 - 0x7fffffff; the default value is 1. This Option
1020 has effect only in balance-tlb and balance-alb modes.
1022 3. Configuring Bonding Devices
1023 ==============================
1025 You can configure bonding using either your distro's network
1026 initialization scripts, or manually using either iproute2 or the
1027 sysfs interface. Distros generally use one of three packages for the
1028 network initialization scripts: initscripts, sysconfig or interfaces.
1029 Recent versions of these packages have support for bonding, while older
1032 We will first describe the options for configuring bonding for
1033 distros using versions of initscripts, sysconfig and interfaces with full
1034 or partial support for bonding, then provide information on enabling
1035 bonding without support from the network initialization scripts (i.e.,
1036 older versions of initscripts or sysconfig).
1038 If you're unsure whether your distro uses sysconfig,
1039 initscripts or interfaces, or don't know if it's new enough, have no fear.
1040 Determining this is fairly straightforward.
1042 First, look for a file called interfaces in /etc/network directory.
1043 If this file is present in your system, then your system use interfaces. See
1044 Configuration with Interfaces Support.
1046 Else, issue the command::
1048 $ rpm -qf /sbin/ifup
1050 It will respond with a line of text starting with either
1051 "initscripts" or "sysconfig," followed by some numbers. This is the
1052 package that provides your network initialization scripts.
1054 Next, to determine if your installation supports bonding,
1057 $ grep ifenslave /sbin/ifup
1059 If this returns any matches, then your initscripts or
1060 sysconfig has support for bonding.
1062 3.1 Configuration with Sysconfig Support
1063 ----------------------------------------
1065 This section applies to distros using a version of sysconfig
1066 with bonding support, for example, SuSE Linux Enterprise Server 9.
1068 SuSE SLES 9's networking configuration system does support
1069 bonding, however, at this writing, the YaST system configuration
1070 front end does not provide any means to work with bonding devices.
1071 Bonding devices can be managed by hand, however, as follows.
1073 First, if they have not already been configured, configure the
1074 slave devices. On SLES 9, this is most easily done by running the
1075 yast2 sysconfig configuration utility. The goal is for to create an
1076 ifcfg-id file for each slave device. The simplest way to accomplish
1077 this is to configure the devices for DHCP (this is only to get the
1078 file ifcfg-id file created; see below for some issues with DHCP). The
1079 name of the configuration file for each device will be of the form::
1081 ifcfg-id-xx:xx:xx:xx:xx:xx
1083 Where the "xx" portion will be replaced with the digits from
1084 the device's permanent MAC address.
1086 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
1087 created, it is necessary to edit the configuration files for the slave
1088 devices (the MAC addresses correspond to those of the slave devices).
1089 Before editing, the file will contain multiple lines, and will look
1090 something like this::
1095 UNIQUE='XNzu.WeZGOGF+4wE'
1096 _nm_name='bus-pci-0001:61:01.0'
1098 Change the BOOTPROTO and STARTMODE lines to the following::
1103 Do not alter the UNIQUE or _nm_name lines. Remove any other
1104 lines (USERCTL, etc).
1106 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
1107 it's time to create the configuration file for the bonding device
1108 itself. This file is named ifcfg-bondX, where X is the number of the
1109 bonding device to create, starting at 0. The first such file is
1110 ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
1111 network configuration system will correctly start multiple instances
1114 The contents of the ifcfg-bondX file is as follows::
1117 BROADCAST="10.0.2.255"
1119 NETMASK="255.255.0.0"
1123 BONDING_MASTER="yes"
1124 BONDING_MODULE_OPTS="mode=active-backup miimon=100"
1125 BONDING_SLAVE0="eth0"
1126 BONDING_SLAVE1="bus-pci-0000:06:08.1"
1128 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
1129 values with the appropriate values for your network.
1131 The STARTMODE specifies when the device is brought online.
1132 The possible values are:
1134 ======== ======================================================
1135 onboot The device is started at boot time. If you're not
1136 sure, this is probably what you want.
1138 manual The device is started only when ifup is called
1139 manually. Bonding devices may be configured this
1140 way if you do not wish them to start automatically
1141 at boot for some reason.
1143 hotplug The device is started by a hotplug event. This is not
1144 a valid choice for a bonding device.
1146 off or The device configuration is ignored.
1148 ======== ======================================================
1150 The line BONDING_MASTER='yes' indicates that the device is a
1151 bonding master device. The only useful value is "yes."
1153 The contents of BONDING_MODULE_OPTS are supplied to the
1154 instance of the bonding module for this device. Specify the options
1155 for the bonding mode, link monitoring, and so on here. Do not include
1156 the max_bonds bonding parameter; this will confuse the configuration
1157 system if you have multiple bonding devices.
1159 Finally, supply one BONDING_SLAVEn="slave device" for each
1160 slave. where "n" is an increasing value, one for each slave. The
1161 "slave device" is either an interface name, e.g., "eth0", or a device
1162 specifier for the network device. The interface name is easier to
1163 find, but the ethN names are subject to change at boot time if, e.g.,
1164 a device early in the sequence has failed. The device specifiers
1165 (bus-pci-0000:06:08.1 in the example above) specify the physical
1166 network device, and will not change unless the device's bus location
1167 changes (for example, it is moved from one PCI slot to another). The
1168 example above uses one of each type for demonstration purposes; most
1169 configurations will choose one or the other for all slave devices.
1171 When all configuration files have been modified or created,
1172 networking must be restarted for the configuration changes to take
1173 effect. This can be accomplished via the following::
1175 # /etc/init.d/network restart
1177 Note that the network control script (/sbin/ifdown) will
1178 remove the bonding module as part of the network shutdown processing,
1179 so it is not necessary to remove the module by hand if, e.g., the
1180 module parameters have changed.
1182 Also, at this writing, YaST/YaST2 will not manage bonding
1183 devices (they do not show bonding interfaces on its list of network
1184 devices). It is necessary to edit the configuration file by hand to
1185 change the bonding configuration.
1187 Additional general options and details of the ifcfg file
1188 format can be found in an example ifcfg template file::
1190 /etc/sysconfig/network/ifcfg.template
1192 Note that the template does not document the various ``BONDING_*``
1193 settings described above, but does describe many of the other options.
1195 3.1.1 Using DHCP with Sysconfig
1196 -------------------------------
1198 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1199 will cause it to query DHCP for its IP address information. At this
1200 writing, this does not function for bonding devices; the scripts
1201 attempt to obtain the device address from DHCP prior to adding any of
1202 the slave devices. Without active slaves, the DHCP requests are not
1203 sent to the network.
1205 3.1.2 Configuring Multiple Bonds with Sysconfig
1206 -----------------------------------------------
1208 The sysconfig network initialization system is capable of
1209 handling multiple bonding devices. All that is necessary is for each
1210 bonding instance to have an appropriately configured ifcfg-bondX file
1211 (as described above). Do not specify the "max_bonds" parameter to any
1212 instance of bonding, as this will confuse sysconfig. If you require
1213 multiple bonding devices with identical parameters, create multiple
1216 Because the sysconfig scripts supply the bonding module
1217 options in the ifcfg-bondX file, it is not necessary to add them to
1218 the system ``/etc/modules.d/*.conf`` configuration files.
1220 3.2 Configuration with Initscripts Support
1221 ------------------------------------------
1223 This section applies to distros using a recent version of
1224 initscripts with bonding support, for example, Red Hat Enterprise Linux
1225 version 3 or later, Fedora, etc. On these systems, the network
1226 initialization scripts have knowledge of bonding, and can be configured to
1227 control bonding devices. Note that older versions of the initscripts
1228 package have lower levels of support for bonding; this will be noted where
1231 These distros will not automatically load the network adapter
1232 driver unless the ethX device is configured with an IP address.
1233 Because of this constraint, users must manually configure a
1234 network-script file for all physical adapters that will be members of
1235 a bondX link. Network script files are located in the directory:
1237 /etc/sysconfig/network-scripts
1239 The file name must be prefixed with "ifcfg-eth" and suffixed
1240 with the adapter's physical adapter number. For example, the script
1241 for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1242 Place the following text in the file::
1251 The DEVICE= line will be different for every ethX device and
1252 must correspond with the name of the file, i.e., ifcfg-eth1 must have
1253 a device line of DEVICE=eth1. The setting of the MASTER= line will
1254 also depend on the final bonding interface name chosen for your bond.
1255 As with other network devices, these typically start at 0, and go up
1256 one for each device, i.e., the first bonding instance is bond0, the
1257 second is bond1, and so on.
1259 Next, create a bond network script. The file name for this
1260 script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1261 the number of the bond. For bond0 the file is named "ifcfg-bond0",
1262 for bond1 it is named "ifcfg-bond1", and so on. Within that file,
1263 place the following text::
1267 NETMASK=255.255.255.0
1269 BROADCAST=192.168.1.255
1274 Be sure to change the networking specific lines (IPADDR,
1275 NETMASK, NETWORK and BROADCAST) to match your network configuration.
1277 For later versions of initscripts, such as that found with Fedora
1278 7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1279 and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
1280 file, e.g. a line of the format::
1282 BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
1284 will configure the bond with the specified options. The options
1285 specified in BONDING_OPTS are identical to the bonding module parameters
1286 except for the arp_ip_target field when using versions of initscripts older
1287 than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
1288 using older versions each target should be included as a separate option and
1289 should be preceded by a '+' to indicate it should be added to the list of
1290 queried targets, e.g.,::
1292 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1294 is the proper syntax to specify multiple targets. When specifying
1295 options via BONDING_OPTS, it is not necessary to edit
1296 ``/etc/modprobe.d/*.conf``.
1298 For even older versions of initscripts that do not support
1299 BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1300 your distro) to load the bonding module with your desired options when the
1301 bond0 interface is brought up. The following lines in /etc/modprobe.d/*.conf
1302 will load the bonding module, and select its options:
1305 options bond0 mode=balance-alb miimon=100
1307 Replace the sample parameters with the appropriate set of
1308 options for your configuration.
1310 Finally run "/etc/rc.d/init.d/network restart" as root. This
1311 will restart the networking subsystem and your bond link should be now
1314 3.2.1 Using DHCP with Initscripts
1315 ---------------------------------
1317 Recent versions of initscripts (the versions supplied with Fedora
1318 Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1319 work) have support for assigning IP information to bonding devices via
1322 To configure bonding for DHCP, configure it as described
1323 above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1324 and add a line consisting of "TYPE=Bonding". Note that the TYPE value
1327 3.2.2 Configuring Multiple Bonds with Initscripts
1328 -------------------------------------------------
1330 Initscripts packages that are included with Fedora 7 and Red Hat
1331 Enterprise Linux 5 support multiple bonding interfaces by simply
1332 specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1333 number of the bond. This support requires sysfs support in the kernel,
1334 and a bonding driver of version 3.0.0 or later. Other configurations may
1335 not support this method for specifying multiple bonding interfaces; for
1336 those instances, see the "Configuring Multiple Bonds Manually" section,
1339 3.3 Configuring Bonding Manually with iproute2
1340 -----------------------------------------------
1342 This section applies to distros whose network initialization
1343 scripts (the sysconfig or initscripts package) do not have specific
1344 knowledge of bonding. One such distro is SuSE Linux Enterprise Server
1347 The general method for these systems is to place the bonding
1348 module parameters into a config file in /etc/modprobe.d/ (as
1349 appropriate for the installed distro), then add modprobe and/or
1350 `ip link` commands to the system's global init script. The name of
1351 the global init script differs; for sysconfig, it is
1352 /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1354 For example, if you wanted to make a simple bond of two e100
1355 devices (presumed to be eth0 and eth1), and have it persist across
1356 reboots, edit the appropriate file (/etc/init.d/boot.local or
1357 /etc/rc.d/rc.local), and add the following::
1359 modprobe bonding mode=balance-alb miimon=100
1361 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1362 ip link set eth0 master bond0
1363 ip link set eth1 master bond0
1365 Replace the example bonding module parameters and bond0
1366 network configuration (IP address, netmask, etc) with the appropriate
1367 values for your configuration.
1369 Unfortunately, this method will not provide support for the
1370 ifup and ifdown scripts on the bond devices. To reload the bonding
1371 configuration, it is necessary to run the initialization script, e.g.,::
1373 # /etc/init.d/boot.local
1377 # /etc/rc.d/rc.local
1379 It may be desirable in such a case to create a separate script
1380 which only initializes the bonding configuration, then call that
1381 separate script from within boot.local. This allows for bonding to be
1382 enabled without re-running the entire global init script.
1384 To shut down the bonding devices, it is necessary to first
1385 mark the bonding device itself as being down, then remove the
1386 appropriate device driver modules. For our example above, you can do
1389 # ifconfig bond0 down
1393 Again, for convenience, it may be desirable to create a script
1394 with these commands.
1397 3.3.1 Configuring Multiple Bonds Manually
1398 -----------------------------------------
1400 This section contains information on configuring multiple
1401 bonding devices with differing options for those systems whose network
1402 initialization scripts lack support for configuring multiple bonds.
1404 If you require multiple bonding devices, but all with the same
1405 options, you may wish to use the "max_bonds" module parameter,
1408 To create multiple bonding devices with differing options, it is
1409 preferable to use bonding parameters exported by sysfs, documented in the
1412 For versions of bonding without sysfs support, the only means to
1413 provide multiple instances of bonding with differing options is to load
1414 the bonding driver multiple times. Note that current versions of the
1415 sysconfig network initialization scripts handle this automatically; if
1416 your distro uses these scripts, no special action is needed. See the
1417 section Configuring Bonding Devices, above, if you're not sure about your
1418 network initialization scripts.
1420 To load multiple instances of the module, it is necessary to
1421 specify a different name for each instance (the module loading system
1422 requires that every loaded module, even multiple instances of the same
1423 module, have a unique name). This is accomplished by supplying multiple
1424 sets of bonding options in ``/etc/modprobe.d/*.conf``, for example::
1427 options bond0 -o bond0 mode=balance-rr miimon=100
1430 options bond1 -o bond1 mode=balance-alb miimon=50
1432 will load the bonding module two times. The first instance is
1433 named "bond0" and creates the bond0 device in balance-rr mode with an
1434 miimon of 100. The second instance is named "bond1" and creates the
1435 bond1 device in balance-alb mode with an miimon of 50.
1437 In some circumstances (typically with older distributions),
1438 the above does not work, and the second bonding instance never sees
1439 its options. In that case, the second options line can be substituted
1442 install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1443 mode=balance-alb miimon=50
1445 This may be repeated any number of times, specifying a new and
1446 unique name in place of bond1 for each subsequent instance.
1448 It has been observed that some Red Hat supplied kernels are unable
1449 to rename modules at load time (the "-o bond1" part). Attempts to pass
1450 that option to modprobe will produce an "Operation not permitted" error.
1451 This has been reported on some Fedora Core kernels, and has been seen on
1452 RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1453 to configure multiple bonds with differing parameters (as they are older
1454 kernels, and also lack sysfs support).
1456 3.4 Configuring Bonding Manually via Sysfs
1457 ------------------------------------------
1459 Starting with version 3.0.0, Channel Bonding may be configured
1460 via the sysfs interface. This interface allows dynamic configuration
1461 of all bonds in the system without unloading the module. It also
1462 allows for adding and removing bonds at runtime. Ifenslave is no
1463 longer required, though it is still supported.
1465 Use of the sysfs interface allows you to use multiple bonds
1466 with different configurations without having to reload the module.
1467 It also allows you to use multiple, differently configured bonds when
1468 bonding is compiled into the kernel.
1470 You must have the sysfs filesystem mounted to configure
1471 bonding this way. The examples in this document assume that you
1472 are using the standard mount point for sysfs, e.g. /sys. If your
1473 sysfs filesystem is mounted elsewhere, you will need to adjust the
1474 example paths accordingly.
1476 Creating and Destroying Bonds
1477 -----------------------------
1478 To add a new bond foo::
1480 # echo +foo > /sys/class/net/bonding_masters
1482 To remove an existing bond bar::
1484 # echo -bar > /sys/class/net/bonding_masters
1486 To show all existing bonds::
1488 # cat /sys/class/net/bonding_masters
1492 due to 4K size limitation of sysfs files, this list may be
1493 truncated if you have more than a few hundred bonds. This is unlikely
1494 to occur under normal operating conditions.
1496 Adding and Removing Slaves
1497 --------------------------
1498 Interfaces may be enslaved to a bond using the file
1499 /sys/class/net/<bond>/bonding/slaves. The semantics for this file
1500 are the same as for the bonding_masters file.
1502 To enslave interface eth0 to bond bond0::
1505 # echo +eth0 > /sys/class/net/bond0/bonding/slaves
1507 To free slave eth0 from bond bond0::
1509 # echo -eth0 > /sys/class/net/bond0/bonding/slaves
1511 When an interface is enslaved to a bond, symlinks between the
1512 two are created in the sysfs filesystem. In this case, you would get
1513 /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1514 /sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1516 This means that you can tell quickly whether or not an
1517 interface is enslaved by looking for the master symlink. Thus:
1518 # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1519 will free eth0 from whatever bond it is enslaved to, regardless of
1520 the name of the bond interface.
1522 Changing a Bond's Configuration
1523 -------------------------------
1524 Each bond may be configured individually by manipulating the
1525 files located in /sys/class/net/<bond name>/bonding
1527 The names of these files correspond directly with the command-
1528 line parameters described elsewhere in this file, and, with the
1529 exception of arp_ip_target, they accept the same values. To see the
1530 current setting, simply cat the appropriate file.
1532 A few examples will be given here; for specific usage
1533 guidelines for each parameter, see the appropriate section in this
1536 To configure bond0 for balance-alb mode::
1538 # ifconfig bond0 down
1539 # echo 6 > /sys/class/net/bond0/bonding/mode
1541 # echo balance-alb > /sys/class/net/bond0/bonding/mode
1545 The bond interface must be down before the mode can be changed.
1547 To enable MII monitoring on bond0 with a 1 second interval::
1549 # echo 1000 > /sys/class/net/bond0/bonding/miimon
1553 If ARP monitoring is enabled, it will disabled when MII
1554 monitoring is enabled, and vice-versa.
1556 To add ARP targets::
1558 # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1559 # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1563 up to 16 target addresses may be specified.
1565 To remove an ARP target::
1567 # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1569 To configure the interval between learning packet transmits::
1571 # echo 12 > /sys/class/net/bond0/bonding/lp_interval
1575 the lp_interval is the number of seconds between instances where
1576 the bonding driver sends learning packets to each slaves peer switch. The
1577 default interval is 1 second.
1579 Example Configuration
1580 ---------------------
1581 We begin with the same example that is shown in section 3.3,
1582 executed with sysfs, and without using ifenslave.
1584 To make a simple bond of two e100 devices (presumed to be eth0
1585 and eth1), and have it persist across reboots, edit the appropriate
1586 file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1591 echo balance-alb > /sys/class/net/bond0/bonding/mode
1592 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1593 echo 100 > /sys/class/net/bond0/bonding/miimon
1594 echo +eth0 > /sys/class/net/bond0/bonding/slaves
1595 echo +eth1 > /sys/class/net/bond0/bonding/slaves
1597 To add a second bond, with two e1000 interfaces in
1598 active-backup mode, using ARP monitoring, add the following lines to
1602 echo +bond1 > /sys/class/net/bonding_masters
1603 echo active-backup > /sys/class/net/bond1/bonding/mode
1604 ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1605 echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1606 echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1607 echo +eth2 > /sys/class/net/bond1/bonding/slaves
1608 echo +eth3 > /sys/class/net/bond1/bonding/slaves
1610 3.5 Configuration with Interfaces Support
1611 -----------------------------------------
1613 This section applies to distros which use /etc/network/interfaces file
1614 to describe network interface configuration, most notably Debian and it's
1617 The ifup and ifdown commands on Debian don't support bonding out of
1618 the box. The ifenslave-2.6 package should be installed to provide bonding
1619 support. Once installed, this package will provide ``bond-*`` options
1620 to be used into /etc/network/interfaces.
1622 Note that ifenslave-2.6 package will load the bonding module and use
1623 the ifenslave command when appropriate.
1625 Example Configurations
1626 ----------------------
1628 In /etc/network/interfaces, the following stanza will configure bond0, in
1629 active-backup mode, with eth0 and eth1 as slaves::
1632 iface bond0 inet dhcp
1633 bond-slaves eth0 eth1
1634 bond-mode active-backup
1636 bond-primary eth0 eth1
1638 If the above configuration doesn't work, you might have a system using
1639 upstart for system startup. This is most notably true for recent
1640 Ubuntu versions. The following stanza in /etc/network/interfaces will
1641 produce the same result on those systems::
1644 iface bond0 inet dhcp
1646 bond-mode active-backup
1650 iface eth0 inet manual
1652 bond-primary eth0 eth1
1655 iface eth1 inet manual
1657 bond-primary eth0 eth1
1659 For a full list of ``bond-*`` supported options in /etc/network/interfaces and
1660 some more advanced examples tailored to you particular distros, see the files in
1661 /usr/share/doc/ifenslave-2.6.
1663 3.6 Overriding Configuration for Special Cases
1664 ----------------------------------------------
1666 When using the bonding driver, the physical port which transmits a frame is
1667 typically selected by the bonding driver, and is not relevant to the user or
1668 system administrator. The output port is simply selected using the policies of
1669 the selected bonding mode. On occasion however, it is helpful to direct certain
1670 classes of traffic to certain physical interfaces on output to implement
1671 slightly more complex policies. For example, to reach a web server over a
1672 bonded interface in which eth0 connects to a private network, while eth1
1673 connects via a public network, it may be desirous to bias the bond to send said
1674 traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1675 can safely be sent over either interface. Such configurations may be achieved
1676 using the traffic control utilities inherent in linux.
1678 By default the bonding driver is multiqueue aware and 16 queues are created
1679 when the driver initializes (see Documentation/networking/multiqueue.rst
1680 for details). If more or less queues are desired the module parameter
1681 tx_queues can be used to change this value. There is no sysfs parameter
1682 available as the allocation is done at module init time.
1684 The output of the file /proc/net/bonding/bondX has changed so the output Queue
1685 ID is now printed for each slave::
1687 Bonding Mode: fault-tolerance (active-backup)
1689 Currently Active Slave: eth0
1691 MII Polling Interval (ms): 0
1695 Slave Interface: eth0
1697 Link Failure Count: 0
1698 Permanent HW addr: 00:1a:a0:12:8f:cb
1701 Slave Interface: eth1
1703 Link Failure Count: 0
1704 Permanent HW addr: 00:1a:a0:12:8f:cc
1707 The queue_id for a slave can be set using the command::
1709 # echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1711 Any interface that needs a queue_id set should set it with multiple calls
1712 like the one above until proper priorities are set for all interfaces. On
1713 distributions that allow configuration via initscripts, multiple 'queue_id'
1714 arguments can be added to BONDING_OPTS to set all needed slave queues.
1716 These queue id's can be used in conjunction with the tc utility to configure
1717 a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1718 slave devices. For instance, say we wanted, in the above configuration to
1719 force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1720 device. The following commands would accomplish this::
1722 # tc qdisc add dev bond0 handle 1 root multiq
1724 # tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip \
1725 dst 192.168.1.100 action skbedit queue_mapping 2
1727 These commands tell the kernel to attach a multiqueue queue discipline to the
1728 bond0 interface and filter traffic enqueued to it, such that packets with a dst
1729 ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1730 This value is then passed into the driver, causing the normal output path
1731 selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1733 Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver
1734 that normal output policy selection should take place. One benefit to simply
1735 leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1736 driver that is now present. This awareness allows tc filters to be placed on
1737 slave devices as well as bond devices and the bonding driver will simply act as
1738 a pass-through for selecting output queues on the slave device rather than
1739 output port selection.
1741 This feature first appeared in bonding driver version 3.7.0 and support for
1742 output slave selection was limited to round-robin and active-backup modes.
1744 3.7 Configuring LACP for 802.3ad mode in a more secure way
1745 ----------------------------------------------------------
1747 When using 802.3ad bonding mode, the Actor (host) and Partner (switch)
1748 exchange LACPDUs. These LACPDUs cannot be sniffed, because they are
1749 destined to link local mac addresses (which switches/bridges are not
1750 supposed to forward). However, most of the values are easily predictable
1751 or are simply the machine's MAC address (which is trivially known to all
1752 other hosts in the same L2). This implies that other machines in the L2
1753 domain can spoof LACPDU packets from other hosts to the switch and potentially
1754 cause mayhem by joining (from the point of view of the switch) another
1755 machine's aggregate, thus receiving a portion of that hosts incoming
1756 traffic and / or spoofing traffic from that machine themselves (potentially
1757 even successfully terminating some portion of flows). Though this is not
1758 a likely scenario, one could avoid this possibility by simply configuring
1759 few bonding parameters:
1761 (a) ad_actor_system : You can set a random mac-address that can be used for
1762 these LACPDU exchanges. The value can not be either NULL or Multicast.
1763 Also it's preferable to set the local-admin bit. Following shell code
1764 generates a random mac-address as described above::
1766 # sys_mac_addr=$(printf '%02x:%02x:%02x:%02x:%02x:%02x' \
1767 $(( (RANDOM & 0xFE) | 0x02 )) \
1768 $(( RANDOM & 0xFF )) \
1769 $(( RANDOM & 0xFF )) \
1770 $(( RANDOM & 0xFF )) \
1771 $(( RANDOM & 0xFF )) \
1772 $(( RANDOM & 0xFF )))
1773 # echo $sys_mac_addr > /sys/class/net/bond0/bonding/ad_actor_system
1775 (b) ad_actor_sys_prio : Randomize the system priority. The default value
1776 is 65535, but system can take the value from 1 - 65535. Following shell
1777 code generates random priority and sets it::
1779 # sys_prio=$(( 1 + RANDOM + RANDOM ))
1780 # echo $sys_prio > /sys/class/net/bond0/bonding/ad_actor_sys_prio
1782 (c) ad_user_port_key : Use the user portion of the port-key. The default
1783 keeps this empty. These are the upper 10 bits of the port-key and value
1784 ranges from 0 - 1023. Following shell code generates these 10 bits and
1787 # usr_port_key=$(( RANDOM & 0x3FF ))
1788 # echo $usr_port_key > /sys/class/net/bond0/bonding/ad_user_port_key
1791 4 Querying Bonding Configuration
1792 =================================
1794 4.1 Bonding Configuration
1795 -------------------------
1797 Each bonding device has a read-only file residing in the
1798 /proc/net/bonding directory. The file contents include information
1799 about the bonding configuration, options and state of each slave.
1801 For example, the contents of /proc/net/bonding/bond0 after the
1802 driver is loaded with parameters of mode=0 and miimon=1000 is
1803 generally as follows::
1805 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1806 Bonding Mode: load balancing (round-robin)
1807 Currently Active Slave: eth0
1809 MII Polling Interval (ms): 1000
1813 Slave Interface: eth1
1815 Link Failure Count: 1
1817 Slave Interface: eth0
1819 Link Failure Count: 1
1821 The precise format and contents will change depending upon the
1822 bonding configuration, state, and version of the bonding driver.
1824 4.2 Network configuration
1825 -------------------------
1827 The network configuration can be inspected using the ifconfig
1828 command. Bonding devices will have the MASTER flag set; Bonding slave
1829 devices will have the SLAVE flag set. The ifconfig output does not
1830 contain information on which slaves are associated with which masters.
1832 In the example below, the bond0 interface is the master
1833 (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1834 bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1835 TLB and ALB that require a unique MAC address for each slave::
1838 bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1839 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1840 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1841 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1842 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1843 collisions:0 txqueuelen:0
1845 eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1846 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1847 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1848 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1849 collisions:0 txqueuelen:100
1850 Interrupt:10 Base address:0x1080
1852 eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1853 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1854 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1855 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1856 collisions:0 txqueuelen:100
1857 Interrupt:9 Base address:0x1400
1859 5. Switch Configuration
1860 =======================
1862 For this section, "switch" refers to whatever system the
1863 bonded devices are directly connected to (i.e., where the other end of
1864 the cable plugs into). This may be an actual dedicated switch device,
1865 or it may be another regular system (e.g., another computer running
1868 The active-backup, balance-tlb and balance-alb modes do not
1869 require any specific configuration of the switch.
1871 The 802.3ad mode requires that the switch have the appropriate
1872 ports configured as an 802.3ad aggregation. The precise method used
1873 to configure this varies from switch to switch, but, for example, a
1874 Cisco 3550 series switch requires that the appropriate ports first be
1875 grouped together in a single etherchannel instance, then that
1876 etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1877 standard EtherChannel).
1879 The balance-rr, balance-xor and broadcast modes generally
1880 require that the switch have the appropriate ports grouped together.
1881 The nomenclature for such a group differs between switches, it may be
1882 called an "etherchannel" (as in the Cisco example, above), a "trunk
1883 group" or some other similar variation. For these modes, each switch
1884 will also have its own configuration options for the switch's transmit
1885 policy to the bond. Typical choices include XOR of either the MAC or
1886 IP addresses. The transmit policy of the two peers does not need to
1887 match. For these three modes, the bonding mode really selects a
1888 transmit policy for an EtherChannel group; all three will interoperate
1889 with another EtherChannel group.
1892 6. 802.1q VLAN Support
1893 ======================
1895 It is possible to configure VLAN devices over a bond interface
1896 using the 8021q driver. However, only packets coming from the 8021q
1897 driver and passing through bonding will be tagged by default. Self
1898 generated packets, for example, bonding's learning packets or ARP
1899 packets generated by either ALB mode or the ARP monitor mechanism, are
1900 tagged internally by bonding itself. As a result, bonding must
1901 "learn" the VLAN IDs configured above it, and use those IDs to tag
1902 self generated packets.
1904 For reasons of simplicity, and to support the use of adapters
1905 that can do VLAN hardware acceleration offloading, the bonding
1906 interface declares itself as fully hardware offloading capable, it gets
1907 the add_vid/kill_vid notifications to gather the necessary
1908 information, and it propagates those actions to the slaves. In case
1909 of mixed adapter types, hardware accelerated tagged packets that
1910 should go through an adapter that is not offloading capable are
1911 "un-accelerated" by the bonding driver so the VLAN tag sits in the
1914 VLAN interfaces *must* be added on top of a bonding interface
1915 only after enslaving at least one slave. The bonding interface has a
1916 hardware address of 00:00:00:00:00:00 until the first slave is added.
1917 If the VLAN interface is created prior to the first enslavement, it
1918 would pick up the all-zeroes hardware address. Once the first slave
1919 is attached to the bond, the bond device itself will pick up the
1920 slave's hardware address, which is then available for the VLAN device.
1922 Also, be aware that a similar problem can occur if all slaves
1923 are released from a bond that still has one or more VLAN interfaces on
1924 top of it. When a new slave is added, the bonding interface will
1925 obtain its hardware address from the first slave, which might not
1926 match the hardware address of the VLAN interfaces (which was
1927 ultimately copied from an earlier slave).
1929 There are two methods to insure that the VLAN device operates
1930 with the correct hardware address if all slaves are removed from a
1933 1. Remove all VLAN interfaces then recreate them
1935 2. Set the bonding interface's hardware address so that it
1936 matches the hardware address of the VLAN interfaces.
1938 Note that changing a VLAN interface's HW address would set the
1939 underlying device -- i.e. the bonding interface -- to promiscuous
1940 mode, which might not be what you want.
1946 The bonding driver at present supports two schemes for
1947 monitoring a slave device's link state: the ARP monitor and the MII
1950 At the present time, due to implementation restrictions in the
1951 bonding driver itself, it is not possible to enable both ARP and MII
1952 monitoring simultaneously.
1954 7.1 ARP Monitor Operation
1955 -------------------------
1957 The ARP monitor operates as its name suggests: it sends ARP
1958 queries to one or more designated peer systems on the network, and
1959 uses the response as an indication that the link is operating. This
1960 gives some assurance that traffic is actually flowing to and from one
1961 or more peers on the local network.
1963 The ARP monitor relies on the device driver itself to verify
1964 that traffic is flowing. In particular, the driver must keep up to
1965 date the last receive time, dev->last_rx. Drivers that use NETIF_F_LLTX
1966 flag must also update netdev_queue->trans_start. If they do not, then the
1967 ARP monitor will immediately fail any slaves using that driver, and
1968 those slaves will stay down. If networking monitoring (tcpdump, etc)
1969 shows the ARP requests and replies on the network, then it may be that
1970 your device driver is not updating last_rx and trans_start.
1972 7.2 Configuring Multiple ARP Targets
1973 ------------------------------------
1975 While ARP monitoring can be done with just one target, it can
1976 be useful in a High Availability setup to have several targets to
1977 monitor. In the case of just one target, the target itself may go
1978 down or have a problem making it unresponsive to ARP requests. Having
1979 an additional target (or several) increases the reliability of the ARP
1982 Multiple ARP targets must be separated by commas as follows::
1984 # example options for ARP monitoring with three targets
1986 options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1988 For just a single target the options would resemble::
1990 # example options for ARP monitoring with one target
1992 options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1995 7.3 MII Monitor Operation
1996 -------------------------
1998 The MII monitor monitors only the carrier state of the local
1999 network interface. It accomplishes this in one of three ways: by
2000 depending upon the device driver to maintain its carrier state, by
2001 querying the device's MII registers, or by making an ethtool query to
2004 If the use_carrier module parameter is 1 (the default value),
2005 then the MII monitor will rely on the driver for carrier state
2006 information (via the netif_carrier subsystem). As explained in the
2007 use_carrier parameter information, above, if the MII monitor fails to
2008 detect carrier loss on the device (e.g., when the cable is physically
2009 disconnected), it may be that the driver does not support
2012 If use_carrier is 0, then the MII monitor will first query the
2013 device's (via ioctl) MII registers and check the link state. If that
2014 request fails (not just that it returns carrier down), then the MII
2015 monitor will make an ethtool ETHTOOL_GLINK request to attempt to obtain
2016 the same information. If both methods fail (i.e., the driver either
2017 does not support or had some error in processing both the MII register
2018 and ethtool requests), then the MII monitor will assume the link is
2021 8. Potential Sources of Trouble
2022 ===============================
2024 8.1 Adventures in Routing
2025 -------------------------
2027 When bonding is configured, it is important that the slave
2028 devices not have routes that supersede routes of the master (or,
2029 generally, not have routes at all). For example, suppose the bonding
2030 device bond0 has two slaves, eth0 and eth1, and the routing table is
2033 Kernel IP routing table
2034 Destination Gateway Genmask Flags MSS Window irtt Iface
2035 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
2036 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
2037 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
2038 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
2040 This routing configuration will likely still update the
2041 receive/transmit times in the driver (needed by the ARP monitor), but
2042 may bypass the bonding driver (because outgoing traffic to, in this
2043 case, another host on network 10 would use eth0 or eth1 before bond0).
2045 The ARP monitor (and ARP itself) may become confused by this
2046 configuration, because ARP requests (generated by the ARP monitor)
2047 will be sent on one interface (bond0), but the corresponding reply
2048 will arrive on a different interface (eth0). This reply looks to ARP
2049 as an unsolicited ARP reply (because ARP matches replies on an
2050 interface basis), and is discarded. The MII monitor is not affected
2051 by the state of the routing table.
2053 The solution here is simply to insure that slaves do not have
2054 routes of their own, and if for some reason they must, those routes do
2055 not supersede routes of their master. This should generally be the
2056 case, but unusual configurations or errant manual or automatic static
2057 route additions may cause trouble.
2059 8.2 Ethernet Device Renaming
2060 ----------------------------
2062 On systems with network configuration scripts that do not
2063 associate physical devices directly with network interface names (so
2064 that the same physical device always has the same "ethX" name), it may
2065 be necessary to add some special logic to config files in
2068 For example, given a modules.conf containing the following::
2071 options bond0 mode=some-mode miimon=50
2077 If neither eth0 and eth1 are slaves to bond0, then when the
2078 bond0 interface comes up, the devices may end up reordered. This
2079 happens because bonding is loaded first, then its slave device's
2080 drivers are loaded next. Since no other drivers have been loaded,
2081 when the e1000 driver loads, it will receive eth0 and eth1 for its
2082 devices, but the bonding configuration tries to enslave eth2 and eth3
2083 (which may later be assigned to the tg3 devices).
2085 Adding the following::
2087 add above bonding e1000 tg3
2089 causes modprobe to load e1000 then tg3, in that order, when
2090 bonding is loaded. This command is fully documented in the
2091 modules.conf manual page.
2093 On systems utilizing modprobe an equivalent problem can occur.
2094 In this case, the following can be added to config files in
2095 /etc/modprobe.d/ as::
2097 softdep bonding pre: tg3 e1000
2099 This will load tg3 and e1000 modules before loading the bonding one.
2100 Full documentation on this can be found in the modprobe.d and modprobe
2103 8.3. Painfully Slow Or No Failed Link Detection By Miimon
2104 ---------------------------------------------------------
2106 By default, bonding enables the use_carrier option, which
2107 instructs bonding to trust the driver to maintain carrier state.
2109 As discussed in the options section, above, some drivers do
2110 not support the netif_carrier_on/_off link state tracking system.
2111 With use_carrier enabled, bonding will always see these links as up,
2112 regardless of their actual state.
2114 Additionally, other drivers do support netif_carrier, but do
2115 not maintain it in real time, e.g., only polling the link state at
2116 some fixed interval. In this case, miimon will detect failures, but
2117 only after some long period of time has expired. If it appears that
2118 miimon is very slow in detecting link failures, try specifying
2119 use_carrier=0 to see if that improves the failure detection time. If
2120 it does, then it may be that the driver checks the carrier state at a
2121 fixed interval, but does not cache the MII register values (so the
2122 use_carrier=0 method of querying the registers directly works). If
2123 use_carrier=0 does not improve the failover, then the driver may cache
2124 the registers, or the problem may be elsewhere.
2126 Also, remember that miimon only checks for the device's
2127 carrier state. It has no way to determine the state of devices on or
2128 beyond other ports of a switch, or if a switch is refusing to pass
2129 traffic while still maintaining carrier on.
2134 If running SNMP agents, the bonding driver should be loaded
2135 before any network drivers participating in a bond. This requirement
2136 is due to the interface index (ipAdEntIfIndex) being associated to
2137 the first interface found with a given IP address. That is, there is
2138 only one ipAdEntIfIndex for each IP address. For example, if eth0 and
2139 eth1 are slaves of bond0 and the driver for eth0 is loaded before the
2140 bonding driver, the interface for the IP address will be associated
2141 with the eth0 interface. This configuration is shown below, the IP
2142 address 192.168.1.1 has an interface index of 2 which indexes to eth0
2143 in the ifDescr table (ifDescr.2).
2147 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2148 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
2149 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
2150 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
2151 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
2152 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
2153 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
2154 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2155 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
2156 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2158 This problem is avoided by loading the bonding driver before
2159 any network drivers participating in a bond. Below is an example of
2160 loading the bonding driver first, the IP address 192.168.1.1 is
2161 correctly associated with ifDescr.2.
2163 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2164 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
2165 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
2166 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
2167 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
2168 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
2169 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
2170 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2171 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
2172 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2174 While some distributions may not report the interface name in
2175 ifDescr, the association between the IP address and IfIndex remains
2176 and SNMP functions such as Interface_Scan_Next will report that
2179 10. Promiscuous mode
2180 ====================
2182 When running network monitoring tools, e.g., tcpdump, it is
2183 common to enable promiscuous mode on the device, so that all traffic
2184 is seen (instead of seeing only traffic destined for the local host).
2185 The bonding driver handles promiscuous mode changes to the bonding
2186 master device (e.g., bond0), and propagates the setting to the slave
2189 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
2190 the promiscuous mode setting is propagated to all slaves.
2192 For the active-backup, balance-tlb and balance-alb modes, the
2193 promiscuous mode setting is propagated only to the active slave.
2195 For balance-tlb mode, the active slave is the slave currently
2196 receiving inbound traffic.
2198 For balance-alb mode, the active slave is the slave used as a
2199 "primary." This slave is used for mode-specific control traffic, for
2200 sending to peers that are unassigned or if the load is unbalanced.
2202 For the active-backup, balance-tlb and balance-alb modes, when
2203 the active slave changes (e.g., due to a link failure), the
2204 promiscuous setting will be propagated to the new active slave.
2206 11. Configuring Bonding for High Availability
2207 =============================================
2209 High Availability refers to configurations that provide
2210 maximum network availability by having redundant or backup devices,
2211 links or switches between the host and the rest of the world. The
2212 goal is to provide the maximum availability of network connectivity
2213 (i.e., the network always works), even though other configurations
2214 could provide higher throughput.
2216 11.1 High Availability in a Single Switch Topology
2217 --------------------------------------------------
2219 If two hosts (or a host and a single switch) are directly
2220 connected via multiple physical links, then there is no availability
2221 penalty to optimizing for maximum bandwidth. In this case, there is
2222 only one switch (or peer), so if it fails, there is no alternative
2223 access to fail over to. Additionally, the bonding load balance modes
2224 support link monitoring of their members, so if individual links fail,
2225 the load will be rebalanced across the remaining devices.
2227 See Section 12, "Configuring Bonding for Maximum Throughput"
2228 for information on configuring bonding with one peer device.
2230 11.2 High Availability in a Multiple Switch Topology
2231 ----------------------------------------------------
2233 With multiple switches, the configuration of bonding and the
2234 network changes dramatically. In multiple switch topologies, there is
2235 a trade off between network availability and usable bandwidth.
2237 Below is a sample network, configured to maximize the
2238 availability of the network::
2242 +-----+----+ +-----+----+
2243 | |port2 ISL port2| |
2244 | switch A +--------------------------+ switch B |
2246 +-----+----+ +-----++---+
2249 +-------------+ host1 +---------------+
2252 In this configuration, there is a link between the two
2253 switches (ISL, or inter switch link), and multiple ports connecting to
2254 the outside world ("port3" on each switch). There is no technical
2255 reason that this could not be extended to a third switch.
2257 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
2258 -------------------------------------------------------------
2260 In a topology such as the example above, the active-backup and
2261 broadcast modes are the only useful bonding modes when optimizing for
2262 availability; the other modes require all links to terminate on the
2263 same peer for them to behave rationally.
2266 This is generally the preferred mode, particularly if
2267 the switches have an ISL and play together well. If the
2268 network configuration is such that one switch is specifically
2269 a backup switch (e.g., has lower capacity, higher cost, etc),
2270 then the primary option can be used to insure that the
2271 preferred link is always used when it is available.
2274 This mode is really a special purpose mode, and is suitable
2275 only for very specific needs. For example, if the two
2276 switches are not connected (no ISL), and the networks beyond
2277 them are totally independent. In this case, if it is
2278 necessary for some specific one-way traffic to reach both
2279 independent networks, then the broadcast mode may be suitable.
2281 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
2282 ----------------------------------------------------------------
2284 The choice of link monitoring ultimately depends upon your
2285 switch. If the switch can reliably fail ports in response to other
2286 failures, then either the MII or ARP monitors should work. For
2287 example, in the above example, if the "port3" link fails at the remote
2288 end, the MII monitor has no direct means to detect this. The ARP
2289 monitor could be configured with a target at the remote end of port3,
2290 thus detecting that failure without switch support.
2292 In general, however, in a multiple switch topology, the ARP
2293 monitor can provide a higher level of reliability in detecting end to
2294 end connectivity failures (which may be caused by the failure of any
2295 individual component to pass traffic for any reason). Additionally,
2296 the ARP monitor should be configured with multiple targets (at least
2297 one for each switch in the network). This will insure that,
2298 regardless of which switch is active, the ARP monitor has a suitable
2301 Note, also, that of late many switches now support a functionality
2302 generally referred to as "trunk failover." This is a feature of the
2303 switch that causes the link state of a particular switch port to be set
2304 down (or up) when the state of another switch port goes down (or up).
2305 Its purpose is to propagate link failures from logically "exterior" ports
2306 to the logically "interior" ports that bonding is able to monitor via
2307 miimon. Availability and configuration for trunk failover varies by
2308 switch, but this can be a viable alternative to the ARP monitor when using
2311 12. Configuring Bonding for Maximum Throughput
2312 ==============================================
2314 12.1 Maximizing Throughput in a Single Switch Topology
2315 ------------------------------------------------------
2317 In a single switch configuration, the best method to maximize
2318 throughput depends upon the application and network environment. The
2319 various load balancing modes each have strengths and weaknesses in
2320 different environments, as detailed below.
2322 For this discussion, we will break down the topologies into
2323 two categories. Depending upon the destination of most traffic, we
2324 categorize them into either "gatewayed" or "local" configurations.
2326 In a gatewayed configuration, the "switch" is acting primarily
2327 as a router, and the majority of traffic passes through this router to
2328 other networks. An example would be the following::
2331 +----------+ +----------+
2332 | |eth0 port1| | to other networks
2333 | Host A +---------------------+ router +------------------->
2334 | +---------------------+ | Hosts B and C are out
2335 | |eth1 port2| | here somewhere
2336 +----------+ +----------+
2338 The router may be a dedicated router device, or another host
2339 acting as a gateway. For our discussion, the important point is that
2340 the majority of traffic from Host A will pass through the router to
2341 some other network before reaching its final destination.
2343 In a gatewayed network configuration, although Host A may
2344 communicate with many other systems, all of its traffic will be sent
2345 and received via one other peer on the local network, the router.
2347 Note that the case of two systems connected directly via
2348 multiple physical links is, for purposes of configuring bonding, the
2349 same as a gatewayed configuration. In that case, it happens that all
2350 traffic is destined for the "gateway" itself, not some other network
2353 In a local configuration, the "switch" is acting primarily as
2354 a switch, and the majority of traffic passes through this switch to
2355 reach other stations on the same network. An example would be the
2358 +----------+ +----------+ +--------+
2359 | |eth0 port1| +-------+ Host B |
2360 | Host A +------------+ switch |port3 +--------+
2361 | +------------+ | +--------+
2362 | |eth1 port2| +------------------+ Host C |
2363 +----------+ +----------+port4 +--------+
2366 Again, the switch may be a dedicated switch device, or another
2367 host acting as a gateway. For our discussion, the important point is
2368 that the majority of traffic from Host A is destined for other hosts
2369 on the same local network (Hosts B and C in the above example).
2371 In summary, in a gatewayed configuration, traffic to and from
2372 the bonded device will be to the same MAC level peer on the network
2373 (the gateway itself, i.e., the router), regardless of its final
2374 destination. In a local configuration, traffic flows directly to and
2375 from the final destinations, thus, each destination (Host B, Host C)
2376 will be addressed directly by their individual MAC addresses.
2378 This distinction between a gatewayed and a local network
2379 configuration is important because many of the load balancing modes
2380 available use the MAC addresses of the local network source and
2381 destination to make load balancing decisions. The behavior of each
2382 mode is described below.
2385 12.1.1 MT Bonding Mode Selection for Single Switch Topology
2386 -----------------------------------------------------------
2388 This configuration is the easiest to set up and to understand,
2389 although you will have to decide which bonding mode best suits your
2390 needs. The trade offs for each mode are detailed below:
2393 This mode is the only mode that will permit a single
2394 TCP/IP connection to stripe traffic across multiple
2395 interfaces. It is therefore the only mode that will allow a
2396 single TCP/IP stream to utilize more than one interface's
2397 worth of throughput. This comes at a cost, however: the
2398 striping generally results in peer systems receiving packets out
2399 of order, causing TCP/IP's congestion control system to kick
2400 in, often by retransmitting segments.
2402 It is possible to adjust TCP/IP's congestion limits by
2403 altering the net.ipv4.tcp_reordering sysctl parameter. The
2404 usual default value is 3. But keep in mind TCP stack is able
2405 to automatically increase this when it detects reorders.
2407 Note that the fraction of packets that will be delivered out of
2408 order is highly variable, and is unlikely to be zero. The level
2409 of reordering depends upon a variety of factors, including the
2410 networking interfaces, the switch, and the topology of the
2411 configuration. Speaking in general terms, higher speed network
2412 cards produce more reordering (due to factors such as packet
2413 coalescing), and a "many to many" topology will reorder at a
2414 higher rate than a "many slow to one fast" configuration.
2416 Many switches do not support any modes that stripe traffic
2417 (instead choosing a port based upon IP or MAC level addresses);
2418 for those devices, traffic for a particular connection flowing
2419 through the switch to a balance-rr bond will not utilize greater
2420 than one interface's worth of bandwidth.
2422 If you are utilizing protocols other than TCP/IP, UDP for
2423 example, and your application can tolerate out of order
2424 delivery, then this mode can allow for single stream datagram
2425 performance that scales near linearly as interfaces are added
2428 This mode requires the switch to have the appropriate ports
2429 configured for "etherchannel" or "trunking."
2432 There is not much advantage in this network topology to
2433 the active-backup mode, as the inactive backup devices are all
2434 connected to the same peer as the primary. In this case, a
2435 load balancing mode (with link monitoring) will provide the
2436 same level of network availability, but with increased
2437 available bandwidth. On the plus side, active-backup mode
2438 does not require any configuration of the switch, so it may
2439 have value if the hardware available does not support any of
2440 the load balance modes.
2443 This mode will limit traffic such that packets destined
2444 for specific peers will always be sent over the same
2445 interface. Since the destination is determined by the MAC
2446 addresses involved, this mode works best in a "local" network
2447 configuration (as described above), with destinations all on
2448 the same local network. This mode is likely to be suboptimal
2449 if all your traffic is passed through a single router (i.e., a
2450 "gatewayed" network configuration, as described above).
2452 As with balance-rr, the switch ports need to be configured for
2453 "etherchannel" or "trunking."
2456 Like active-backup, there is not much advantage to this
2457 mode in this type of network topology.
2460 This mode can be a good choice for this type of network
2461 topology. The 802.3ad mode is an IEEE standard, so all peers
2462 that implement 802.3ad should interoperate well. The 802.3ad
2463 protocol includes automatic configuration of the aggregates,
2464 so minimal manual configuration of the switch is needed
2465 (typically only to designate that some set of devices is
2466 available for 802.3ad). The 802.3ad standard also mandates
2467 that frames be delivered in order (within certain limits), so
2468 in general single connections will not see misordering of
2469 packets. The 802.3ad mode does have some drawbacks: the
2470 standard mandates that all devices in the aggregate operate at
2471 the same speed and duplex. Also, as with all bonding load
2472 balance modes other than balance-rr, no single connection will
2473 be able to utilize more than a single interface's worth of
2476 Additionally, the linux bonding 802.3ad implementation
2477 distributes traffic by peer (using an XOR of MAC addresses
2478 and packet type ID), so in a "gatewayed" configuration, all
2479 outgoing traffic will generally use the same device. Incoming
2480 traffic may also end up on a single device, but that is
2481 dependent upon the balancing policy of the peer's 802.3ad
2482 implementation. In a "local" configuration, traffic will be
2483 distributed across the devices in the bond.
2485 Finally, the 802.3ad mode mandates the use of the MII monitor,
2486 therefore, the ARP monitor is not available in this mode.
2489 The balance-tlb mode balances outgoing traffic by peer.
2490 Since the balancing is done according to MAC address, in a
2491 "gatewayed" configuration (as described above), this mode will
2492 send all traffic across a single device. However, in a
2493 "local" network configuration, this mode balances multiple
2494 local network peers across devices in a vaguely intelligent
2495 manner (not a simple XOR as in balance-xor or 802.3ad mode),
2496 so that mathematically unlucky MAC addresses (i.e., ones that
2497 XOR to the same value) will not all "bunch up" on a single
2500 Unlike 802.3ad, interfaces may be of differing speeds, and no
2501 special switch configuration is required. On the down side,
2502 in this mode all incoming traffic arrives over a single
2503 interface, this mode requires certain ethtool support in the
2504 network device driver of the slave interfaces, and the ARP
2505 monitor is not available.
2508 This mode is everything that balance-tlb is, and more.
2509 It has all of the features (and restrictions) of balance-tlb,
2510 and will also balance incoming traffic from local network
2511 peers (as described in the Bonding Module Options section,
2514 The only additional down side to this mode is that the network
2515 device driver must support changing the hardware address while
2518 12.1.2 MT Link Monitoring for Single Switch Topology
2519 ----------------------------------------------------
2521 The choice of link monitoring may largely depend upon which
2522 mode you choose to use. The more advanced load balancing modes do not
2523 support the use of the ARP monitor, and are thus restricted to using
2524 the MII monitor (which does not provide as high a level of end to end
2525 assurance as the ARP monitor).
2527 12.2 Maximum Throughput in a Multiple Switch Topology
2528 -----------------------------------------------------
2530 Multiple switches may be utilized to optimize for throughput
2531 when they are configured in parallel as part of an isolated network
2532 between two or more systems, for example::
2538 +--------+ | +---------+
2540 +------+---+ +-----+----+ +-----+----+
2541 | Switch A | | Switch B | | Switch C |
2542 +------+---+ +-----+----+ +-----+----+
2544 +--------+ | +---------+
2550 In this configuration, the switches are isolated from one
2551 another. One reason to employ a topology such as this is for an
2552 isolated network with many hosts (a cluster configured for high
2553 performance, for example), using multiple smaller switches can be more
2554 cost effective than a single larger switch, e.g., on a network with 24
2555 hosts, three 24 port switches can be significantly less expensive than
2556 a single 72 port switch.
2558 If access beyond the network is required, an individual host
2559 can be equipped with an additional network device connected to an
2560 external network; this host then additionally acts as a gateway.
2562 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2563 -------------------------------------------------------------
2565 In actual practice, the bonding mode typically employed in
2566 configurations of this type is balance-rr. Historically, in this
2567 network configuration, the usual caveats about out of order packet
2568 delivery are mitigated by the use of network adapters that do not do
2569 any kind of packet coalescing (via the use of NAPI, or because the
2570 device itself does not generate interrupts until some number of
2571 packets has arrived). When employed in this fashion, the balance-rr
2572 mode allows individual connections between two hosts to effectively
2573 utilize greater than one interface's bandwidth.
2575 12.2.2 MT Link Monitoring for Multiple Switch Topology
2576 ------------------------------------------------------
2578 Again, in actual practice, the MII monitor is most often used
2579 in this configuration, as performance is given preference over
2580 availability. The ARP monitor will function in this topology, but its
2581 advantages over the MII monitor are mitigated by the volume of probes
2582 needed as the number of systems involved grows (remember that each
2583 host in the network is configured with bonding).
2585 13. Switch Behavior Issues
2586 ==========================
2588 13.1 Link Establishment and Failover Delays
2589 -------------------------------------------
2591 Some switches exhibit undesirable behavior with regard to the
2592 timing of link up and down reporting by the switch.
2594 First, when a link comes up, some switches may indicate that
2595 the link is up (carrier available), but not pass traffic over the
2596 interface for some period of time. This delay is typically due to
2597 some type of autonegotiation or routing protocol, but may also occur
2598 during switch initialization (e.g., during recovery after a switch
2599 failure). If you find this to be a problem, specify an appropriate
2600 value to the updelay bonding module option to delay the use of the
2601 relevant interface(s).
2603 Second, some switches may "bounce" the link state one or more
2604 times while a link is changing state. This occurs most commonly while
2605 the switch is initializing. Again, an appropriate updelay value may
2608 Note that when a bonding interface has no active links, the
2609 driver will immediately reuse the first link that goes up, even if the
2610 updelay parameter has been specified (the updelay is ignored in this
2611 case). If there are slave interfaces waiting for the updelay timeout
2612 to expire, the interface that first went into that state will be
2613 immediately reused. This reduces down time of the network if the
2614 value of updelay has been overestimated, and since this occurs only in
2615 cases with no connectivity, there is no additional penalty for
2616 ignoring the updelay.
2618 In addition to the concerns about switch timings, if your
2619 switches take a long time to go into backup mode, it may be desirable
2620 to not activate a backup interface immediately after a link goes down.
2621 Failover may be delayed via the downdelay bonding module option.
2623 13.2 Duplicated Incoming Packets
2624 --------------------------------
2626 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2627 suppress duplicate packets, which should largely eliminate this problem.
2628 The following description is kept for reference.
2630 It is not uncommon to observe a short burst of duplicated
2631 traffic when the bonding device is first used, or after it has been
2632 idle for some period of time. This is most easily observed by issuing
2633 a "ping" to some other host on the network, and noticing that the
2634 output from ping flags duplicates (typically one per slave).
2636 For example, on a bond in active-backup mode with five slaves
2637 all connected to one switch, the output may appear as follows::
2640 PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
2641 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
2642 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2643 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2644 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2645 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2646 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
2647 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
2648 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2650 This is not due to an error in the bonding driver, rather, it
2651 is a side effect of how many switches update their MAC forwarding
2652 tables. Initially, the switch does not associate the MAC address in
2653 the packet with a particular switch port, and so it may send the
2654 traffic to all ports until its MAC forwarding table is updated. Since
2655 the interfaces attached to the bond may occupy multiple ports on a
2656 single switch, when the switch (temporarily) floods the traffic to all
2657 ports, the bond device receives multiple copies of the same packet
2658 (one per slave device).
2660 The duplicated packet behavior is switch dependent, some
2661 switches exhibit this, and some do not. On switches that display this
2662 behavior, it can be induced by clearing the MAC forwarding table (on
2663 most Cisco switches, the privileged command "clear mac address-table
2664 dynamic" will accomplish this).
2666 14. Hardware Specific Considerations
2667 ====================================
2669 This section contains additional information for configuring
2670 bonding on specific hardware platforms, or for interfacing bonding
2671 with particular switches or other devices.
2673 14.1 IBM BladeCenter
2674 --------------------
2676 This applies to the JS20 and similar systems.
2678 On the JS20 blades, the bonding driver supports only
2679 balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2680 largely due to the network topology inside the BladeCenter, detailed
2683 JS20 network adapter information
2684 --------------------------------
2686 All JS20s come with two Broadcom Gigabit Ethernet ports
2687 integrated on the planar (that's "motherboard" in IBM-speak). In the
2688 BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2689 I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2690 An add-on Broadcom daughter card can be installed on a JS20 to provide
2691 two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2692 wired to I/O Modules 3 and 4, respectively.
2694 Each I/O Module may contain either a switch or a passthrough
2695 module (which allows ports to be directly connected to an external
2696 switch). Some bonding modes require a specific BladeCenter internal
2697 network topology in order to function; these are detailed below.
2699 Additional BladeCenter-specific networking information can be
2700 found in two IBM Redbooks (www.ibm.com/redbooks):
2702 - "IBM eServer BladeCenter Networking Options"
2703 - "IBM eServer BladeCenter Layer 2-7 Network Switching"
2705 BladeCenter networking configuration
2706 ------------------------------------
2708 Because a BladeCenter can be configured in a very large number
2709 of ways, this discussion will be confined to describing basic
2712 Normally, Ethernet Switch Modules (ESMs) are used in I/O
2713 modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2714 JS20 will be connected to different internal switches (in the
2715 respective I/O modules).
2717 A passthrough module (OPM or CPM, optical or copper,
2718 passthrough module) connects the I/O module directly to an external
2719 switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2720 interfaces of a JS20 can be redirected to the outside world and
2721 connected to a common external switch.
2723 Depending upon the mix of ESMs and PMs, the network will
2724 appear to bonding as either a single switch topology (all PMs) or as a
2725 multiple switch topology (one or more ESMs, zero or more PMs). It is
2726 also possible to connect ESMs together, resulting in a configuration
2727 much like the example in "High Availability in a Multiple Switch
2730 Requirements for specific modes
2731 -------------------------------
2733 The balance-rr mode requires the use of passthrough modules
2734 for devices in the bond, all connected to an common external switch.
2735 That switch must be configured for "etherchannel" or "trunking" on the
2736 appropriate ports, as is usual for balance-rr.
2738 The balance-alb and balance-tlb modes will function with
2739 either switch modules or passthrough modules (or a mix). The only
2740 specific requirement for these modes is that all network interfaces
2741 must be able to reach all destinations for traffic sent over the
2742 bonding device (i.e., the network must converge at some point outside
2745 The active-backup mode has no additional requirements.
2747 Link monitoring issues
2748 ----------------------
2750 When an Ethernet Switch Module is in place, only the ARP
2751 monitor will reliably detect link loss to an external switch. This is
2752 nothing unusual, but examination of the BladeCenter cabinet would
2753 suggest that the "external" network ports are the ethernet ports for
2754 the system, when it fact there is a switch between these "external"
2755 ports and the devices on the JS20 system itself. The MII monitor is
2756 only able to detect link failures between the ESM and the JS20 system.
2758 When a passthrough module is in place, the MII monitor does
2759 detect failures to the "external" port, which is then directly
2760 connected to the JS20 system.
2765 The Serial Over LAN (SoL) link is established over the primary
2766 ethernet (eth0) only, therefore, any loss of link to eth0 will result
2767 in losing your SoL connection. It will not fail over with other
2768 network traffic, as the SoL system is beyond the control of the
2771 It may be desirable to disable spanning tree on the switch
2772 (either the internal Ethernet Switch Module, or an external switch) to
2773 avoid fail-over delay issues when using bonding.
2776 15. Frequently Asked Questions
2777 ==============================
2782 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2783 The new driver was designed to be SMP safe from the start.
2785 2. What type of cards will work with it?
2786 -----------------------------------------
2788 Any Ethernet type cards (you can even mix cards - a Intel
2789 EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2790 devices need not be of the same speed.
2792 Starting with version 3.2.1, bonding also supports Infiniband
2793 slaves in active-backup mode.
2795 3. How many bonding devices can I have?
2796 ----------------------------------------
2800 4. How many slaves can a bonding device have?
2801 ----------------------------------------------
2803 This is limited only by the number of network interfaces Linux
2804 supports and/or the number of network cards you can place in your
2807 5. What happens when a slave link dies?
2808 ----------------------------------------
2810 If link monitoring is enabled, then the failing device will be
2811 disabled. The active-backup mode will fail over to a backup link, and
2812 other modes will ignore the failed link. The link will continue to be
2813 monitored, and should it recover, it will rejoin the bond (in whatever
2814 manner is appropriate for the mode). See the sections on High
2815 Availability and the documentation for each mode for additional
2818 Link monitoring can be enabled via either the miimon or
2819 arp_interval parameters (described in the module parameters section,
2820 above). In general, miimon monitors the carrier state as sensed by
2821 the underlying network device, and the arp monitor (arp_interval)
2822 monitors connectivity to another host on the local network.
2824 If no link monitoring is configured, the bonding driver will
2825 be unable to detect link failures, and will assume that all links are
2826 always available. This will likely result in lost packets, and a
2827 resulting degradation of performance. The precise performance loss
2828 depends upon the bonding mode and network configuration.
2830 6. Can bonding be used for High Availability?
2831 ----------------------------------------------
2833 Yes. See the section on High Availability for details.
2835 7. Which switches/systems does it work with?
2836 ---------------------------------------------
2838 The full answer to this depends upon the desired mode.
2840 In the basic balance modes (balance-rr and balance-xor), it
2841 works with any system that supports etherchannel (also called
2842 trunking). Most managed switches currently available have such
2843 support, and many unmanaged switches as well.
2845 The advanced balance modes (balance-tlb and balance-alb) do
2846 not have special switch requirements, but do need device drivers that
2847 support specific features (described in the appropriate section under
2848 module parameters, above).
2850 In 802.3ad mode, it works with systems that support IEEE
2851 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2852 switches currently available support 802.3ad.
2854 The active-backup mode should work with any Layer-II switch.
2856 8. Where does a bonding device get its MAC address from?
2857 ---------------------------------------------------------
2859 When using slave devices that have fixed MAC addresses, or when
2860 the fail_over_mac option is enabled, the bonding device's MAC address is
2861 the MAC address of the active slave.
2863 For other configurations, if not explicitly configured (with
2864 ifconfig or ip link), the MAC address of the bonding device is taken from
2865 its first slave device. This MAC address is then passed to all following
2866 slaves and remains persistent (even if the first slave is removed) until
2867 the bonding device is brought down or reconfigured.
2869 If you wish to change the MAC address, you can set it with
2870 ifconfig or ip link::
2872 # ifconfig bond0 hw ether 00:11:22:33:44:55
2874 # ip link set bond0 address 66:77:88:99:aa:bb
2876 The MAC address can be also changed by bringing down/up the
2877 device and then changing its slaves (or their order)::
2879 # ifconfig bond0 down ; modprobe -r bonding
2880 # ifconfig bond0 .... up
2881 # ifenslave bond0 eth...
2883 This method will automatically take the address from the next
2884 slave that is added.
2886 To restore your slaves' MAC addresses, you need to detach them
2887 from the bond (``ifenslave -d bond0 eth0``). The bonding driver will
2888 then restore the MAC addresses that the slaves had before they were
2891 16. Resources and Links
2892 =======================
2894 The latest version of the bonding driver can be found in the latest
2895 version of the linux kernel, found on http://kernel.org
2897 The latest version of this document can be found in the latest kernel
2898 source (named Documentation/networking/bonding.rst).
2900 Discussions regarding the development of the bonding driver take place
2901 on the main Linux network mailing list, hosted at vger.kernel.org. The list
2904 netdev@vger.kernel.org
2906 The administrative interface (to subscribe or unsubscribe) can
2909 http://vger.kernel.org/vger-lists.html#netdev