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 NULL or
200 multicast. It is preferred to have the local-admin bit set for this
201 mac but driver does not enforce it. If the value is not given then
202 system defaults to using the masters' mac address as actors' system
205 This parameter has effect only in 802.3ad mode and is available through
210 Specifies the 802.3ad aggregation selection logic to use. The
211 possible values and their effects are:
215 The active aggregator is chosen by largest aggregate
218 Reselection of the active aggregator occurs only when all
219 slaves of the active aggregator are down or the active
220 aggregator has no slaves.
222 This is the default value.
226 The active aggregator is chosen by largest aggregate
227 bandwidth. Reselection occurs if:
229 - A slave is added to or removed from the bond
231 - Any slave's link state changes
233 - Any slave's 802.3ad association state changes
235 - The bond's administrative state changes to up
239 The active aggregator is chosen by the largest number of
240 ports (slaves). Reselection occurs as described under the
241 "bandwidth" setting, above.
243 The bandwidth and count selection policies permit failover of
244 802.3ad aggregations when partial failure of the active aggregator
245 occurs. This keeps the aggregator with the highest availability
246 (either in bandwidth or in number of ports) active at all times.
248 This option was added in bonding version 3.4.0.
252 In an AD system, the port-key has three parts as shown below -
262 This defines the upper 10 bits of the port key. The values can be
263 from 0 - 1023. If not given, the system defaults to 0.
265 This parameter has effect only in 802.3ad mode and is available through
270 Specifies that duplicate frames (received on inactive ports) should be
271 dropped (0) or delivered (1).
273 Normally, bonding will drop duplicate frames (received on inactive
274 ports), which is desirable for most users. But there are some times
275 it is nice to allow duplicate frames to be delivered.
277 The default value is 0 (drop duplicate frames received on inactive
282 Specifies the ARP link monitoring frequency in milliseconds.
284 The ARP monitor works by periodically checking the slave
285 devices to determine whether they have sent or received
286 traffic recently (the precise criteria depends upon the
287 bonding mode, and the state of the slave). Regular traffic is
288 generated via ARP probes issued for the addresses specified by
289 the arp_ip_target option.
291 This behavior can be modified by the arp_validate option,
294 If ARP monitoring is used in an etherchannel compatible mode
295 (modes 0 and 2), the switch should be configured in a mode
296 that evenly distributes packets across all links. If the
297 switch is configured to distribute the packets in an XOR
298 fashion, all replies from the ARP targets will be received on
299 the same link which could cause the other team members to
300 fail. ARP monitoring should not be used in conjunction with
301 miimon. A value of 0 disables ARP monitoring. The default
306 Specifies the IP addresses to use as ARP monitoring peers when
307 arp_interval is > 0. These are the targets of the ARP request
308 sent to determine the health of the link to the targets.
309 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
310 addresses must be separated by a comma. At least one IP
311 address must be given for ARP monitoring to function. The
312 maximum number of targets that can be specified is 16. The
313 default value is no IP addresses.
317 Specifies whether or not ARP probes and replies should be
318 validated in any mode that supports arp monitoring, or whether
319 non-ARP traffic should be filtered (disregarded) for link
326 No validation or filtering is performed.
330 Validation is performed only for the active slave.
334 Validation is performed only for backup slaves.
338 Validation is performed for all slaves.
342 Filtering is applied to all slaves. No validation is
347 Filtering is applied to all slaves, validation is performed
348 only for the active slave.
352 Filtering is applied to all slaves, validation is performed
353 only for backup slaves.
357 Enabling validation causes the ARP monitor to examine the incoming
358 ARP requests and replies, and only consider a slave to be up if it
359 is receiving the appropriate ARP traffic.
361 For an active slave, the validation checks ARP replies to confirm
362 that they were generated by an arp_ip_target. Since backup slaves
363 do not typically receive these replies, the validation performed
364 for backup slaves is on the broadcast ARP request sent out via the
365 active slave. It is possible that some switch or network
366 configurations may result in situations wherein the backup slaves
367 do not receive the ARP requests; in such a situation, validation
368 of backup slaves must be disabled.
370 The validation of ARP requests on backup slaves is mainly helping
371 bonding to decide which slaves are more likely to work in case of
372 the active slave failure, it doesn't really guarantee that the
373 backup slave will work if it's selected as the next active slave.
375 Validation is useful in network configurations in which multiple
376 bonding hosts are concurrently issuing ARPs to one or more targets
377 beyond a common switch. Should the link between the switch and
378 target fail (but not the switch itself), the probe traffic
379 generated by the multiple bonding instances will fool the standard
380 ARP monitor into considering the links as still up. Use of
381 validation can resolve this, as the ARP monitor will only consider
382 ARP requests and replies associated with its own instance of
387 Enabling filtering causes the ARP monitor to only use incoming ARP
388 packets for link availability purposes. Arriving packets that are
389 not ARPs are delivered normally, but do not count when determining
390 if a slave is available.
392 Filtering operates by only considering the reception of ARP
393 packets (any ARP packet, regardless of source or destination) when
394 determining if a slave has received traffic for link availability
397 Filtering is useful in network configurations in which significant
398 levels of third party broadcast traffic would fool the standard
399 ARP monitor into considering the links as still up. Use of
400 filtering can resolve this, as only ARP traffic is considered for
401 link availability purposes.
403 This option was added in bonding version 3.1.0.
407 Specifies the quantity of arp_ip_targets that must be reachable
408 in order for the ARP monitor to consider a slave as being up.
409 This option affects only active-backup mode for slaves with
410 arp_validation enabled.
416 consider the slave up only when any of the arp_ip_targets
421 consider the slave up only when all of the arp_ip_targets
426 Specifies the time, in milliseconds, to wait before disabling
427 a slave after a link failure has been detected. This option
428 is only valid for the miimon link monitor. The downdelay
429 value should be a multiple of the miimon value; if not, it
430 will be rounded down to the nearest multiple. The default
435 Specifies whether active-backup mode should set all slaves to
436 the same MAC address at enslavement (the traditional
437 behavior), or, when enabled, perform special handling of the
438 bond's MAC address in accordance with the selected policy.
444 This setting disables fail_over_mac, and causes
445 bonding to set all slaves of an active-backup bond to
446 the same MAC address at enslavement time. This is the
451 The "active" fail_over_mac policy indicates that the
452 MAC address of the bond should always be the MAC
453 address of the currently active slave. The MAC
454 address of the slaves is not changed; instead, the MAC
455 address of the bond changes during a failover.
457 This policy is useful for devices that cannot ever
458 alter their MAC address, or for devices that refuse
459 incoming broadcasts with their own source MAC (which
460 interferes with the ARP monitor).
462 The down side of this policy is that every device on
463 the network must be updated via gratuitous ARP,
464 vs. just updating a switch or set of switches (which
465 often takes place for any traffic, not just ARP
466 traffic, if the switch snoops incoming traffic to
467 update its tables) for the traditional method. If the
468 gratuitous ARP is lost, communication may be
471 When this policy is used in conjunction with the mii
472 monitor, devices which assert link up prior to being
473 able to actually transmit and receive are particularly
474 susceptible to loss of the gratuitous ARP, and an
475 appropriate updelay setting may be required.
479 The "follow" fail_over_mac policy causes the MAC
480 address of the bond to be selected normally (normally
481 the MAC address of the first slave added to the bond).
482 However, the second and subsequent slaves are not set
483 to this MAC address while they are in a backup role; a
484 slave is programmed with the bond's MAC address at
485 failover time (and the formerly active slave receives
486 the newly active slave's MAC address).
488 This policy is useful for multiport devices that
489 either become confused or incur a performance penalty
490 when multiple ports are programmed with the same MAC
494 The default policy is none, unless the first slave cannot
495 change its MAC address, in which case the active policy is
498 This option may be modified via sysfs only when no slaves are
501 This option was added in bonding version 3.2.0. The "follow"
502 policy was added in bonding version 3.3.0.
506 Option specifying the rate in which we'll ask our link partner
507 to transmit LACPDU packets in 802.3ad mode. Possible values
511 Request partner to transmit LACPDUs every 30 seconds
514 Request partner to transmit LACPDUs every 1 second
520 Specifies the number of bonding devices to create for this
521 instance of the bonding driver. E.g., if max_bonds is 3, and
522 the bonding driver is not already loaded, then bond0, bond1
523 and bond2 will be created. The default value is 1. Specifying
524 a value of 0 will load bonding, but will not create any devices.
528 Specifies the MII link monitoring frequency in milliseconds.
529 This determines how often the link state of each slave is
530 inspected for link failures. A value of zero disables MII
531 link monitoring. A value of 100 is a good starting point.
532 The use_carrier option, below, affects how the link state is
533 determined. See the High Availability section for additional
534 information. The default value is 0.
538 Specifies the minimum number of links that must be active before
539 asserting carrier. It is similar to the Cisco EtherChannel min-links
540 feature. This allows setting the minimum number of member ports that
541 must be up (link-up state) before marking the bond device as up
542 (carrier on). This is useful for situations where higher level services
543 such as clustering want to ensure a minimum number of low bandwidth
544 links are active before switchover. This option only affect 802.3ad
547 The default value is 0. This will cause carrier to be asserted (for
548 802.3ad mode) whenever there is an active aggregator, regardless of the
549 number of available links in that aggregator. Note that, because an
550 aggregator cannot be active without at least one available link,
551 setting this option to 0 or to 1 has the exact same effect.
555 Specifies one of the bonding policies. The default is
556 balance-rr (round robin). Possible values are:
560 Round-robin policy: Transmit packets in sequential
561 order from the first available slave through the
562 last. This mode provides load balancing and fault
567 Active-backup policy: Only one slave in the bond is
568 active. A different slave becomes active if, and only
569 if, the active slave fails. The bond's MAC address is
570 externally visible on only one port (network adapter)
571 to avoid confusing the switch.
573 In bonding version 2.6.2 or later, when a failover
574 occurs in active-backup mode, bonding will issue one
575 or more gratuitous ARPs on the newly active slave.
576 One gratuitous ARP is issued for the bonding master
577 interface and each VLAN interfaces configured above
578 it, provided that the interface has at least one IP
579 address configured. Gratuitous ARPs issued for VLAN
580 interfaces are tagged with the appropriate VLAN id.
582 This mode provides fault tolerance. The primary
583 option, documented below, affects the behavior of this
588 XOR policy: Transmit based on the selected transmit
589 hash policy. The default policy is a simple [(source
590 MAC address XOR'd with destination MAC address XOR
591 packet type ID) modulo slave count]. Alternate transmit
592 policies may be selected via the xmit_hash_policy option,
595 This mode provides load balancing and fault tolerance.
599 Broadcast policy: transmits everything on all slave
600 interfaces. This mode provides fault tolerance.
604 IEEE 802.3ad Dynamic link aggregation. Creates
605 aggregation groups that share the same speed and
606 duplex settings. Utilizes all slaves in the active
607 aggregator according to the 802.3ad specification.
609 Slave selection for outgoing traffic is done according
610 to the transmit hash policy, which may be changed from
611 the default simple XOR policy via the xmit_hash_policy
612 option, documented below. Note that not all transmit
613 policies may be 802.3ad compliant, particularly in
614 regards to the packet mis-ordering requirements of
615 section 43.2.4 of the 802.3ad standard. Differing
616 peer implementations will have varying tolerances for
621 1. Ethtool support in the base drivers for retrieving
622 the speed and duplex of each slave.
624 2. A switch that supports IEEE 802.3ad Dynamic link
627 Most switches will require some type of configuration
628 to enable 802.3ad mode.
632 Adaptive transmit load balancing: channel bonding that
633 does not require any special switch support.
635 In tlb_dynamic_lb=1 mode; the outgoing traffic is
636 distributed according to the current load (computed
637 relative to the speed) on each slave.
639 In tlb_dynamic_lb=0 mode; the load balancing based on
640 current load is disabled and the load is distributed
641 only using the hash distribution.
643 Incoming traffic is received by the current slave.
644 If the receiving slave fails, another slave takes over
645 the MAC address of the failed receiving slave.
649 Ethtool support in the base drivers for retrieving the
654 Adaptive load balancing: includes balance-tlb plus
655 receive load balancing (rlb) for IPV4 traffic, and
656 does not require any special switch support. The
657 receive load balancing is achieved by ARP negotiation.
658 The bonding driver intercepts the ARP Replies sent by
659 the local system on their way out and overwrites the
660 source hardware address with the unique hardware
661 address of one of the slaves in the bond such that
662 different peers use different hardware addresses for
665 Receive traffic from connections created by the server
666 is also balanced. When the local system sends an ARP
667 Request the bonding driver copies and saves the peer's
668 IP information from the ARP packet. When the ARP
669 Reply arrives from the peer, its hardware address is
670 retrieved and the bonding driver initiates an ARP
671 reply to this peer assigning it to one of the slaves
672 in the bond. A problematic outcome of using ARP
673 negotiation for balancing is that each time that an
674 ARP request is broadcast it uses the hardware address
675 of the bond. Hence, peers learn the hardware address
676 of the bond and the balancing of receive traffic
677 collapses to the current slave. This is handled by
678 sending updates (ARP Replies) to all the peers with
679 their individually assigned hardware address such that
680 the traffic is redistributed. Receive traffic is also
681 redistributed when a new slave is added to the bond
682 and when an inactive slave is re-activated. The
683 receive load is distributed sequentially (round robin)
684 among the group of highest speed slaves in the bond.
686 When a link is reconnected or a new slave joins the
687 bond the receive traffic is redistributed among all
688 active slaves in the bond by initiating ARP Replies
689 with the selected MAC address to each of the
690 clients. The updelay parameter (detailed below) must
691 be set to a value equal or greater than the switch's
692 forwarding delay so that the ARP Replies sent to the
693 peers will not be blocked by the switch.
697 1. Ethtool support in the base drivers for retrieving
698 the speed of each slave.
700 2. Base driver support for setting the hardware
701 address of a device while it is open. This is
702 required so that there will always be one slave in the
703 team using the bond hardware address (the
704 curr_active_slave) while having a unique hardware
705 address for each slave in the bond. If the
706 curr_active_slave fails its hardware address is
707 swapped with the new curr_active_slave that was
713 Specify the number of peer notifications (gratuitous ARPs and
714 unsolicited IPv6 Neighbor Advertisements) to be issued after a
715 failover event. As soon as the link is up on the new slave
716 (possibly immediately) a peer notification is sent on the
717 bonding device and each VLAN sub-device. This is repeated at
718 the rate specified by peer_notif_delay if the number is
721 The valid range is 0 - 255; the default value is 1. These options
722 affect only the active-backup mode. These options were added for
723 bonding versions 3.3.0 and 3.4.0 respectively.
725 From Linux 3.0 and bonding version 3.7.1, these notifications
726 are generated by the ipv4 and ipv6 code and the numbers of
727 repetitions cannot be set independently.
731 Specify the number of packets to transmit through a slave before
732 moving to the next one. When set to 0 then a slave is chosen at
735 The valid range is 0 - 65535; the default value is 1. This option
736 has effect only in balance-rr mode.
740 Specify the delay, in milliseconds, between each peer
741 notification (gratuitous ARP and unsolicited IPv6 Neighbor
742 Advertisement) when they are issued after a failover event.
743 This delay should be a multiple of the link monitor interval
744 (arp_interval or miimon, whichever is active). The default
745 value is 0 which means to match the value of the link monitor
750 A string (eth0, eth2, etc) specifying which slave is the
751 primary device. The specified device will always be the
752 active slave while it is available. Only when the primary is
753 off-line will alternate devices be used. This is useful when
754 one slave is preferred over another, e.g., when one slave has
755 higher throughput than another.
757 The primary option is only valid for active-backup(1),
758 balance-tlb (5) and balance-alb (6) mode.
762 Specifies the reselection policy for the primary slave. This
763 affects how the primary slave is chosen to become the active slave
764 when failure of the active slave or recovery of the primary slave
765 occurs. This option is designed to prevent flip-flopping between
766 the primary slave and other slaves. Possible values are:
768 always or 0 (default)
770 The primary slave becomes the active slave whenever it
775 The primary slave becomes the active slave when it comes
776 back up, if the speed and duplex of the primary slave is
777 better than the speed and duplex of the current active
782 The primary slave becomes the active slave only if the
783 current active slave fails and the primary slave is up.
785 The primary_reselect setting is ignored in two cases:
787 If no slaves are active, the first slave to recover is
788 made the active slave.
790 When initially enslaved, the primary slave is always made
793 Changing the primary_reselect policy via sysfs will cause an
794 immediate selection of the best active slave according to the new
795 policy. This may or may not result in a change of the active
796 slave, depending upon the circumstances.
798 This option was added for bonding version 3.6.0.
802 Specifies if dynamic shuffling of flows is enabled in tlb
803 mode. The value has no effect on any other modes.
805 The default behavior of tlb mode is to shuffle active flows across
806 slaves based on the load in that interval. This gives nice lb
807 characteristics but can cause packet reordering. If re-ordering is
808 a concern use this variable to disable flow shuffling and rely on
809 load balancing provided solely by the hash distribution.
810 xmit-hash-policy can be used to select the appropriate hashing for
813 The sysfs entry can be used to change the setting per bond device
814 and the initial value is derived from the module parameter. The
815 sysfs entry is allowed to be changed only if the bond device is
818 The default value is "1" that enables flow shuffling while value "0"
819 disables it. This option was added in bonding driver 3.7.1
824 Specifies the time, in milliseconds, to wait before enabling a
825 slave after a link recovery has been detected. This option is
826 only valid for the miimon link monitor. The updelay value
827 should be a multiple of the miimon value; if not, it will be
828 rounded down to the nearest multiple. The default value is 0.
832 Specifies whether or not miimon should use MII or ETHTOOL
833 ioctls vs. netif_carrier_ok() to determine the link
834 status. The MII or ETHTOOL ioctls are less efficient and
835 utilize a deprecated calling sequence within the kernel. The
836 netif_carrier_ok() relies on the device driver to maintain its
837 state with netif_carrier_on/off; at this writing, most, but
838 not all, device drivers support this facility.
840 If bonding insists that the link is up when it should not be,
841 it may be that your network device driver does not support
842 netif_carrier_on/off. The default state for netif_carrier is
843 "carrier on," so if a driver does not support netif_carrier,
844 it will appear as if the link is always up. In this case,
845 setting use_carrier to 0 will cause bonding to revert to the
846 MII / ETHTOOL ioctl method to determine the link state.
848 A value of 1 enables the use of netif_carrier_ok(), a value of
849 0 will use the deprecated MII / ETHTOOL ioctls. The default
854 Selects the transmit hash policy to use for slave selection in
855 balance-xor, 802.3ad, and tlb modes. Possible values are:
859 Uses XOR of hardware MAC addresses and packet type ID
860 field to generate the hash. The formula is
862 hash = source MAC XOR destination MAC XOR packet type ID
863 slave number = hash modulo slave count
865 This algorithm will place all traffic to a particular
866 network peer on the same slave.
868 This algorithm is 802.3ad compliant.
872 This policy uses a combination of layer2 and layer3
873 protocol information to generate the hash.
875 Uses XOR of hardware MAC addresses and IP addresses to
876 generate the hash. The formula is
878 hash = source MAC XOR destination MAC XOR packet type ID
879 hash = hash XOR source IP XOR destination IP
880 hash = hash XOR (hash RSHIFT 16)
881 hash = hash XOR (hash RSHIFT 8)
882 And then hash is reduced modulo slave count.
884 If the protocol is IPv6 then the source and destination
885 addresses are first hashed using ipv6_addr_hash.
887 This algorithm will place all traffic to a particular
888 network peer on the same slave. For non-IP traffic,
889 the formula is the same as for the layer2 transmit
892 This policy is intended to provide a more balanced
893 distribution of traffic than layer2 alone, especially
894 in environments where a layer3 gateway device is
895 required to reach most destinations.
897 This algorithm is 802.3ad compliant.
901 This policy uses upper layer protocol information,
902 when available, to generate the hash. This allows for
903 traffic to a particular network peer to span multiple
904 slaves, although a single connection will not span
907 The formula for unfragmented TCP and UDP packets is
909 hash = source port, destination port (as in the header)
910 hash = hash XOR source IP XOR destination IP
911 hash = hash XOR (hash RSHIFT 16)
912 hash = hash XOR (hash RSHIFT 8)
913 And then hash is reduced modulo slave count.
915 If the protocol is IPv6 then the source and destination
916 addresses are first hashed using ipv6_addr_hash.
918 For fragmented TCP or UDP packets and all other IPv4 and
919 IPv6 protocol traffic, the source and destination port
920 information is omitted. For non-IP traffic, the
921 formula is the same as for the layer2 transmit hash
924 This algorithm is not fully 802.3ad compliant. A
925 single TCP or UDP conversation containing both
926 fragmented and unfragmented packets will see packets
927 striped across two interfaces. This may result in out
928 of order delivery. Most traffic types will not meet
929 this criteria, as TCP rarely fragments traffic, and
930 most UDP traffic is not involved in extended
931 conversations. Other implementations of 802.3ad may
932 or may not tolerate this noncompliance.
936 This policy uses the same formula as layer2+3 but it
937 relies on skb_flow_dissect to obtain the header fields
938 which might result in the use of inner headers if an
939 encapsulation protocol is used. For example this will
940 improve the performance for tunnel users because the
941 packets will be distributed according to the encapsulated
946 This policy uses the same formula as layer3+4 but it
947 relies on skb_flow_dissect to obtain the header fields
948 which might result in the use of inner headers if an
949 encapsulation protocol is used. For example this will
950 improve the performance for tunnel users because the
951 packets will be distributed according to the encapsulated
956 This policy uses a very rudimentary vlan ID and source mac
957 hash to load-balance traffic per-vlan, with failover
958 should one leg fail. The intended use case is for a bond
959 shared by multiple virtual machines, all configured to
960 use their own vlan, to give lacp-like functionality
961 without requiring lacp-capable switching hardware.
963 The formula for the hash is simply
965 hash = (vlan ID) XOR (source MAC vendor) XOR (source MAC dev)
967 The default value is layer2. This option was added in bonding
968 version 2.6.3. In earlier versions of bonding, this parameter
969 does not exist, and the layer2 policy is the only policy. The
970 layer2+3 value was added for bonding version 3.2.2.
974 Specifies the number of IGMP membership reports to be issued after
975 a failover event. One membership report is issued immediately after
976 the failover, subsequent packets are sent in each 200ms interval.
978 The valid range is 0 - 255; the default value is 1. A value of 0
979 prevents the IGMP membership report from being issued in response
980 to the failover event.
982 This option is useful for bonding modes balance-rr (0), active-backup
983 (1), balance-tlb (5) and balance-alb (6), in which a failover can
984 switch the IGMP traffic from one slave to another. Therefore a fresh
985 IGMP report must be issued to cause the switch to forward the incoming
986 IGMP traffic over the newly selected slave.
988 This option was added for bonding version 3.7.0.
992 Specifies the number of seconds between instances where the bonding
993 driver sends learning packets to each slaves peer switch.
995 The valid range is 1 - 0x7fffffff; the default value is 1. This Option
996 has effect only in balance-tlb and balance-alb modes.
998 3. Configuring Bonding Devices
999 ==============================
1001 You can configure bonding using either your distro's network
1002 initialization scripts, or manually using either iproute2 or the
1003 sysfs interface. Distros generally use one of three packages for the
1004 network initialization scripts: initscripts, sysconfig or interfaces.
1005 Recent versions of these packages have support for bonding, while older
1008 We will first describe the options for configuring bonding for
1009 distros using versions of initscripts, sysconfig and interfaces with full
1010 or partial support for bonding, then provide information on enabling
1011 bonding without support from the network initialization scripts (i.e.,
1012 older versions of initscripts or sysconfig).
1014 If you're unsure whether your distro uses sysconfig,
1015 initscripts or interfaces, or don't know if it's new enough, have no fear.
1016 Determining this is fairly straightforward.
1018 First, look for a file called interfaces in /etc/network directory.
1019 If this file is present in your system, then your system use interfaces. See
1020 Configuration with Interfaces Support.
1022 Else, issue the command::
1024 $ rpm -qf /sbin/ifup
1026 It will respond with a line of text starting with either
1027 "initscripts" or "sysconfig," followed by some numbers. This is the
1028 package that provides your network initialization scripts.
1030 Next, to determine if your installation supports bonding,
1033 $ grep ifenslave /sbin/ifup
1035 If this returns any matches, then your initscripts or
1036 sysconfig has support for bonding.
1038 3.1 Configuration with Sysconfig Support
1039 ----------------------------------------
1041 This section applies to distros using a version of sysconfig
1042 with bonding support, for example, SuSE Linux Enterprise Server 9.
1044 SuSE SLES 9's networking configuration system does support
1045 bonding, however, at this writing, the YaST system configuration
1046 front end does not provide any means to work with bonding devices.
1047 Bonding devices can be managed by hand, however, as follows.
1049 First, if they have not already been configured, configure the
1050 slave devices. On SLES 9, this is most easily done by running the
1051 yast2 sysconfig configuration utility. The goal is for to create an
1052 ifcfg-id file for each slave device. The simplest way to accomplish
1053 this is to configure the devices for DHCP (this is only to get the
1054 file ifcfg-id file created; see below for some issues with DHCP). The
1055 name of the configuration file for each device will be of the form::
1057 ifcfg-id-xx:xx:xx:xx:xx:xx
1059 Where the "xx" portion will be replaced with the digits from
1060 the device's permanent MAC address.
1062 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
1063 created, it is necessary to edit the configuration files for the slave
1064 devices (the MAC addresses correspond to those of the slave devices).
1065 Before editing, the file will contain multiple lines, and will look
1066 something like this::
1071 UNIQUE='XNzu.WeZGOGF+4wE'
1072 _nm_name='bus-pci-0001:61:01.0'
1074 Change the BOOTPROTO and STARTMODE lines to the following::
1079 Do not alter the UNIQUE or _nm_name lines. Remove any other
1080 lines (USERCTL, etc).
1082 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
1083 it's time to create the configuration file for the bonding device
1084 itself. This file is named ifcfg-bondX, where X is the number of the
1085 bonding device to create, starting at 0. The first such file is
1086 ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
1087 network configuration system will correctly start multiple instances
1090 The contents of the ifcfg-bondX file is as follows::
1093 BROADCAST="10.0.2.255"
1095 NETMASK="255.255.0.0"
1099 BONDING_MASTER="yes"
1100 BONDING_MODULE_OPTS="mode=active-backup miimon=100"
1101 BONDING_SLAVE0="eth0"
1102 BONDING_SLAVE1="bus-pci-0000:06:08.1"
1104 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
1105 values with the appropriate values for your network.
1107 The STARTMODE specifies when the device is brought online.
1108 The possible values are:
1110 ======== ======================================================
1111 onboot The device is started at boot time. If you're not
1112 sure, this is probably what you want.
1114 manual The device is started only when ifup is called
1115 manually. Bonding devices may be configured this
1116 way if you do not wish them to start automatically
1117 at boot for some reason.
1119 hotplug The device is started by a hotplug event. This is not
1120 a valid choice for a bonding device.
1122 off or The device configuration is ignored.
1124 ======== ======================================================
1126 The line BONDING_MASTER='yes' indicates that the device is a
1127 bonding master device. The only useful value is "yes."
1129 The contents of BONDING_MODULE_OPTS are supplied to the
1130 instance of the bonding module for this device. Specify the options
1131 for the bonding mode, link monitoring, and so on here. Do not include
1132 the max_bonds bonding parameter; this will confuse the configuration
1133 system if you have multiple bonding devices.
1135 Finally, supply one BONDING_SLAVEn="slave device" for each
1136 slave. where "n" is an increasing value, one for each slave. The
1137 "slave device" is either an interface name, e.g., "eth0", or a device
1138 specifier for the network device. The interface name is easier to
1139 find, but the ethN names are subject to change at boot time if, e.g.,
1140 a device early in the sequence has failed. The device specifiers
1141 (bus-pci-0000:06:08.1 in the example above) specify the physical
1142 network device, and will not change unless the device's bus location
1143 changes (for example, it is moved from one PCI slot to another). The
1144 example above uses one of each type for demonstration purposes; most
1145 configurations will choose one or the other for all slave devices.
1147 When all configuration files have been modified or created,
1148 networking must be restarted for the configuration changes to take
1149 effect. This can be accomplished via the following::
1151 # /etc/init.d/network restart
1153 Note that the network control script (/sbin/ifdown) will
1154 remove the bonding module as part of the network shutdown processing,
1155 so it is not necessary to remove the module by hand if, e.g., the
1156 module parameters have changed.
1158 Also, at this writing, YaST/YaST2 will not manage bonding
1159 devices (they do not show bonding interfaces on its list of network
1160 devices). It is necessary to edit the configuration file by hand to
1161 change the bonding configuration.
1163 Additional general options and details of the ifcfg file
1164 format can be found in an example ifcfg template file::
1166 /etc/sysconfig/network/ifcfg.template
1168 Note that the template does not document the various ``BONDING_*``
1169 settings described above, but does describe many of the other options.
1171 3.1.1 Using DHCP with Sysconfig
1172 -------------------------------
1174 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1175 will cause it to query DHCP for its IP address information. At this
1176 writing, this does not function for bonding devices; the scripts
1177 attempt to obtain the device address from DHCP prior to adding any of
1178 the slave devices. Without active slaves, the DHCP requests are not
1179 sent to the network.
1181 3.1.2 Configuring Multiple Bonds with Sysconfig
1182 -----------------------------------------------
1184 The sysconfig network initialization system is capable of
1185 handling multiple bonding devices. All that is necessary is for each
1186 bonding instance to have an appropriately configured ifcfg-bondX file
1187 (as described above). Do not specify the "max_bonds" parameter to any
1188 instance of bonding, as this will confuse sysconfig. If you require
1189 multiple bonding devices with identical parameters, create multiple
1192 Because the sysconfig scripts supply the bonding module
1193 options in the ifcfg-bondX file, it is not necessary to add them to
1194 the system ``/etc/modules.d/*.conf`` configuration files.
1196 3.2 Configuration with Initscripts Support
1197 ------------------------------------------
1199 This section applies to distros using a recent version of
1200 initscripts with bonding support, for example, Red Hat Enterprise Linux
1201 version 3 or later, Fedora, etc. On these systems, the network
1202 initialization scripts have knowledge of bonding, and can be configured to
1203 control bonding devices. Note that older versions of the initscripts
1204 package have lower levels of support for bonding; this will be noted where
1207 These distros will not automatically load the network adapter
1208 driver unless the ethX device is configured with an IP address.
1209 Because of this constraint, users must manually configure a
1210 network-script file for all physical adapters that will be members of
1211 a bondX link. Network script files are located in the directory:
1213 /etc/sysconfig/network-scripts
1215 The file name must be prefixed with "ifcfg-eth" and suffixed
1216 with the adapter's physical adapter number. For example, the script
1217 for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1218 Place the following text in the file::
1227 The DEVICE= line will be different for every ethX device and
1228 must correspond with the name of the file, i.e., ifcfg-eth1 must have
1229 a device line of DEVICE=eth1. The setting of the MASTER= line will
1230 also depend on the final bonding interface name chosen for your bond.
1231 As with other network devices, these typically start at 0, and go up
1232 one for each device, i.e., the first bonding instance is bond0, the
1233 second is bond1, and so on.
1235 Next, create a bond network script. The file name for this
1236 script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1237 the number of the bond. For bond0 the file is named "ifcfg-bond0",
1238 for bond1 it is named "ifcfg-bond1", and so on. Within that file,
1239 place the following text::
1243 NETMASK=255.255.255.0
1245 BROADCAST=192.168.1.255
1250 Be sure to change the networking specific lines (IPADDR,
1251 NETMASK, NETWORK and BROADCAST) to match your network configuration.
1253 For later versions of initscripts, such as that found with Fedora
1254 7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1255 and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
1256 file, e.g. a line of the format::
1258 BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
1260 will configure the bond with the specified options. The options
1261 specified in BONDING_OPTS are identical to the bonding module parameters
1262 except for the arp_ip_target field when using versions of initscripts older
1263 than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
1264 using older versions each target should be included as a separate option and
1265 should be preceded by a '+' to indicate it should be added to the list of
1266 queried targets, e.g.,::
1268 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1270 is the proper syntax to specify multiple targets. When specifying
1271 options via BONDING_OPTS, it is not necessary to edit
1272 ``/etc/modprobe.d/*.conf``.
1274 For even older versions of initscripts that do not support
1275 BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1276 your distro) to load the bonding module with your desired options when the
1277 bond0 interface is brought up. The following lines in /etc/modprobe.d/*.conf
1278 will load the bonding module, and select its options:
1281 options bond0 mode=balance-alb miimon=100
1283 Replace the sample parameters with the appropriate set of
1284 options for your configuration.
1286 Finally run "/etc/rc.d/init.d/network restart" as root. This
1287 will restart the networking subsystem and your bond link should be now
1290 3.2.1 Using DHCP with Initscripts
1291 ---------------------------------
1293 Recent versions of initscripts (the versions supplied with Fedora
1294 Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1295 work) have support for assigning IP information to bonding devices via
1298 To configure bonding for DHCP, configure it as described
1299 above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1300 and add a line consisting of "TYPE=Bonding". Note that the TYPE value
1303 3.2.2 Configuring Multiple Bonds with Initscripts
1304 -------------------------------------------------
1306 Initscripts packages that are included with Fedora 7 and Red Hat
1307 Enterprise Linux 5 support multiple bonding interfaces by simply
1308 specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1309 number of the bond. This support requires sysfs support in the kernel,
1310 and a bonding driver of version 3.0.0 or later. Other configurations may
1311 not support this method for specifying multiple bonding interfaces; for
1312 those instances, see the "Configuring Multiple Bonds Manually" section,
1315 3.3 Configuring Bonding Manually with iproute2
1316 -----------------------------------------------
1318 This section applies to distros whose network initialization
1319 scripts (the sysconfig or initscripts package) do not have specific
1320 knowledge of bonding. One such distro is SuSE Linux Enterprise Server
1323 The general method for these systems is to place the bonding
1324 module parameters into a config file in /etc/modprobe.d/ (as
1325 appropriate for the installed distro), then add modprobe and/or
1326 `ip link` commands to the system's global init script. The name of
1327 the global init script differs; for sysconfig, it is
1328 /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1330 For example, if you wanted to make a simple bond of two e100
1331 devices (presumed to be eth0 and eth1), and have it persist across
1332 reboots, edit the appropriate file (/etc/init.d/boot.local or
1333 /etc/rc.d/rc.local), and add the following::
1335 modprobe bonding mode=balance-alb miimon=100
1337 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1338 ip link set eth0 master bond0
1339 ip link set eth1 master bond0
1341 Replace the example bonding module parameters and bond0
1342 network configuration (IP address, netmask, etc) with the appropriate
1343 values for your configuration.
1345 Unfortunately, this method will not provide support for the
1346 ifup and ifdown scripts on the bond devices. To reload the bonding
1347 configuration, it is necessary to run the initialization script, e.g.,::
1349 # /etc/init.d/boot.local
1353 # /etc/rc.d/rc.local
1355 It may be desirable in such a case to create a separate script
1356 which only initializes the bonding configuration, then call that
1357 separate script from within boot.local. This allows for bonding to be
1358 enabled without re-running the entire global init script.
1360 To shut down the bonding devices, it is necessary to first
1361 mark the bonding device itself as being down, then remove the
1362 appropriate device driver modules. For our example above, you can do
1365 # ifconfig bond0 down
1369 Again, for convenience, it may be desirable to create a script
1370 with these commands.
1373 3.3.1 Configuring Multiple Bonds Manually
1374 -----------------------------------------
1376 This section contains information on configuring multiple
1377 bonding devices with differing options for those systems whose network
1378 initialization scripts lack support for configuring multiple bonds.
1380 If you require multiple bonding devices, but all with the same
1381 options, you may wish to use the "max_bonds" module parameter,
1384 To create multiple bonding devices with differing options, it is
1385 preferable to use bonding parameters exported by sysfs, documented in the
1388 For versions of bonding without sysfs support, the only means to
1389 provide multiple instances of bonding with differing options is to load
1390 the bonding driver multiple times. Note that current versions of the
1391 sysconfig network initialization scripts handle this automatically; if
1392 your distro uses these scripts, no special action is needed. See the
1393 section Configuring Bonding Devices, above, if you're not sure about your
1394 network initialization scripts.
1396 To load multiple instances of the module, it is necessary to
1397 specify a different name for each instance (the module loading system
1398 requires that every loaded module, even multiple instances of the same
1399 module, have a unique name). This is accomplished by supplying multiple
1400 sets of bonding options in ``/etc/modprobe.d/*.conf``, for example::
1403 options bond0 -o bond0 mode=balance-rr miimon=100
1406 options bond1 -o bond1 mode=balance-alb miimon=50
1408 will load the bonding module two times. The first instance is
1409 named "bond0" and creates the bond0 device in balance-rr mode with an
1410 miimon of 100. The second instance is named "bond1" and creates the
1411 bond1 device in balance-alb mode with an miimon of 50.
1413 In some circumstances (typically with older distributions),
1414 the above does not work, and the second bonding instance never sees
1415 its options. In that case, the second options line can be substituted
1418 install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1419 mode=balance-alb miimon=50
1421 This may be repeated any number of times, specifying a new and
1422 unique name in place of bond1 for each subsequent instance.
1424 It has been observed that some Red Hat supplied kernels are unable
1425 to rename modules at load time (the "-o bond1" part). Attempts to pass
1426 that option to modprobe will produce an "Operation not permitted" error.
1427 This has been reported on some Fedora Core kernels, and has been seen on
1428 RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1429 to configure multiple bonds with differing parameters (as they are older
1430 kernels, and also lack sysfs support).
1432 3.4 Configuring Bonding Manually via Sysfs
1433 ------------------------------------------
1435 Starting with version 3.0.0, Channel Bonding may be configured
1436 via the sysfs interface. This interface allows dynamic configuration
1437 of all bonds in the system without unloading the module. It also
1438 allows for adding and removing bonds at runtime. Ifenslave is no
1439 longer required, though it is still supported.
1441 Use of the sysfs interface allows you to use multiple bonds
1442 with different configurations without having to reload the module.
1443 It also allows you to use multiple, differently configured bonds when
1444 bonding is compiled into the kernel.
1446 You must have the sysfs filesystem mounted to configure
1447 bonding this way. The examples in this document assume that you
1448 are using the standard mount point for sysfs, e.g. /sys. If your
1449 sysfs filesystem is mounted elsewhere, you will need to adjust the
1450 example paths accordingly.
1452 Creating and Destroying Bonds
1453 -----------------------------
1454 To add a new bond foo::
1456 # echo +foo > /sys/class/net/bonding_masters
1458 To remove an existing bond bar::
1460 # echo -bar > /sys/class/net/bonding_masters
1462 To show all existing bonds::
1464 # cat /sys/class/net/bonding_masters
1468 due to 4K size limitation of sysfs files, this list may be
1469 truncated if you have more than a few hundred bonds. This is unlikely
1470 to occur under normal operating conditions.
1472 Adding and Removing Slaves
1473 --------------------------
1474 Interfaces may be enslaved to a bond using the file
1475 /sys/class/net/<bond>/bonding/slaves. The semantics for this file
1476 are the same as for the bonding_masters file.
1478 To enslave interface eth0 to bond bond0::
1481 # echo +eth0 > /sys/class/net/bond0/bonding/slaves
1483 To free slave eth0 from bond bond0::
1485 # echo -eth0 > /sys/class/net/bond0/bonding/slaves
1487 When an interface is enslaved to a bond, symlinks between the
1488 two are created in the sysfs filesystem. In this case, you would get
1489 /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1490 /sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1492 This means that you can tell quickly whether or not an
1493 interface is enslaved by looking for the master symlink. Thus:
1494 # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1495 will free eth0 from whatever bond it is enslaved to, regardless of
1496 the name of the bond interface.
1498 Changing a Bond's Configuration
1499 -------------------------------
1500 Each bond may be configured individually by manipulating the
1501 files located in /sys/class/net/<bond name>/bonding
1503 The names of these files correspond directly with the command-
1504 line parameters described elsewhere in this file, and, with the
1505 exception of arp_ip_target, they accept the same values. To see the
1506 current setting, simply cat the appropriate file.
1508 A few examples will be given here; for specific usage
1509 guidelines for each parameter, see the appropriate section in this
1512 To configure bond0 for balance-alb mode::
1514 # ifconfig bond0 down
1515 # echo 6 > /sys/class/net/bond0/bonding/mode
1517 # echo balance-alb > /sys/class/net/bond0/bonding/mode
1521 The bond interface must be down before the mode can be changed.
1523 To enable MII monitoring on bond0 with a 1 second interval::
1525 # echo 1000 > /sys/class/net/bond0/bonding/miimon
1529 If ARP monitoring is enabled, it will disabled when MII
1530 monitoring is enabled, and vice-versa.
1532 To add ARP targets::
1534 # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1535 # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1539 up to 16 target addresses may be specified.
1541 To remove an ARP target::
1543 # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1545 To configure the interval between learning packet transmits::
1547 # echo 12 > /sys/class/net/bond0/bonding/lp_interval
1551 the lp_interval is the number of seconds between instances where
1552 the bonding driver sends learning packets to each slaves peer switch. The
1553 default interval is 1 second.
1555 Example Configuration
1556 ---------------------
1557 We begin with the same example that is shown in section 3.3,
1558 executed with sysfs, and without using ifenslave.
1560 To make a simple bond of two e100 devices (presumed to be eth0
1561 and eth1), and have it persist across reboots, edit the appropriate
1562 file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1567 echo balance-alb > /sys/class/net/bond0/bonding/mode
1568 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1569 echo 100 > /sys/class/net/bond0/bonding/miimon
1570 echo +eth0 > /sys/class/net/bond0/bonding/slaves
1571 echo +eth1 > /sys/class/net/bond0/bonding/slaves
1573 To add a second bond, with two e1000 interfaces in
1574 active-backup mode, using ARP monitoring, add the following lines to
1578 echo +bond1 > /sys/class/net/bonding_masters
1579 echo active-backup > /sys/class/net/bond1/bonding/mode
1580 ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1581 echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1582 echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1583 echo +eth2 > /sys/class/net/bond1/bonding/slaves
1584 echo +eth3 > /sys/class/net/bond1/bonding/slaves
1586 3.5 Configuration with Interfaces Support
1587 -----------------------------------------
1589 This section applies to distros which use /etc/network/interfaces file
1590 to describe network interface configuration, most notably Debian and it's
1593 The ifup and ifdown commands on Debian don't support bonding out of
1594 the box. The ifenslave-2.6 package should be installed to provide bonding
1595 support. Once installed, this package will provide ``bond-*`` options
1596 to be used into /etc/network/interfaces.
1598 Note that ifenslave-2.6 package will load the bonding module and use
1599 the ifenslave command when appropriate.
1601 Example Configurations
1602 ----------------------
1604 In /etc/network/interfaces, the following stanza will configure bond0, in
1605 active-backup mode, with eth0 and eth1 as slaves::
1608 iface bond0 inet dhcp
1609 bond-slaves eth0 eth1
1610 bond-mode active-backup
1612 bond-primary eth0 eth1
1614 If the above configuration doesn't work, you might have a system using
1615 upstart for system startup. This is most notably true for recent
1616 Ubuntu versions. The following stanza in /etc/network/interfaces will
1617 produce the same result on those systems::
1620 iface bond0 inet dhcp
1622 bond-mode active-backup
1626 iface eth0 inet manual
1628 bond-primary eth0 eth1
1631 iface eth1 inet manual
1633 bond-primary eth0 eth1
1635 For a full list of ``bond-*`` supported options in /etc/network/interfaces and
1636 some more advanced examples tailored to you particular distros, see the files in
1637 /usr/share/doc/ifenslave-2.6.
1639 3.6 Overriding Configuration for Special Cases
1640 ----------------------------------------------
1642 When using the bonding driver, the physical port which transmits a frame is
1643 typically selected by the bonding driver, and is not relevant to the user or
1644 system administrator. The output port is simply selected using the policies of
1645 the selected bonding mode. On occasion however, it is helpful to direct certain
1646 classes of traffic to certain physical interfaces on output to implement
1647 slightly more complex policies. For example, to reach a web server over a
1648 bonded interface in which eth0 connects to a private network, while eth1
1649 connects via a public network, it may be desirous to bias the bond to send said
1650 traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1651 can safely be sent over either interface. Such configurations may be achieved
1652 using the traffic control utilities inherent in linux.
1654 By default the bonding driver is multiqueue aware and 16 queues are created
1655 when the driver initializes (see Documentation/networking/multiqueue.rst
1656 for details). If more or less queues are desired the module parameter
1657 tx_queues can be used to change this value. There is no sysfs parameter
1658 available as the allocation is done at module init time.
1660 The output of the file /proc/net/bonding/bondX has changed so the output Queue
1661 ID is now printed for each slave::
1663 Bonding Mode: fault-tolerance (active-backup)
1665 Currently Active Slave: eth0
1667 MII Polling Interval (ms): 0
1671 Slave Interface: eth0
1673 Link Failure Count: 0
1674 Permanent HW addr: 00:1a:a0:12:8f:cb
1677 Slave Interface: eth1
1679 Link Failure Count: 0
1680 Permanent HW addr: 00:1a:a0:12:8f:cc
1683 The queue_id for a slave can be set using the command::
1685 # echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1687 Any interface that needs a queue_id set should set it with multiple calls
1688 like the one above until proper priorities are set for all interfaces. On
1689 distributions that allow configuration via initscripts, multiple 'queue_id'
1690 arguments can be added to BONDING_OPTS to set all needed slave queues.
1692 These queue id's can be used in conjunction with the tc utility to configure
1693 a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1694 slave devices. For instance, say we wanted, in the above configuration to
1695 force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1696 device. The following commands would accomplish this::
1698 # tc qdisc add dev bond0 handle 1 root multiq
1700 # tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip \
1701 dst 192.168.1.100 action skbedit queue_mapping 2
1703 These commands tell the kernel to attach a multiqueue queue discipline to the
1704 bond0 interface and filter traffic enqueued to it, such that packets with a dst
1705 ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1706 This value is then passed into the driver, causing the normal output path
1707 selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1709 Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver
1710 that normal output policy selection should take place. One benefit to simply
1711 leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1712 driver that is now present. This awareness allows tc filters to be placed on
1713 slave devices as well as bond devices and the bonding driver will simply act as
1714 a pass-through for selecting output queues on the slave device rather than
1715 output port selection.
1717 This feature first appeared in bonding driver version 3.7.0 and support for
1718 output slave selection was limited to round-robin and active-backup modes.
1720 3.7 Configuring LACP for 802.3ad mode in a more secure way
1721 ----------------------------------------------------------
1723 When using 802.3ad bonding mode, the Actor (host) and Partner (switch)
1724 exchange LACPDUs. These LACPDUs cannot be sniffed, because they are
1725 destined to link local mac addresses (which switches/bridges are not
1726 supposed to forward). However, most of the values are easily predictable
1727 or are simply the machine's MAC address (which is trivially known to all
1728 other hosts in the same L2). This implies that other machines in the L2
1729 domain can spoof LACPDU packets from other hosts to the switch and potentially
1730 cause mayhem by joining (from the point of view of the switch) another
1731 machine's aggregate, thus receiving a portion of that hosts incoming
1732 traffic and / or spoofing traffic from that machine themselves (potentially
1733 even successfully terminating some portion of flows). Though this is not
1734 a likely scenario, one could avoid this possibility by simply configuring
1735 few bonding parameters:
1737 (a) ad_actor_system : You can set a random mac-address that can be used for
1738 these LACPDU exchanges. The value can not be either NULL or Multicast.
1739 Also it's preferable to set the local-admin bit. Following shell code
1740 generates a random mac-address as described above::
1742 # sys_mac_addr=$(printf '%02x:%02x:%02x:%02x:%02x:%02x' \
1743 $(( (RANDOM & 0xFE) | 0x02 )) \
1744 $(( RANDOM & 0xFF )) \
1745 $(( RANDOM & 0xFF )) \
1746 $(( RANDOM & 0xFF )) \
1747 $(( RANDOM & 0xFF )) \
1748 $(( RANDOM & 0xFF )))
1749 # echo $sys_mac_addr > /sys/class/net/bond0/bonding/ad_actor_system
1751 (b) ad_actor_sys_prio : Randomize the system priority. The default value
1752 is 65535, but system can take the value from 1 - 65535. Following shell
1753 code generates random priority and sets it::
1755 # sys_prio=$(( 1 + RANDOM + RANDOM ))
1756 # echo $sys_prio > /sys/class/net/bond0/bonding/ad_actor_sys_prio
1758 (c) ad_user_port_key : Use the user portion of the port-key. The default
1759 keeps this empty. These are the upper 10 bits of the port-key and value
1760 ranges from 0 - 1023. Following shell code generates these 10 bits and
1763 # usr_port_key=$(( RANDOM & 0x3FF ))
1764 # echo $usr_port_key > /sys/class/net/bond0/bonding/ad_user_port_key
1767 4 Querying Bonding Configuration
1768 =================================
1770 4.1 Bonding Configuration
1771 -------------------------
1773 Each bonding device has a read-only file residing in the
1774 /proc/net/bonding directory. The file contents include information
1775 about the bonding configuration, options and state of each slave.
1777 For example, the contents of /proc/net/bonding/bond0 after the
1778 driver is loaded with parameters of mode=0 and miimon=1000 is
1779 generally as follows::
1781 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1782 Bonding Mode: load balancing (round-robin)
1783 Currently Active Slave: eth0
1785 MII Polling Interval (ms): 1000
1789 Slave Interface: eth1
1791 Link Failure Count: 1
1793 Slave Interface: eth0
1795 Link Failure Count: 1
1797 The precise format and contents will change depending upon the
1798 bonding configuration, state, and version of the bonding driver.
1800 4.2 Network configuration
1801 -------------------------
1803 The network configuration can be inspected using the ifconfig
1804 command. Bonding devices will have the MASTER flag set; Bonding slave
1805 devices will have the SLAVE flag set. The ifconfig output does not
1806 contain information on which slaves are associated with which masters.
1808 In the example below, the bond0 interface is the master
1809 (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1810 bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1811 TLB and ALB that require a unique MAC address for each slave::
1814 bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1815 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1816 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1817 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1818 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1819 collisions:0 txqueuelen:0
1821 eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1822 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1823 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1824 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1825 collisions:0 txqueuelen:100
1826 Interrupt:10 Base address:0x1080
1828 eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1829 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1830 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1831 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1832 collisions:0 txqueuelen:100
1833 Interrupt:9 Base address:0x1400
1835 5. Switch Configuration
1836 =======================
1838 For this section, "switch" refers to whatever system the
1839 bonded devices are directly connected to (i.e., where the other end of
1840 the cable plugs into). This may be an actual dedicated switch device,
1841 or it may be another regular system (e.g., another computer running
1844 The active-backup, balance-tlb and balance-alb modes do not
1845 require any specific configuration of the switch.
1847 The 802.3ad mode requires that the switch have the appropriate
1848 ports configured as an 802.3ad aggregation. The precise method used
1849 to configure this varies from switch to switch, but, for example, a
1850 Cisco 3550 series switch requires that the appropriate ports first be
1851 grouped together in a single etherchannel instance, then that
1852 etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1853 standard EtherChannel).
1855 The balance-rr, balance-xor and broadcast modes generally
1856 require that the switch have the appropriate ports grouped together.
1857 The nomenclature for such a group differs between switches, it may be
1858 called an "etherchannel" (as in the Cisco example, above), a "trunk
1859 group" or some other similar variation. For these modes, each switch
1860 will also have its own configuration options for the switch's transmit
1861 policy to the bond. Typical choices include XOR of either the MAC or
1862 IP addresses. The transmit policy of the two peers does not need to
1863 match. For these three modes, the bonding mode really selects a
1864 transmit policy for an EtherChannel group; all three will interoperate
1865 with another EtherChannel group.
1868 6. 802.1q VLAN Support
1869 ======================
1871 It is possible to configure VLAN devices over a bond interface
1872 using the 8021q driver. However, only packets coming from the 8021q
1873 driver and passing through bonding will be tagged by default. Self
1874 generated packets, for example, bonding's learning packets or ARP
1875 packets generated by either ALB mode or the ARP monitor mechanism, are
1876 tagged internally by bonding itself. As a result, bonding must
1877 "learn" the VLAN IDs configured above it, and use those IDs to tag
1878 self generated packets.
1880 For reasons of simplicity, and to support the use of adapters
1881 that can do VLAN hardware acceleration offloading, the bonding
1882 interface declares itself as fully hardware offloading capable, it gets
1883 the add_vid/kill_vid notifications to gather the necessary
1884 information, and it propagates those actions to the slaves. In case
1885 of mixed adapter types, hardware accelerated tagged packets that
1886 should go through an adapter that is not offloading capable are
1887 "un-accelerated" by the bonding driver so the VLAN tag sits in the
1890 VLAN interfaces *must* be added on top of a bonding interface
1891 only after enslaving at least one slave. The bonding interface has a
1892 hardware address of 00:00:00:00:00:00 until the first slave is added.
1893 If the VLAN interface is created prior to the first enslavement, it
1894 would pick up the all-zeroes hardware address. Once the first slave
1895 is attached to the bond, the bond device itself will pick up the
1896 slave's hardware address, which is then available for the VLAN device.
1898 Also, be aware that a similar problem can occur if all slaves
1899 are released from a bond that still has one or more VLAN interfaces on
1900 top of it. When a new slave is added, the bonding interface will
1901 obtain its hardware address from the first slave, which might not
1902 match the hardware address of the VLAN interfaces (which was
1903 ultimately copied from an earlier slave).
1905 There are two methods to insure that the VLAN device operates
1906 with the correct hardware address if all slaves are removed from a
1909 1. Remove all VLAN interfaces then recreate them
1911 2. Set the bonding interface's hardware address so that it
1912 matches the hardware address of the VLAN interfaces.
1914 Note that changing a VLAN interface's HW address would set the
1915 underlying device -- i.e. the bonding interface -- to promiscuous
1916 mode, which might not be what you want.
1922 The bonding driver at present supports two schemes for
1923 monitoring a slave device's link state: the ARP monitor and the MII
1926 At the present time, due to implementation restrictions in the
1927 bonding driver itself, it is not possible to enable both ARP and MII
1928 monitoring simultaneously.
1930 7.1 ARP Monitor Operation
1931 -------------------------
1933 The ARP monitor operates as its name suggests: it sends ARP
1934 queries to one or more designated peer systems on the network, and
1935 uses the response as an indication that the link is operating. This
1936 gives some assurance that traffic is actually flowing to and from one
1937 or more peers on the local network.
1939 The ARP monitor relies on the device driver itself to verify
1940 that traffic is flowing. In particular, the driver must keep up to
1941 date the last receive time, dev->last_rx. Drivers that use NETIF_F_LLTX
1942 flag must also update netdev_queue->trans_start. If they do not, then the
1943 ARP monitor will immediately fail any slaves using that driver, and
1944 those slaves will stay down. If networking monitoring (tcpdump, etc)
1945 shows the ARP requests and replies on the network, then it may be that
1946 your device driver is not updating last_rx and trans_start.
1948 7.2 Configuring Multiple ARP Targets
1949 ------------------------------------
1951 While ARP monitoring can be done with just one target, it can
1952 be useful in a High Availability setup to have several targets to
1953 monitor. In the case of just one target, the target itself may go
1954 down or have a problem making it unresponsive to ARP requests. Having
1955 an additional target (or several) increases the reliability of the ARP
1958 Multiple ARP targets must be separated by commas as follows::
1960 # example options for ARP monitoring with three targets
1962 options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1964 For just a single target the options would resemble::
1966 # example options for ARP monitoring with one target
1968 options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1971 7.3 MII Monitor Operation
1972 -------------------------
1974 The MII monitor monitors only the carrier state of the local
1975 network interface. It accomplishes this in one of three ways: by
1976 depending upon the device driver to maintain its carrier state, by
1977 querying the device's MII registers, or by making an ethtool query to
1980 If the use_carrier module parameter is 1 (the default value),
1981 then the MII monitor will rely on the driver for carrier state
1982 information (via the netif_carrier subsystem). As explained in the
1983 use_carrier parameter information, above, if the MII monitor fails to
1984 detect carrier loss on the device (e.g., when the cable is physically
1985 disconnected), it may be that the driver does not support
1988 If use_carrier is 0, then the MII monitor will first query the
1989 device's (via ioctl) MII registers and check the link state. If that
1990 request fails (not just that it returns carrier down), then the MII
1991 monitor will make an ethtool ETHTOOL_GLINK request to attempt to obtain
1992 the same information. If both methods fail (i.e., the driver either
1993 does not support or had some error in processing both the MII register
1994 and ethtool requests), then the MII monitor will assume the link is
1997 8. Potential Sources of Trouble
1998 ===============================
2000 8.1 Adventures in Routing
2001 -------------------------
2003 When bonding is configured, it is important that the slave
2004 devices not have routes that supersede routes of the master (or,
2005 generally, not have routes at all). For example, suppose the bonding
2006 device bond0 has two slaves, eth0 and eth1, and the routing table is
2009 Kernel IP routing table
2010 Destination Gateway Genmask Flags MSS Window irtt Iface
2011 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
2012 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
2013 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
2014 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
2016 This routing configuration will likely still update the
2017 receive/transmit times in the driver (needed by the ARP monitor), but
2018 may bypass the bonding driver (because outgoing traffic to, in this
2019 case, another host on network 10 would use eth0 or eth1 before bond0).
2021 The ARP monitor (and ARP itself) may become confused by this
2022 configuration, because ARP requests (generated by the ARP monitor)
2023 will be sent on one interface (bond0), but the corresponding reply
2024 will arrive on a different interface (eth0). This reply looks to ARP
2025 as an unsolicited ARP reply (because ARP matches replies on an
2026 interface basis), and is discarded. The MII monitor is not affected
2027 by the state of the routing table.
2029 The solution here is simply to insure that slaves do not have
2030 routes of their own, and if for some reason they must, those routes do
2031 not supersede routes of their master. This should generally be the
2032 case, but unusual configurations or errant manual or automatic static
2033 route additions may cause trouble.
2035 8.2 Ethernet Device Renaming
2036 ----------------------------
2038 On systems with network configuration scripts that do not
2039 associate physical devices directly with network interface names (so
2040 that the same physical device always has the same "ethX" name), it may
2041 be necessary to add some special logic to config files in
2044 For example, given a modules.conf containing the following::
2047 options bond0 mode=some-mode miimon=50
2053 If neither eth0 and eth1 are slaves to bond0, then when the
2054 bond0 interface comes up, the devices may end up reordered. This
2055 happens because bonding is loaded first, then its slave device's
2056 drivers are loaded next. Since no other drivers have been loaded,
2057 when the e1000 driver loads, it will receive eth0 and eth1 for its
2058 devices, but the bonding configuration tries to enslave eth2 and eth3
2059 (which may later be assigned to the tg3 devices).
2061 Adding the following::
2063 add above bonding e1000 tg3
2065 causes modprobe to load e1000 then tg3, in that order, when
2066 bonding is loaded. This command is fully documented in the
2067 modules.conf manual page.
2069 On systems utilizing modprobe an equivalent problem can occur.
2070 In this case, the following can be added to config files in
2071 /etc/modprobe.d/ as::
2073 softdep bonding pre: tg3 e1000
2075 This will load tg3 and e1000 modules before loading the bonding one.
2076 Full documentation on this can be found in the modprobe.d and modprobe
2079 8.3. Painfully Slow Or No Failed Link Detection By Miimon
2080 ---------------------------------------------------------
2082 By default, bonding enables the use_carrier option, which
2083 instructs bonding to trust the driver to maintain carrier state.
2085 As discussed in the options section, above, some drivers do
2086 not support the netif_carrier_on/_off link state tracking system.
2087 With use_carrier enabled, bonding will always see these links as up,
2088 regardless of their actual state.
2090 Additionally, other drivers do support netif_carrier, but do
2091 not maintain it in real time, e.g., only polling the link state at
2092 some fixed interval. In this case, miimon will detect failures, but
2093 only after some long period of time has expired. If it appears that
2094 miimon is very slow in detecting link failures, try specifying
2095 use_carrier=0 to see if that improves the failure detection time. If
2096 it does, then it may be that the driver checks the carrier state at a
2097 fixed interval, but does not cache the MII register values (so the
2098 use_carrier=0 method of querying the registers directly works). If
2099 use_carrier=0 does not improve the failover, then the driver may cache
2100 the registers, or the problem may be elsewhere.
2102 Also, remember that miimon only checks for the device's
2103 carrier state. It has no way to determine the state of devices on or
2104 beyond other ports of a switch, or if a switch is refusing to pass
2105 traffic while still maintaining carrier on.
2110 If running SNMP agents, the bonding driver should be loaded
2111 before any network drivers participating in a bond. This requirement
2112 is due to the interface index (ipAdEntIfIndex) being associated to
2113 the first interface found with a given IP address. That is, there is
2114 only one ipAdEntIfIndex for each IP address. For example, if eth0 and
2115 eth1 are slaves of bond0 and the driver for eth0 is loaded before the
2116 bonding driver, the interface for the IP address will be associated
2117 with the eth0 interface. This configuration is shown below, the IP
2118 address 192.168.1.1 has an interface index of 2 which indexes to eth0
2119 in the ifDescr table (ifDescr.2).
2123 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2124 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
2125 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
2126 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
2127 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
2128 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
2129 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
2130 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2131 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
2132 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2134 This problem is avoided by loading the bonding driver before
2135 any network drivers participating in a bond. Below is an example of
2136 loading the bonding driver first, the IP address 192.168.1.1 is
2137 correctly associated with ifDescr.2.
2139 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2140 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
2141 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
2142 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
2143 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
2144 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
2145 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
2146 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2147 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
2148 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2150 While some distributions may not report the interface name in
2151 ifDescr, the association between the IP address and IfIndex remains
2152 and SNMP functions such as Interface_Scan_Next will report that
2155 10. Promiscuous mode
2156 ====================
2158 When running network monitoring tools, e.g., tcpdump, it is
2159 common to enable promiscuous mode on the device, so that all traffic
2160 is seen (instead of seeing only traffic destined for the local host).
2161 The bonding driver handles promiscuous mode changes to the bonding
2162 master device (e.g., bond0), and propagates the setting to the slave
2165 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
2166 the promiscuous mode setting is propagated to all slaves.
2168 For the active-backup, balance-tlb and balance-alb modes, the
2169 promiscuous mode setting is propagated only to the active slave.
2171 For balance-tlb mode, the active slave is the slave currently
2172 receiving inbound traffic.
2174 For balance-alb mode, the active slave is the slave used as a
2175 "primary." This slave is used for mode-specific control traffic, for
2176 sending to peers that are unassigned or if the load is unbalanced.
2178 For the active-backup, balance-tlb and balance-alb modes, when
2179 the active slave changes (e.g., due to a link failure), the
2180 promiscuous setting will be propagated to the new active slave.
2182 11. Configuring Bonding for High Availability
2183 =============================================
2185 High Availability refers to configurations that provide
2186 maximum network availability by having redundant or backup devices,
2187 links or switches between the host and the rest of the world. The
2188 goal is to provide the maximum availability of network connectivity
2189 (i.e., the network always works), even though other configurations
2190 could provide higher throughput.
2192 11.1 High Availability in a Single Switch Topology
2193 --------------------------------------------------
2195 If two hosts (or a host and a single switch) are directly
2196 connected via multiple physical links, then there is no availability
2197 penalty to optimizing for maximum bandwidth. In this case, there is
2198 only one switch (or peer), so if it fails, there is no alternative
2199 access to fail over to. Additionally, the bonding load balance modes
2200 support link monitoring of their members, so if individual links fail,
2201 the load will be rebalanced across the remaining devices.
2203 See Section 12, "Configuring Bonding for Maximum Throughput"
2204 for information on configuring bonding with one peer device.
2206 11.2 High Availability in a Multiple Switch Topology
2207 ----------------------------------------------------
2209 With multiple switches, the configuration of bonding and the
2210 network changes dramatically. In multiple switch topologies, there is
2211 a trade off between network availability and usable bandwidth.
2213 Below is a sample network, configured to maximize the
2214 availability of the network::
2218 +-----+----+ +-----+----+
2219 | |port2 ISL port2| |
2220 | switch A +--------------------------+ switch B |
2222 +-----+----+ +-----++---+
2225 +-------------+ host1 +---------------+
2228 In this configuration, there is a link between the two
2229 switches (ISL, or inter switch link), and multiple ports connecting to
2230 the outside world ("port3" on each switch). There is no technical
2231 reason that this could not be extended to a third switch.
2233 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
2234 -------------------------------------------------------------
2236 In a topology such as the example above, the active-backup and
2237 broadcast modes are the only useful bonding modes when optimizing for
2238 availability; the other modes require all links to terminate on the
2239 same peer for them to behave rationally.
2242 This is generally the preferred mode, particularly if
2243 the switches have an ISL and play together well. If the
2244 network configuration is such that one switch is specifically
2245 a backup switch (e.g., has lower capacity, higher cost, etc),
2246 then the primary option can be used to insure that the
2247 preferred link is always used when it is available.
2250 This mode is really a special purpose mode, and is suitable
2251 only for very specific needs. For example, if the two
2252 switches are not connected (no ISL), and the networks beyond
2253 them are totally independent. In this case, if it is
2254 necessary for some specific one-way traffic to reach both
2255 independent networks, then the broadcast mode may be suitable.
2257 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
2258 ----------------------------------------------------------------
2260 The choice of link monitoring ultimately depends upon your
2261 switch. If the switch can reliably fail ports in response to other
2262 failures, then either the MII or ARP monitors should work. For
2263 example, in the above example, if the "port3" link fails at the remote
2264 end, the MII monitor has no direct means to detect this. The ARP
2265 monitor could be configured with a target at the remote end of port3,
2266 thus detecting that failure without switch support.
2268 In general, however, in a multiple switch topology, the ARP
2269 monitor can provide a higher level of reliability in detecting end to
2270 end connectivity failures (which may be caused by the failure of any
2271 individual component to pass traffic for any reason). Additionally,
2272 the ARP monitor should be configured with multiple targets (at least
2273 one for each switch in the network). This will insure that,
2274 regardless of which switch is active, the ARP monitor has a suitable
2277 Note, also, that of late many switches now support a functionality
2278 generally referred to as "trunk failover." This is a feature of the
2279 switch that causes the link state of a particular switch port to be set
2280 down (or up) when the state of another switch port goes down (or up).
2281 Its purpose is to propagate link failures from logically "exterior" ports
2282 to the logically "interior" ports that bonding is able to monitor via
2283 miimon. Availability and configuration for trunk failover varies by
2284 switch, but this can be a viable alternative to the ARP monitor when using
2287 12. Configuring Bonding for Maximum Throughput
2288 ==============================================
2290 12.1 Maximizing Throughput in a Single Switch Topology
2291 ------------------------------------------------------
2293 In a single switch configuration, the best method to maximize
2294 throughput depends upon the application and network environment. The
2295 various load balancing modes each have strengths and weaknesses in
2296 different environments, as detailed below.
2298 For this discussion, we will break down the topologies into
2299 two categories. Depending upon the destination of most traffic, we
2300 categorize them into either "gatewayed" or "local" configurations.
2302 In a gatewayed configuration, the "switch" is acting primarily
2303 as a router, and the majority of traffic passes through this router to
2304 other networks. An example would be the following::
2307 +----------+ +----------+
2308 | |eth0 port1| | to other networks
2309 | Host A +---------------------+ router +------------------->
2310 | +---------------------+ | Hosts B and C are out
2311 | |eth1 port2| | here somewhere
2312 +----------+ +----------+
2314 The router may be a dedicated router device, or another host
2315 acting as a gateway. For our discussion, the important point is that
2316 the majority of traffic from Host A will pass through the router to
2317 some other network before reaching its final destination.
2319 In a gatewayed network configuration, although Host A may
2320 communicate with many other systems, all of its traffic will be sent
2321 and received via one other peer on the local network, the router.
2323 Note that the case of two systems connected directly via
2324 multiple physical links is, for purposes of configuring bonding, the
2325 same as a gatewayed configuration. In that case, it happens that all
2326 traffic is destined for the "gateway" itself, not some other network
2329 In a local configuration, the "switch" is acting primarily as
2330 a switch, and the majority of traffic passes through this switch to
2331 reach other stations on the same network. An example would be the
2334 +----------+ +----------+ +--------+
2335 | |eth0 port1| +-------+ Host B |
2336 | Host A +------------+ switch |port3 +--------+
2337 | +------------+ | +--------+
2338 | |eth1 port2| +------------------+ Host C |
2339 +----------+ +----------+port4 +--------+
2342 Again, the switch may be a dedicated switch device, or another
2343 host acting as a gateway. For our discussion, the important point is
2344 that the majority of traffic from Host A is destined for other hosts
2345 on the same local network (Hosts B and C in the above example).
2347 In summary, in a gatewayed configuration, traffic to and from
2348 the bonded device will be to the same MAC level peer on the network
2349 (the gateway itself, i.e., the router), regardless of its final
2350 destination. In a local configuration, traffic flows directly to and
2351 from the final destinations, thus, each destination (Host B, Host C)
2352 will be addressed directly by their individual MAC addresses.
2354 This distinction between a gatewayed and a local network
2355 configuration is important because many of the load balancing modes
2356 available use the MAC addresses of the local network source and
2357 destination to make load balancing decisions. The behavior of each
2358 mode is described below.
2361 12.1.1 MT Bonding Mode Selection for Single Switch Topology
2362 -----------------------------------------------------------
2364 This configuration is the easiest to set up and to understand,
2365 although you will have to decide which bonding mode best suits your
2366 needs. The trade offs for each mode are detailed below:
2369 This mode is the only mode that will permit a single
2370 TCP/IP connection to stripe traffic across multiple
2371 interfaces. It is therefore the only mode that will allow a
2372 single TCP/IP stream to utilize more than one interface's
2373 worth of throughput. This comes at a cost, however: the
2374 striping generally results in peer systems receiving packets out
2375 of order, causing TCP/IP's congestion control system to kick
2376 in, often by retransmitting segments.
2378 It is possible to adjust TCP/IP's congestion limits by
2379 altering the net.ipv4.tcp_reordering sysctl parameter. The
2380 usual default value is 3. But keep in mind TCP stack is able
2381 to automatically increase this when it detects reorders.
2383 Note that the fraction of packets that will be delivered out of
2384 order is highly variable, and is unlikely to be zero. The level
2385 of reordering depends upon a variety of factors, including the
2386 networking interfaces, the switch, and the topology of the
2387 configuration. Speaking in general terms, higher speed network
2388 cards produce more reordering (due to factors such as packet
2389 coalescing), and a "many to many" topology will reorder at a
2390 higher rate than a "many slow to one fast" configuration.
2392 Many switches do not support any modes that stripe traffic
2393 (instead choosing a port based upon IP or MAC level addresses);
2394 for those devices, traffic for a particular connection flowing
2395 through the switch to a balance-rr bond will not utilize greater
2396 than one interface's worth of bandwidth.
2398 If you are utilizing protocols other than TCP/IP, UDP for
2399 example, and your application can tolerate out of order
2400 delivery, then this mode can allow for single stream datagram
2401 performance that scales near linearly as interfaces are added
2404 This mode requires the switch to have the appropriate ports
2405 configured for "etherchannel" or "trunking."
2408 There is not much advantage in this network topology to
2409 the active-backup mode, as the inactive backup devices are all
2410 connected to the same peer as the primary. In this case, a
2411 load balancing mode (with link monitoring) will provide the
2412 same level of network availability, but with increased
2413 available bandwidth. On the plus side, active-backup mode
2414 does not require any configuration of the switch, so it may
2415 have value if the hardware available does not support any of
2416 the load balance modes.
2419 This mode will limit traffic such that packets destined
2420 for specific peers will always be sent over the same
2421 interface. Since the destination is determined by the MAC
2422 addresses involved, this mode works best in a "local" network
2423 configuration (as described above), with destinations all on
2424 the same local network. This mode is likely to be suboptimal
2425 if all your traffic is passed through a single router (i.e., a
2426 "gatewayed" network configuration, as described above).
2428 As with balance-rr, the switch ports need to be configured for
2429 "etherchannel" or "trunking."
2432 Like active-backup, there is not much advantage to this
2433 mode in this type of network topology.
2436 This mode can be a good choice for this type of network
2437 topology. The 802.3ad mode is an IEEE standard, so all peers
2438 that implement 802.3ad should interoperate well. The 802.3ad
2439 protocol includes automatic configuration of the aggregates,
2440 so minimal manual configuration of the switch is needed
2441 (typically only to designate that some set of devices is
2442 available for 802.3ad). The 802.3ad standard also mandates
2443 that frames be delivered in order (within certain limits), so
2444 in general single connections will not see misordering of
2445 packets. The 802.3ad mode does have some drawbacks: the
2446 standard mandates that all devices in the aggregate operate at
2447 the same speed and duplex. Also, as with all bonding load
2448 balance modes other than balance-rr, no single connection will
2449 be able to utilize more than a single interface's worth of
2452 Additionally, the linux bonding 802.3ad implementation
2453 distributes traffic by peer (using an XOR of MAC addresses
2454 and packet type ID), so in a "gatewayed" configuration, all
2455 outgoing traffic will generally use the same device. Incoming
2456 traffic may also end up on a single device, but that is
2457 dependent upon the balancing policy of the peer's 802.3ad
2458 implementation. In a "local" configuration, traffic will be
2459 distributed across the devices in the bond.
2461 Finally, the 802.3ad mode mandates the use of the MII monitor,
2462 therefore, the ARP monitor is not available in this mode.
2465 The balance-tlb mode balances outgoing traffic by peer.
2466 Since the balancing is done according to MAC address, in a
2467 "gatewayed" configuration (as described above), this mode will
2468 send all traffic across a single device. However, in a
2469 "local" network configuration, this mode balances multiple
2470 local network peers across devices in a vaguely intelligent
2471 manner (not a simple XOR as in balance-xor or 802.3ad mode),
2472 so that mathematically unlucky MAC addresses (i.e., ones that
2473 XOR to the same value) will not all "bunch up" on a single
2476 Unlike 802.3ad, interfaces may be of differing speeds, and no
2477 special switch configuration is required. On the down side,
2478 in this mode all incoming traffic arrives over a single
2479 interface, this mode requires certain ethtool support in the
2480 network device driver of the slave interfaces, and the ARP
2481 monitor is not available.
2484 This mode is everything that balance-tlb is, and more.
2485 It has all of the features (and restrictions) of balance-tlb,
2486 and will also balance incoming traffic from local network
2487 peers (as described in the Bonding Module Options section,
2490 The only additional down side to this mode is that the network
2491 device driver must support changing the hardware address while
2494 12.1.2 MT Link Monitoring for Single Switch Topology
2495 ----------------------------------------------------
2497 The choice of link monitoring may largely depend upon which
2498 mode you choose to use. The more advanced load balancing modes do not
2499 support the use of the ARP monitor, and are thus restricted to using
2500 the MII monitor (which does not provide as high a level of end to end
2501 assurance as the ARP monitor).
2503 12.2 Maximum Throughput in a Multiple Switch Topology
2504 -----------------------------------------------------
2506 Multiple switches may be utilized to optimize for throughput
2507 when they are configured in parallel as part of an isolated network
2508 between two or more systems, for example::
2514 +--------+ | +---------+
2516 +------+---+ +-----+----+ +-----+----+
2517 | Switch A | | Switch B | | Switch C |
2518 +------+---+ +-----+----+ +-----+----+
2520 +--------+ | +---------+
2526 In this configuration, the switches are isolated from one
2527 another. One reason to employ a topology such as this is for an
2528 isolated network with many hosts (a cluster configured for high
2529 performance, for example), using multiple smaller switches can be more
2530 cost effective than a single larger switch, e.g., on a network with 24
2531 hosts, three 24 port switches can be significantly less expensive than
2532 a single 72 port switch.
2534 If access beyond the network is required, an individual host
2535 can be equipped with an additional network device connected to an
2536 external network; this host then additionally acts as a gateway.
2538 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2539 -------------------------------------------------------------
2541 In actual practice, the bonding mode typically employed in
2542 configurations of this type is balance-rr. Historically, in this
2543 network configuration, the usual caveats about out of order packet
2544 delivery are mitigated by the use of network adapters that do not do
2545 any kind of packet coalescing (via the use of NAPI, or because the
2546 device itself does not generate interrupts until some number of
2547 packets has arrived). When employed in this fashion, the balance-rr
2548 mode allows individual connections between two hosts to effectively
2549 utilize greater than one interface's bandwidth.
2551 12.2.2 MT Link Monitoring for Multiple Switch Topology
2552 ------------------------------------------------------
2554 Again, in actual practice, the MII monitor is most often used
2555 in this configuration, as performance is given preference over
2556 availability. The ARP monitor will function in this topology, but its
2557 advantages over the MII monitor are mitigated by the volume of probes
2558 needed as the number of systems involved grows (remember that each
2559 host in the network is configured with bonding).
2561 13. Switch Behavior Issues
2562 ==========================
2564 13.1 Link Establishment and Failover Delays
2565 -------------------------------------------
2567 Some switches exhibit undesirable behavior with regard to the
2568 timing of link up and down reporting by the switch.
2570 First, when a link comes up, some switches may indicate that
2571 the link is up (carrier available), but not pass traffic over the
2572 interface for some period of time. This delay is typically due to
2573 some type of autonegotiation or routing protocol, but may also occur
2574 during switch initialization (e.g., during recovery after a switch
2575 failure). If you find this to be a problem, specify an appropriate
2576 value to the updelay bonding module option to delay the use of the
2577 relevant interface(s).
2579 Second, some switches may "bounce" the link state one or more
2580 times while a link is changing state. This occurs most commonly while
2581 the switch is initializing. Again, an appropriate updelay value may
2584 Note that when a bonding interface has no active links, the
2585 driver will immediately reuse the first link that goes up, even if the
2586 updelay parameter has been specified (the updelay is ignored in this
2587 case). If there are slave interfaces waiting for the updelay timeout
2588 to expire, the interface that first went into that state will be
2589 immediately reused. This reduces down time of the network if the
2590 value of updelay has been overestimated, and since this occurs only in
2591 cases with no connectivity, there is no additional penalty for
2592 ignoring the updelay.
2594 In addition to the concerns about switch timings, if your
2595 switches take a long time to go into backup mode, it may be desirable
2596 to not activate a backup interface immediately after a link goes down.
2597 Failover may be delayed via the downdelay bonding module option.
2599 13.2 Duplicated Incoming Packets
2600 --------------------------------
2602 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2603 suppress duplicate packets, which should largely eliminate this problem.
2604 The following description is kept for reference.
2606 It is not uncommon to observe a short burst of duplicated
2607 traffic when the bonding device is first used, or after it has been
2608 idle for some period of time. This is most easily observed by issuing
2609 a "ping" to some other host on the network, and noticing that the
2610 output from ping flags duplicates (typically one per slave).
2612 For example, on a bond in active-backup mode with five slaves
2613 all connected to one switch, the output may appear as follows::
2616 PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
2617 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
2618 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2619 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2620 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2621 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2622 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
2623 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
2624 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2626 This is not due to an error in the bonding driver, rather, it
2627 is a side effect of how many switches update their MAC forwarding
2628 tables. Initially, the switch does not associate the MAC address in
2629 the packet with a particular switch port, and so it may send the
2630 traffic to all ports until its MAC forwarding table is updated. Since
2631 the interfaces attached to the bond may occupy multiple ports on a
2632 single switch, when the switch (temporarily) floods the traffic to all
2633 ports, the bond device receives multiple copies of the same packet
2634 (one per slave device).
2636 The duplicated packet behavior is switch dependent, some
2637 switches exhibit this, and some do not. On switches that display this
2638 behavior, it can be induced by clearing the MAC forwarding table (on
2639 most Cisco switches, the privileged command "clear mac address-table
2640 dynamic" will accomplish this).
2642 14. Hardware Specific Considerations
2643 ====================================
2645 This section contains additional information for configuring
2646 bonding on specific hardware platforms, or for interfacing bonding
2647 with particular switches or other devices.
2649 14.1 IBM BladeCenter
2650 --------------------
2652 This applies to the JS20 and similar systems.
2654 On the JS20 blades, the bonding driver supports only
2655 balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2656 largely due to the network topology inside the BladeCenter, detailed
2659 JS20 network adapter information
2660 --------------------------------
2662 All JS20s come with two Broadcom Gigabit Ethernet ports
2663 integrated on the planar (that's "motherboard" in IBM-speak). In the
2664 BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2665 I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2666 An add-on Broadcom daughter card can be installed on a JS20 to provide
2667 two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2668 wired to I/O Modules 3 and 4, respectively.
2670 Each I/O Module may contain either a switch or a passthrough
2671 module (which allows ports to be directly connected to an external
2672 switch). Some bonding modes require a specific BladeCenter internal
2673 network topology in order to function; these are detailed below.
2675 Additional BladeCenter-specific networking information can be
2676 found in two IBM Redbooks (www.ibm.com/redbooks):
2678 - "IBM eServer BladeCenter Networking Options"
2679 - "IBM eServer BladeCenter Layer 2-7 Network Switching"
2681 BladeCenter networking configuration
2682 ------------------------------------
2684 Because a BladeCenter can be configured in a very large number
2685 of ways, this discussion will be confined to describing basic
2688 Normally, Ethernet Switch Modules (ESMs) are used in I/O
2689 modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2690 JS20 will be connected to different internal switches (in the
2691 respective I/O modules).
2693 A passthrough module (OPM or CPM, optical or copper,
2694 passthrough module) connects the I/O module directly to an external
2695 switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2696 interfaces of a JS20 can be redirected to the outside world and
2697 connected to a common external switch.
2699 Depending upon the mix of ESMs and PMs, the network will
2700 appear to bonding as either a single switch topology (all PMs) or as a
2701 multiple switch topology (one or more ESMs, zero or more PMs). It is
2702 also possible to connect ESMs together, resulting in a configuration
2703 much like the example in "High Availability in a Multiple Switch
2706 Requirements for specific modes
2707 -------------------------------
2709 The balance-rr mode requires the use of passthrough modules
2710 for devices in the bond, all connected to an common external switch.
2711 That switch must be configured for "etherchannel" or "trunking" on the
2712 appropriate ports, as is usual for balance-rr.
2714 The balance-alb and balance-tlb modes will function with
2715 either switch modules or passthrough modules (or a mix). The only
2716 specific requirement for these modes is that all network interfaces
2717 must be able to reach all destinations for traffic sent over the
2718 bonding device (i.e., the network must converge at some point outside
2721 The active-backup mode has no additional requirements.
2723 Link monitoring issues
2724 ----------------------
2726 When an Ethernet Switch Module is in place, only the ARP
2727 monitor will reliably detect link loss to an external switch. This is
2728 nothing unusual, but examination of the BladeCenter cabinet would
2729 suggest that the "external" network ports are the ethernet ports for
2730 the system, when it fact there is a switch between these "external"
2731 ports and the devices on the JS20 system itself. The MII monitor is
2732 only able to detect link failures between the ESM and the JS20 system.
2734 When a passthrough module is in place, the MII monitor does
2735 detect failures to the "external" port, which is then directly
2736 connected to the JS20 system.
2741 The Serial Over LAN (SoL) link is established over the primary
2742 ethernet (eth0) only, therefore, any loss of link to eth0 will result
2743 in losing your SoL connection. It will not fail over with other
2744 network traffic, as the SoL system is beyond the control of the
2747 It may be desirable to disable spanning tree on the switch
2748 (either the internal Ethernet Switch Module, or an external switch) to
2749 avoid fail-over delay issues when using bonding.
2752 15. Frequently Asked Questions
2753 ==============================
2758 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2759 The new driver was designed to be SMP safe from the start.
2761 2. What type of cards will work with it?
2762 -----------------------------------------
2764 Any Ethernet type cards (you can even mix cards - a Intel
2765 EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2766 devices need not be of the same speed.
2768 Starting with version 3.2.1, bonding also supports Infiniband
2769 slaves in active-backup mode.
2771 3. How many bonding devices can I have?
2772 ----------------------------------------
2776 4. How many slaves can a bonding device have?
2777 ----------------------------------------------
2779 This is limited only by the number of network interfaces Linux
2780 supports and/or the number of network cards you can place in your
2783 5. What happens when a slave link dies?
2784 ----------------------------------------
2786 If link monitoring is enabled, then the failing device will be
2787 disabled. The active-backup mode will fail over to a backup link, and
2788 other modes will ignore the failed link. The link will continue to be
2789 monitored, and should it recover, it will rejoin the bond (in whatever
2790 manner is appropriate for the mode). See the sections on High
2791 Availability and the documentation for each mode for additional
2794 Link monitoring can be enabled via either the miimon or
2795 arp_interval parameters (described in the module parameters section,
2796 above). In general, miimon monitors the carrier state as sensed by
2797 the underlying network device, and the arp monitor (arp_interval)
2798 monitors connectivity to another host on the local network.
2800 If no link monitoring is configured, the bonding driver will
2801 be unable to detect link failures, and will assume that all links are
2802 always available. This will likely result in lost packets, and a
2803 resulting degradation of performance. The precise performance loss
2804 depends upon the bonding mode and network configuration.
2806 6. Can bonding be used for High Availability?
2807 ----------------------------------------------
2809 Yes. See the section on High Availability for details.
2811 7. Which switches/systems does it work with?
2812 ---------------------------------------------
2814 The full answer to this depends upon the desired mode.
2816 In the basic balance modes (balance-rr and balance-xor), it
2817 works with any system that supports etherchannel (also called
2818 trunking). Most managed switches currently available have such
2819 support, and many unmanaged switches as well.
2821 The advanced balance modes (balance-tlb and balance-alb) do
2822 not have special switch requirements, but do need device drivers that
2823 support specific features (described in the appropriate section under
2824 module parameters, above).
2826 In 802.3ad mode, it works with systems that support IEEE
2827 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2828 switches currently available support 802.3ad.
2830 The active-backup mode should work with any Layer-II switch.
2832 8. Where does a bonding device get its MAC address from?
2833 ---------------------------------------------------------
2835 When using slave devices that have fixed MAC addresses, or when
2836 the fail_over_mac option is enabled, the bonding device's MAC address is
2837 the MAC address of the active slave.
2839 For other configurations, if not explicitly configured (with
2840 ifconfig or ip link), the MAC address of the bonding device is taken from
2841 its first slave device. This MAC address is then passed to all following
2842 slaves and remains persistent (even if the first slave is removed) until
2843 the bonding device is brought down or reconfigured.
2845 If you wish to change the MAC address, you can set it with
2846 ifconfig or ip link::
2848 # ifconfig bond0 hw ether 00:11:22:33:44:55
2850 # ip link set bond0 address 66:77:88:99:aa:bb
2852 The MAC address can be also changed by bringing down/up the
2853 device and then changing its slaves (or their order)::
2855 # ifconfig bond0 down ; modprobe -r bonding
2856 # ifconfig bond0 .... up
2857 # ifenslave bond0 eth...
2859 This method will automatically take the address from the next
2860 slave that is added.
2862 To restore your slaves' MAC addresses, you need to detach them
2863 from the bond (``ifenslave -d bond0 eth0``). The bonding driver will
2864 then restore the MAC addresses that the slaves had before they were
2867 16. Resources and Links
2868 =======================
2870 The latest version of the bonding driver can be found in the latest
2871 version of the linux kernel, found on http://kernel.org
2873 The latest version of this document can be found in the latest kernel
2874 source (named Documentation/networking/bonding.rst).
2876 Discussions regarding the development of the bonding driver take place
2877 on the main Linux network mailing list, hosted at vger.kernel.org. The list
2880 netdev@vger.kernel.org
2882 The administrative interface (to subscribe or unsubscribe) can
2885 http://vger.kernel.org/vger-lists.html#netdev