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.
505 Option specifying whether to send LACPDU frames periodically.
508 LACPDU frames acts as "speak when spoken to".
511 LACPDU frames are sent along the configured links
512 periodically. See lacp_rate for more details.
518 Option specifying the rate in which we'll ask our link partner
519 to transmit LACPDU packets in 802.3ad mode. Possible values
523 Request partner to transmit LACPDUs every 30 seconds
526 Request partner to transmit LACPDUs every 1 second
532 Specifies the number of bonding devices to create for this
533 instance of the bonding driver. E.g., if max_bonds is 3, and
534 the bonding driver is not already loaded, then bond0, bond1
535 and bond2 will be created. The default value is 1. Specifying
536 a value of 0 will load bonding, but will not create any devices.
540 Specifies the MII link monitoring frequency in milliseconds.
541 This determines how often the link state of each slave is
542 inspected for link failures. A value of zero disables MII
543 link monitoring. A value of 100 is a good starting point.
544 The use_carrier option, below, affects how the link state is
545 determined. See the High Availability section for additional
546 information. The default value is 0.
550 Specifies the minimum number of links that must be active before
551 asserting carrier. It is similar to the Cisco EtherChannel min-links
552 feature. This allows setting the minimum number of member ports that
553 must be up (link-up state) before marking the bond device as up
554 (carrier on). This is useful for situations where higher level services
555 such as clustering want to ensure a minimum number of low bandwidth
556 links are active before switchover. This option only affect 802.3ad
559 The default value is 0. This will cause carrier to be asserted (for
560 802.3ad mode) whenever there is an active aggregator, regardless of the
561 number of available links in that aggregator. Note that, because an
562 aggregator cannot be active without at least one available link,
563 setting this option to 0 or to 1 has the exact same effect.
567 Specifies one of the bonding policies. The default is
568 balance-rr (round robin). Possible values are:
572 Round-robin policy: Transmit packets in sequential
573 order from the first available slave through the
574 last. This mode provides load balancing and fault
579 Active-backup policy: Only one slave in the bond is
580 active. A different slave becomes active if, and only
581 if, the active slave fails. The bond's MAC address is
582 externally visible on only one port (network adapter)
583 to avoid confusing the switch.
585 In bonding version 2.6.2 or later, when a failover
586 occurs in active-backup mode, bonding will issue one
587 or more gratuitous ARPs on the newly active slave.
588 One gratuitous ARP is issued for the bonding master
589 interface and each VLAN interfaces configured above
590 it, provided that the interface has at least one IP
591 address configured. Gratuitous ARPs issued for VLAN
592 interfaces are tagged with the appropriate VLAN id.
594 This mode provides fault tolerance. The primary
595 option, documented below, affects the behavior of this
600 XOR policy: Transmit based on the selected transmit
601 hash policy. The default policy is a simple [(source
602 MAC address XOR'd with destination MAC address XOR
603 packet type ID) modulo slave count]. Alternate transmit
604 policies may be selected via the xmit_hash_policy option,
607 This mode provides load balancing and fault tolerance.
611 Broadcast policy: transmits everything on all slave
612 interfaces. This mode provides fault tolerance.
616 IEEE 802.3ad Dynamic link aggregation. Creates
617 aggregation groups that share the same speed and
618 duplex settings. Utilizes all slaves in the active
619 aggregator according to the 802.3ad specification.
621 Slave selection for outgoing traffic is done according
622 to the transmit hash policy, which may be changed from
623 the default simple XOR policy via the xmit_hash_policy
624 option, documented below. Note that not all transmit
625 policies may be 802.3ad compliant, particularly in
626 regards to the packet mis-ordering requirements of
627 section 43.2.4 of the 802.3ad standard. Differing
628 peer implementations will have varying tolerances for
633 1. Ethtool support in the base drivers for retrieving
634 the speed and duplex of each slave.
636 2. A switch that supports IEEE 802.3ad Dynamic link
639 Most switches will require some type of configuration
640 to enable 802.3ad mode.
644 Adaptive transmit load balancing: channel bonding that
645 does not require any special switch support.
647 In tlb_dynamic_lb=1 mode; the outgoing traffic is
648 distributed according to the current load (computed
649 relative to the speed) on each slave.
651 In tlb_dynamic_lb=0 mode; the load balancing based on
652 current load is disabled and the load is distributed
653 only using the hash distribution.
655 Incoming traffic is received by the current slave.
656 If the receiving slave fails, another slave takes over
657 the MAC address of the failed receiving slave.
661 Ethtool support in the base drivers for retrieving the
666 Adaptive load balancing: includes balance-tlb plus
667 receive load balancing (rlb) for IPV4 traffic, and
668 does not require any special switch support. The
669 receive load balancing is achieved by ARP negotiation.
670 The bonding driver intercepts the ARP Replies sent by
671 the local system on their way out and overwrites the
672 source hardware address with the unique hardware
673 address of one of the slaves in the bond such that
674 different peers use different hardware addresses for
677 Receive traffic from connections created by the server
678 is also balanced. When the local system sends an ARP
679 Request the bonding driver copies and saves the peer's
680 IP information from the ARP packet. When the ARP
681 Reply arrives from the peer, its hardware address is
682 retrieved and the bonding driver initiates an ARP
683 reply to this peer assigning it to one of the slaves
684 in the bond. A problematic outcome of using ARP
685 negotiation for balancing is that each time that an
686 ARP request is broadcast it uses the hardware address
687 of the bond. Hence, peers learn the hardware address
688 of the bond and the balancing of receive traffic
689 collapses to the current slave. This is handled by
690 sending updates (ARP Replies) to all the peers with
691 their individually assigned hardware address such that
692 the traffic is redistributed. Receive traffic is also
693 redistributed when a new slave is added to the bond
694 and when an inactive slave is re-activated. The
695 receive load is distributed sequentially (round robin)
696 among the group of highest speed slaves in the bond.
698 When a link is reconnected or a new slave joins the
699 bond the receive traffic is redistributed among all
700 active slaves in the bond by initiating ARP Replies
701 with the selected MAC address to each of the
702 clients. The updelay parameter (detailed below) must
703 be set to a value equal or greater than the switch's
704 forwarding delay so that the ARP Replies sent to the
705 peers will not be blocked by the switch.
709 1. Ethtool support in the base drivers for retrieving
710 the speed of each slave.
712 2. Base driver support for setting the hardware
713 address of a device while it is open. This is
714 required so that there will always be one slave in the
715 team using the bond hardware address (the
716 curr_active_slave) while having a unique hardware
717 address for each slave in the bond. If the
718 curr_active_slave fails its hardware address is
719 swapped with the new curr_active_slave that was
725 Specify the number of peer notifications (gratuitous ARPs and
726 unsolicited IPv6 Neighbor Advertisements) to be issued after a
727 failover event. As soon as the link is up on the new slave
728 (possibly immediately) a peer notification is sent on the
729 bonding device and each VLAN sub-device. This is repeated at
730 the rate specified by peer_notif_delay if the number is
733 The valid range is 0 - 255; the default value is 1. These options
734 affect only the active-backup mode. These options were added for
735 bonding versions 3.3.0 and 3.4.0 respectively.
737 From Linux 3.0 and bonding version 3.7.1, these notifications
738 are generated by the ipv4 and ipv6 code and the numbers of
739 repetitions cannot be set independently.
743 Specify the number of packets to transmit through a slave before
744 moving to the next one. When set to 0 then a slave is chosen at
747 The valid range is 0 - 65535; the default value is 1. This option
748 has effect only in balance-rr mode.
752 Specify the delay, in milliseconds, between each peer
753 notification (gratuitous ARP and unsolicited IPv6 Neighbor
754 Advertisement) when they are issued after a failover event.
755 This delay should be a multiple of the link monitor interval
756 (arp_interval or miimon, whichever is active). The default
757 value is 0 which means to match the value of the link monitor
762 A string (eth0, eth2, etc) specifying which slave is the
763 primary device. The specified device will always be the
764 active slave while it is available. Only when the primary is
765 off-line will alternate devices be used. This is useful when
766 one slave is preferred over another, e.g., when one slave has
767 higher throughput than another.
769 The primary option is only valid for active-backup(1),
770 balance-tlb (5) and balance-alb (6) mode.
774 Specifies the reselection policy for the primary slave. This
775 affects how the primary slave is chosen to become the active slave
776 when failure of the active slave or recovery of the primary slave
777 occurs. This option is designed to prevent flip-flopping between
778 the primary slave and other slaves. Possible values are:
780 always or 0 (default)
782 The primary slave becomes the active slave whenever it
787 The primary slave becomes the active slave when it comes
788 back up, if the speed and duplex of the primary slave is
789 better than the speed and duplex of the current active
794 The primary slave becomes the active slave only if the
795 current active slave fails and the primary slave is up.
797 The primary_reselect setting is ignored in two cases:
799 If no slaves are active, the first slave to recover is
800 made the active slave.
802 When initially enslaved, the primary slave is always made
805 Changing the primary_reselect policy via sysfs will cause an
806 immediate selection of the best active slave according to the new
807 policy. This may or may not result in a change of the active
808 slave, depending upon the circumstances.
810 This option was added for bonding version 3.6.0.
814 Specifies if dynamic shuffling of flows is enabled in tlb
815 mode. The value has no effect on any other modes.
817 The default behavior of tlb mode is to shuffle active flows across
818 slaves based on the load in that interval. This gives nice lb
819 characteristics but can cause packet reordering. If re-ordering is
820 a concern use this variable to disable flow shuffling and rely on
821 load balancing provided solely by the hash distribution.
822 xmit-hash-policy can be used to select the appropriate hashing for
825 The sysfs entry can be used to change the setting per bond device
826 and the initial value is derived from the module parameter. The
827 sysfs entry is allowed to be changed only if the bond device is
830 The default value is "1" that enables flow shuffling while value "0"
831 disables it. This option was added in bonding driver 3.7.1
836 Specifies the time, in milliseconds, to wait before enabling a
837 slave after a link recovery has been detected. This option is
838 only valid for the miimon link monitor. The updelay value
839 should be a multiple of the miimon value; if not, it will be
840 rounded down to the nearest multiple. The default value is 0.
844 Specifies whether or not miimon should use MII or ETHTOOL
845 ioctls vs. netif_carrier_ok() to determine the link
846 status. The MII or ETHTOOL ioctls are less efficient and
847 utilize a deprecated calling sequence within the kernel. The
848 netif_carrier_ok() relies on the device driver to maintain its
849 state with netif_carrier_on/off; at this writing, most, but
850 not all, device drivers support this facility.
852 If bonding insists that the link is up when it should not be,
853 it may be that your network device driver does not support
854 netif_carrier_on/off. The default state for netif_carrier is
855 "carrier on," so if a driver does not support netif_carrier,
856 it will appear as if the link is always up. In this case,
857 setting use_carrier to 0 will cause bonding to revert to the
858 MII / ETHTOOL ioctl method to determine the link state.
860 A value of 1 enables the use of netif_carrier_ok(), a value of
861 0 will use the deprecated MII / ETHTOOL ioctls. The default
866 Selects the transmit hash policy to use for slave selection in
867 balance-xor, 802.3ad, and tlb modes. Possible values are:
871 Uses XOR of hardware MAC addresses and packet type ID
872 field to generate the hash. The formula is
874 hash = source MAC XOR destination MAC XOR packet type ID
875 slave number = hash modulo slave count
877 This algorithm will place all traffic to a particular
878 network peer on the same slave.
880 This algorithm is 802.3ad compliant.
884 This policy uses a combination of layer2 and layer3
885 protocol information to generate the hash.
887 Uses XOR of hardware MAC addresses and IP addresses to
888 generate the hash. The formula is
890 hash = source MAC XOR destination MAC XOR packet type ID
891 hash = hash XOR source IP XOR destination IP
892 hash = hash XOR (hash RSHIFT 16)
893 hash = hash XOR (hash RSHIFT 8)
894 And then hash is reduced modulo slave count.
896 If the protocol is IPv6 then the source and destination
897 addresses are first hashed using ipv6_addr_hash.
899 This algorithm will place all traffic to a particular
900 network peer on the same slave. For non-IP traffic,
901 the formula is the same as for the layer2 transmit
904 This policy is intended to provide a more balanced
905 distribution of traffic than layer2 alone, especially
906 in environments where a layer3 gateway device is
907 required to reach most destinations.
909 This algorithm is 802.3ad compliant.
913 This policy uses upper layer protocol information,
914 when available, to generate the hash. This allows for
915 traffic to a particular network peer to span multiple
916 slaves, although a single connection will not span
919 The formula for unfragmented TCP and UDP packets is
921 hash = source port, destination port (as in the header)
922 hash = hash XOR source IP XOR destination IP
923 hash = hash XOR (hash RSHIFT 16)
924 hash = hash XOR (hash RSHIFT 8)
925 And then hash is reduced modulo slave count.
927 If the protocol is IPv6 then the source and destination
928 addresses are first hashed using ipv6_addr_hash.
930 For fragmented TCP or UDP packets and all other IPv4 and
931 IPv6 protocol traffic, the source and destination port
932 information is omitted. For non-IP traffic, the
933 formula is the same as for the layer2 transmit hash
936 This algorithm is not fully 802.3ad compliant. A
937 single TCP or UDP conversation containing both
938 fragmented and unfragmented packets will see packets
939 striped across two interfaces. This may result in out
940 of order delivery. Most traffic types will not meet
941 this criteria, as TCP rarely fragments traffic, and
942 most UDP traffic is not involved in extended
943 conversations. Other implementations of 802.3ad may
944 or may not tolerate this noncompliance.
948 This policy uses the same formula as layer2+3 but it
949 relies on skb_flow_dissect to obtain the header fields
950 which might result in the use of inner headers if an
951 encapsulation protocol is used. For example this will
952 improve the performance for tunnel users because the
953 packets will be distributed according to the encapsulated
958 This policy uses the same formula as layer3+4 but it
959 relies on skb_flow_dissect to obtain the header fields
960 which might result in the use of inner headers if an
961 encapsulation protocol is used. For example this will
962 improve the performance for tunnel users because the
963 packets will be distributed according to the encapsulated
968 This policy uses a very rudimentary vlan ID and source mac
969 hash to load-balance traffic per-vlan, with failover
970 should one leg fail. The intended use case is for a bond
971 shared by multiple virtual machines, all configured to
972 use their own vlan, to give lacp-like functionality
973 without requiring lacp-capable switching hardware.
975 The formula for the hash is simply
977 hash = (vlan ID) XOR (source MAC vendor) XOR (source MAC dev)
979 The default value is layer2. This option was added in bonding
980 version 2.6.3. In earlier versions of bonding, this parameter
981 does not exist, and the layer2 policy is the only policy. The
982 layer2+3 value was added for bonding version 3.2.2.
986 Specifies the number of IGMP membership reports to be issued after
987 a failover event. One membership report is issued immediately after
988 the failover, subsequent packets are sent in each 200ms interval.
990 The valid range is 0 - 255; the default value is 1. A value of 0
991 prevents the IGMP membership report from being issued in response
992 to the failover event.
994 This option is useful for bonding modes balance-rr (0), active-backup
995 (1), balance-tlb (5) and balance-alb (6), in which a failover can
996 switch the IGMP traffic from one slave to another. Therefore a fresh
997 IGMP report must be issued to cause the switch to forward the incoming
998 IGMP traffic over the newly selected slave.
1000 This option was added for bonding version 3.7.0.
1004 Specifies the number of seconds between instances where the bonding
1005 driver sends learning packets to each slaves peer switch.
1007 The valid range is 1 - 0x7fffffff; the default value is 1. This Option
1008 has effect only in balance-tlb and balance-alb modes.
1010 3. Configuring Bonding Devices
1011 ==============================
1013 You can configure bonding using either your distro's network
1014 initialization scripts, or manually using either iproute2 or the
1015 sysfs interface. Distros generally use one of three packages for the
1016 network initialization scripts: initscripts, sysconfig or interfaces.
1017 Recent versions of these packages have support for bonding, while older
1020 We will first describe the options for configuring bonding for
1021 distros using versions of initscripts, sysconfig and interfaces with full
1022 or partial support for bonding, then provide information on enabling
1023 bonding without support from the network initialization scripts (i.e.,
1024 older versions of initscripts or sysconfig).
1026 If you're unsure whether your distro uses sysconfig,
1027 initscripts or interfaces, or don't know if it's new enough, have no fear.
1028 Determining this is fairly straightforward.
1030 First, look for a file called interfaces in /etc/network directory.
1031 If this file is present in your system, then your system use interfaces. See
1032 Configuration with Interfaces Support.
1034 Else, issue the command::
1036 $ rpm -qf /sbin/ifup
1038 It will respond with a line of text starting with either
1039 "initscripts" or "sysconfig," followed by some numbers. This is the
1040 package that provides your network initialization scripts.
1042 Next, to determine if your installation supports bonding,
1045 $ grep ifenslave /sbin/ifup
1047 If this returns any matches, then your initscripts or
1048 sysconfig has support for bonding.
1050 3.1 Configuration with Sysconfig Support
1051 ----------------------------------------
1053 This section applies to distros using a version of sysconfig
1054 with bonding support, for example, SuSE Linux Enterprise Server 9.
1056 SuSE SLES 9's networking configuration system does support
1057 bonding, however, at this writing, the YaST system configuration
1058 front end does not provide any means to work with bonding devices.
1059 Bonding devices can be managed by hand, however, as follows.
1061 First, if they have not already been configured, configure the
1062 slave devices. On SLES 9, this is most easily done by running the
1063 yast2 sysconfig configuration utility. The goal is for to create an
1064 ifcfg-id file for each slave device. The simplest way to accomplish
1065 this is to configure the devices for DHCP (this is only to get the
1066 file ifcfg-id file created; see below for some issues with DHCP). The
1067 name of the configuration file for each device will be of the form::
1069 ifcfg-id-xx:xx:xx:xx:xx:xx
1071 Where the "xx" portion will be replaced with the digits from
1072 the device's permanent MAC address.
1074 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
1075 created, it is necessary to edit the configuration files for the slave
1076 devices (the MAC addresses correspond to those of the slave devices).
1077 Before editing, the file will contain multiple lines, and will look
1078 something like this::
1083 UNIQUE='XNzu.WeZGOGF+4wE'
1084 _nm_name='bus-pci-0001:61:01.0'
1086 Change the BOOTPROTO and STARTMODE lines to the following::
1091 Do not alter the UNIQUE or _nm_name lines. Remove any other
1092 lines (USERCTL, etc).
1094 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
1095 it's time to create the configuration file for the bonding device
1096 itself. This file is named ifcfg-bondX, where X is the number of the
1097 bonding device to create, starting at 0. The first such file is
1098 ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
1099 network configuration system will correctly start multiple instances
1102 The contents of the ifcfg-bondX file is as follows::
1105 BROADCAST="10.0.2.255"
1107 NETMASK="255.255.0.0"
1111 BONDING_MASTER="yes"
1112 BONDING_MODULE_OPTS="mode=active-backup miimon=100"
1113 BONDING_SLAVE0="eth0"
1114 BONDING_SLAVE1="bus-pci-0000:06:08.1"
1116 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
1117 values with the appropriate values for your network.
1119 The STARTMODE specifies when the device is brought online.
1120 The possible values are:
1122 ======== ======================================================
1123 onboot The device is started at boot time. If you're not
1124 sure, this is probably what you want.
1126 manual The device is started only when ifup is called
1127 manually. Bonding devices may be configured this
1128 way if you do not wish them to start automatically
1129 at boot for some reason.
1131 hotplug The device is started by a hotplug event. This is not
1132 a valid choice for a bonding device.
1134 off or The device configuration is ignored.
1136 ======== ======================================================
1138 The line BONDING_MASTER='yes' indicates that the device is a
1139 bonding master device. The only useful value is "yes."
1141 The contents of BONDING_MODULE_OPTS are supplied to the
1142 instance of the bonding module for this device. Specify the options
1143 for the bonding mode, link monitoring, and so on here. Do not include
1144 the max_bonds bonding parameter; this will confuse the configuration
1145 system if you have multiple bonding devices.
1147 Finally, supply one BONDING_SLAVEn="slave device" for each
1148 slave. where "n" is an increasing value, one for each slave. The
1149 "slave device" is either an interface name, e.g., "eth0", or a device
1150 specifier for the network device. The interface name is easier to
1151 find, but the ethN names are subject to change at boot time if, e.g.,
1152 a device early in the sequence has failed. The device specifiers
1153 (bus-pci-0000:06:08.1 in the example above) specify the physical
1154 network device, and will not change unless the device's bus location
1155 changes (for example, it is moved from one PCI slot to another). The
1156 example above uses one of each type for demonstration purposes; most
1157 configurations will choose one or the other for all slave devices.
1159 When all configuration files have been modified or created,
1160 networking must be restarted for the configuration changes to take
1161 effect. This can be accomplished via the following::
1163 # /etc/init.d/network restart
1165 Note that the network control script (/sbin/ifdown) will
1166 remove the bonding module as part of the network shutdown processing,
1167 so it is not necessary to remove the module by hand if, e.g., the
1168 module parameters have changed.
1170 Also, at this writing, YaST/YaST2 will not manage bonding
1171 devices (they do not show bonding interfaces on its list of network
1172 devices). It is necessary to edit the configuration file by hand to
1173 change the bonding configuration.
1175 Additional general options and details of the ifcfg file
1176 format can be found in an example ifcfg template file::
1178 /etc/sysconfig/network/ifcfg.template
1180 Note that the template does not document the various ``BONDING_*``
1181 settings described above, but does describe many of the other options.
1183 3.1.1 Using DHCP with Sysconfig
1184 -------------------------------
1186 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1187 will cause it to query DHCP for its IP address information. At this
1188 writing, this does not function for bonding devices; the scripts
1189 attempt to obtain the device address from DHCP prior to adding any of
1190 the slave devices. Without active slaves, the DHCP requests are not
1191 sent to the network.
1193 3.1.2 Configuring Multiple Bonds with Sysconfig
1194 -----------------------------------------------
1196 The sysconfig network initialization system is capable of
1197 handling multiple bonding devices. All that is necessary is for each
1198 bonding instance to have an appropriately configured ifcfg-bondX file
1199 (as described above). Do not specify the "max_bonds" parameter to any
1200 instance of bonding, as this will confuse sysconfig. If you require
1201 multiple bonding devices with identical parameters, create multiple
1204 Because the sysconfig scripts supply the bonding module
1205 options in the ifcfg-bondX file, it is not necessary to add them to
1206 the system ``/etc/modules.d/*.conf`` configuration files.
1208 3.2 Configuration with Initscripts Support
1209 ------------------------------------------
1211 This section applies to distros using a recent version of
1212 initscripts with bonding support, for example, Red Hat Enterprise Linux
1213 version 3 or later, Fedora, etc. On these systems, the network
1214 initialization scripts have knowledge of bonding, and can be configured to
1215 control bonding devices. Note that older versions of the initscripts
1216 package have lower levels of support for bonding; this will be noted where
1219 These distros will not automatically load the network adapter
1220 driver unless the ethX device is configured with an IP address.
1221 Because of this constraint, users must manually configure a
1222 network-script file for all physical adapters that will be members of
1223 a bondX link. Network script files are located in the directory:
1225 /etc/sysconfig/network-scripts
1227 The file name must be prefixed with "ifcfg-eth" and suffixed
1228 with the adapter's physical adapter number. For example, the script
1229 for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1230 Place the following text in the file::
1239 The DEVICE= line will be different for every ethX device and
1240 must correspond with the name of the file, i.e., ifcfg-eth1 must have
1241 a device line of DEVICE=eth1. The setting of the MASTER= line will
1242 also depend on the final bonding interface name chosen for your bond.
1243 As with other network devices, these typically start at 0, and go up
1244 one for each device, i.e., the first bonding instance is bond0, the
1245 second is bond1, and so on.
1247 Next, create a bond network script. The file name for this
1248 script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1249 the number of the bond. For bond0 the file is named "ifcfg-bond0",
1250 for bond1 it is named "ifcfg-bond1", and so on. Within that file,
1251 place the following text::
1255 NETMASK=255.255.255.0
1257 BROADCAST=192.168.1.255
1262 Be sure to change the networking specific lines (IPADDR,
1263 NETMASK, NETWORK and BROADCAST) to match your network configuration.
1265 For later versions of initscripts, such as that found with Fedora
1266 7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1267 and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
1268 file, e.g. a line of the format::
1270 BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
1272 will configure the bond with the specified options. The options
1273 specified in BONDING_OPTS are identical to the bonding module parameters
1274 except for the arp_ip_target field when using versions of initscripts older
1275 than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
1276 using older versions each target should be included as a separate option and
1277 should be preceded by a '+' to indicate it should be added to the list of
1278 queried targets, e.g.,::
1280 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1282 is the proper syntax to specify multiple targets. When specifying
1283 options via BONDING_OPTS, it is not necessary to edit
1284 ``/etc/modprobe.d/*.conf``.
1286 For even older versions of initscripts that do not support
1287 BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1288 your distro) to load the bonding module with your desired options when the
1289 bond0 interface is brought up. The following lines in /etc/modprobe.d/*.conf
1290 will load the bonding module, and select its options:
1293 options bond0 mode=balance-alb miimon=100
1295 Replace the sample parameters with the appropriate set of
1296 options for your configuration.
1298 Finally run "/etc/rc.d/init.d/network restart" as root. This
1299 will restart the networking subsystem and your bond link should be now
1302 3.2.1 Using DHCP with Initscripts
1303 ---------------------------------
1305 Recent versions of initscripts (the versions supplied with Fedora
1306 Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1307 work) have support for assigning IP information to bonding devices via
1310 To configure bonding for DHCP, configure it as described
1311 above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1312 and add a line consisting of "TYPE=Bonding". Note that the TYPE value
1315 3.2.2 Configuring Multiple Bonds with Initscripts
1316 -------------------------------------------------
1318 Initscripts packages that are included with Fedora 7 and Red Hat
1319 Enterprise Linux 5 support multiple bonding interfaces by simply
1320 specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1321 number of the bond. This support requires sysfs support in the kernel,
1322 and a bonding driver of version 3.0.0 or later. Other configurations may
1323 not support this method for specifying multiple bonding interfaces; for
1324 those instances, see the "Configuring Multiple Bonds Manually" section,
1327 3.3 Configuring Bonding Manually with iproute2
1328 -----------------------------------------------
1330 This section applies to distros whose network initialization
1331 scripts (the sysconfig or initscripts package) do not have specific
1332 knowledge of bonding. One such distro is SuSE Linux Enterprise Server
1335 The general method for these systems is to place the bonding
1336 module parameters into a config file in /etc/modprobe.d/ (as
1337 appropriate for the installed distro), then add modprobe and/or
1338 `ip link` commands to the system's global init script. The name of
1339 the global init script differs; for sysconfig, it is
1340 /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1342 For example, if you wanted to make a simple bond of two e100
1343 devices (presumed to be eth0 and eth1), and have it persist across
1344 reboots, edit the appropriate file (/etc/init.d/boot.local or
1345 /etc/rc.d/rc.local), and add the following::
1347 modprobe bonding mode=balance-alb miimon=100
1349 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1350 ip link set eth0 master bond0
1351 ip link set eth1 master bond0
1353 Replace the example bonding module parameters and bond0
1354 network configuration (IP address, netmask, etc) with the appropriate
1355 values for your configuration.
1357 Unfortunately, this method will not provide support for the
1358 ifup and ifdown scripts on the bond devices. To reload the bonding
1359 configuration, it is necessary to run the initialization script, e.g.,::
1361 # /etc/init.d/boot.local
1365 # /etc/rc.d/rc.local
1367 It may be desirable in such a case to create a separate script
1368 which only initializes the bonding configuration, then call that
1369 separate script from within boot.local. This allows for bonding to be
1370 enabled without re-running the entire global init script.
1372 To shut down the bonding devices, it is necessary to first
1373 mark the bonding device itself as being down, then remove the
1374 appropriate device driver modules. For our example above, you can do
1377 # ifconfig bond0 down
1381 Again, for convenience, it may be desirable to create a script
1382 with these commands.
1385 3.3.1 Configuring Multiple Bonds Manually
1386 -----------------------------------------
1388 This section contains information on configuring multiple
1389 bonding devices with differing options for those systems whose network
1390 initialization scripts lack support for configuring multiple bonds.
1392 If you require multiple bonding devices, but all with the same
1393 options, you may wish to use the "max_bonds" module parameter,
1396 To create multiple bonding devices with differing options, it is
1397 preferable to use bonding parameters exported by sysfs, documented in the
1400 For versions of bonding without sysfs support, the only means to
1401 provide multiple instances of bonding with differing options is to load
1402 the bonding driver multiple times. Note that current versions of the
1403 sysconfig network initialization scripts handle this automatically; if
1404 your distro uses these scripts, no special action is needed. See the
1405 section Configuring Bonding Devices, above, if you're not sure about your
1406 network initialization scripts.
1408 To load multiple instances of the module, it is necessary to
1409 specify a different name for each instance (the module loading system
1410 requires that every loaded module, even multiple instances of the same
1411 module, have a unique name). This is accomplished by supplying multiple
1412 sets of bonding options in ``/etc/modprobe.d/*.conf``, for example::
1415 options bond0 -o bond0 mode=balance-rr miimon=100
1418 options bond1 -o bond1 mode=balance-alb miimon=50
1420 will load the bonding module two times. The first instance is
1421 named "bond0" and creates the bond0 device in balance-rr mode with an
1422 miimon of 100. The second instance is named "bond1" and creates the
1423 bond1 device in balance-alb mode with an miimon of 50.
1425 In some circumstances (typically with older distributions),
1426 the above does not work, and the second bonding instance never sees
1427 its options. In that case, the second options line can be substituted
1430 install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1431 mode=balance-alb miimon=50
1433 This may be repeated any number of times, specifying a new and
1434 unique name in place of bond1 for each subsequent instance.
1436 It has been observed that some Red Hat supplied kernels are unable
1437 to rename modules at load time (the "-o bond1" part). Attempts to pass
1438 that option to modprobe will produce an "Operation not permitted" error.
1439 This has been reported on some Fedora Core kernels, and has been seen on
1440 RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1441 to configure multiple bonds with differing parameters (as they are older
1442 kernels, and also lack sysfs support).
1444 3.4 Configuring Bonding Manually via Sysfs
1445 ------------------------------------------
1447 Starting with version 3.0.0, Channel Bonding may be configured
1448 via the sysfs interface. This interface allows dynamic configuration
1449 of all bonds in the system without unloading the module. It also
1450 allows for adding and removing bonds at runtime. Ifenslave is no
1451 longer required, though it is still supported.
1453 Use of the sysfs interface allows you to use multiple bonds
1454 with different configurations without having to reload the module.
1455 It also allows you to use multiple, differently configured bonds when
1456 bonding is compiled into the kernel.
1458 You must have the sysfs filesystem mounted to configure
1459 bonding this way. The examples in this document assume that you
1460 are using the standard mount point for sysfs, e.g. /sys. If your
1461 sysfs filesystem is mounted elsewhere, you will need to adjust the
1462 example paths accordingly.
1464 Creating and Destroying Bonds
1465 -----------------------------
1466 To add a new bond foo::
1468 # echo +foo > /sys/class/net/bonding_masters
1470 To remove an existing bond bar::
1472 # echo -bar > /sys/class/net/bonding_masters
1474 To show all existing bonds::
1476 # cat /sys/class/net/bonding_masters
1480 due to 4K size limitation of sysfs files, this list may be
1481 truncated if you have more than a few hundred bonds. This is unlikely
1482 to occur under normal operating conditions.
1484 Adding and Removing Slaves
1485 --------------------------
1486 Interfaces may be enslaved to a bond using the file
1487 /sys/class/net/<bond>/bonding/slaves. The semantics for this file
1488 are the same as for the bonding_masters file.
1490 To enslave interface eth0 to bond bond0::
1493 # echo +eth0 > /sys/class/net/bond0/bonding/slaves
1495 To free slave eth0 from bond bond0::
1497 # echo -eth0 > /sys/class/net/bond0/bonding/slaves
1499 When an interface is enslaved to a bond, symlinks between the
1500 two are created in the sysfs filesystem. In this case, you would get
1501 /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1502 /sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1504 This means that you can tell quickly whether or not an
1505 interface is enslaved by looking for the master symlink. Thus:
1506 # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1507 will free eth0 from whatever bond it is enslaved to, regardless of
1508 the name of the bond interface.
1510 Changing a Bond's Configuration
1511 -------------------------------
1512 Each bond may be configured individually by manipulating the
1513 files located in /sys/class/net/<bond name>/bonding
1515 The names of these files correspond directly with the command-
1516 line parameters described elsewhere in this file, and, with the
1517 exception of arp_ip_target, they accept the same values. To see the
1518 current setting, simply cat the appropriate file.
1520 A few examples will be given here; for specific usage
1521 guidelines for each parameter, see the appropriate section in this
1524 To configure bond0 for balance-alb mode::
1526 # ifconfig bond0 down
1527 # echo 6 > /sys/class/net/bond0/bonding/mode
1529 # echo balance-alb > /sys/class/net/bond0/bonding/mode
1533 The bond interface must be down before the mode can be changed.
1535 To enable MII monitoring on bond0 with a 1 second interval::
1537 # echo 1000 > /sys/class/net/bond0/bonding/miimon
1541 If ARP monitoring is enabled, it will disabled when MII
1542 monitoring is enabled, and vice-versa.
1544 To add ARP targets::
1546 # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1547 # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1551 up to 16 target addresses may be specified.
1553 To remove an ARP target::
1555 # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1557 To configure the interval between learning packet transmits::
1559 # echo 12 > /sys/class/net/bond0/bonding/lp_interval
1563 the lp_interval is the number of seconds between instances where
1564 the bonding driver sends learning packets to each slaves peer switch. The
1565 default interval is 1 second.
1567 Example Configuration
1568 ---------------------
1569 We begin with the same example that is shown in section 3.3,
1570 executed with sysfs, and without using ifenslave.
1572 To make a simple bond of two e100 devices (presumed to be eth0
1573 and eth1), and have it persist across reboots, edit the appropriate
1574 file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1579 echo balance-alb > /sys/class/net/bond0/bonding/mode
1580 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1581 echo 100 > /sys/class/net/bond0/bonding/miimon
1582 echo +eth0 > /sys/class/net/bond0/bonding/slaves
1583 echo +eth1 > /sys/class/net/bond0/bonding/slaves
1585 To add a second bond, with two e1000 interfaces in
1586 active-backup mode, using ARP monitoring, add the following lines to
1590 echo +bond1 > /sys/class/net/bonding_masters
1591 echo active-backup > /sys/class/net/bond1/bonding/mode
1592 ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1593 echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1594 echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1595 echo +eth2 > /sys/class/net/bond1/bonding/slaves
1596 echo +eth3 > /sys/class/net/bond1/bonding/slaves
1598 3.5 Configuration with Interfaces Support
1599 -----------------------------------------
1601 This section applies to distros which use /etc/network/interfaces file
1602 to describe network interface configuration, most notably Debian and it's
1605 The ifup and ifdown commands on Debian don't support bonding out of
1606 the box. The ifenslave-2.6 package should be installed to provide bonding
1607 support. Once installed, this package will provide ``bond-*`` options
1608 to be used into /etc/network/interfaces.
1610 Note that ifenslave-2.6 package will load the bonding module and use
1611 the ifenslave command when appropriate.
1613 Example Configurations
1614 ----------------------
1616 In /etc/network/interfaces, the following stanza will configure bond0, in
1617 active-backup mode, with eth0 and eth1 as slaves::
1620 iface bond0 inet dhcp
1621 bond-slaves eth0 eth1
1622 bond-mode active-backup
1624 bond-primary eth0 eth1
1626 If the above configuration doesn't work, you might have a system using
1627 upstart for system startup. This is most notably true for recent
1628 Ubuntu versions. The following stanza in /etc/network/interfaces will
1629 produce the same result on those systems::
1632 iface bond0 inet dhcp
1634 bond-mode active-backup
1638 iface eth0 inet manual
1640 bond-primary eth0 eth1
1643 iface eth1 inet manual
1645 bond-primary eth0 eth1
1647 For a full list of ``bond-*`` supported options in /etc/network/interfaces and
1648 some more advanced examples tailored to you particular distros, see the files in
1649 /usr/share/doc/ifenslave-2.6.
1651 3.6 Overriding Configuration for Special Cases
1652 ----------------------------------------------
1654 When using the bonding driver, the physical port which transmits a frame is
1655 typically selected by the bonding driver, and is not relevant to the user or
1656 system administrator. The output port is simply selected using the policies of
1657 the selected bonding mode. On occasion however, it is helpful to direct certain
1658 classes of traffic to certain physical interfaces on output to implement
1659 slightly more complex policies. For example, to reach a web server over a
1660 bonded interface in which eth0 connects to a private network, while eth1
1661 connects via a public network, it may be desirous to bias the bond to send said
1662 traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1663 can safely be sent over either interface. Such configurations may be achieved
1664 using the traffic control utilities inherent in linux.
1666 By default the bonding driver is multiqueue aware and 16 queues are created
1667 when the driver initializes (see Documentation/networking/multiqueue.rst
1668 for details). If more or less queues are desired the module parameter
1669 tx_queues can be used to change this value. There is no sysfs parameter
1670 available as the allocation is done at module init time.
1672 The output of the file /proc/net/bonding/bondX has changed so the output Queue
1673 ID is now printed for each slave::
1675 Bonding Mode: fault-tolerance (active-backup)
1677 Currently Active Slave: eth0
1679 MII Polling Interval (ms): 0
1683 Slave Interface: eth0
1685 Link Failure Count: 0
1686 Permanent HW addr: 00:1a:a0:12:8f:cb
1689 Slave Interface: eth1
1691 Link Failure Count: 0
1692 Permanent HW addr: 00:1a:a0:12:8f:cc
1695 The queue_id for a slave can be set using the command::
1697 # echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1699 Any interface that needs a queue_id set should set it with multiple calls
1700 like the one above until proper priorities are set for all interfaces. On
1701 distributions that allow configuration via initscripts, multiple 'queue_id'
1702 arguments can be added to BONDING_OPTS to set all needed slave queues.
1704 These queue id's can be used in conjunction with the tc utility to configure
1705 a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1706 slave devices. For instance, say we wanted, in the above configuration to
1707 force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1708 device. The following commands would accomplish this::
1710 # tc qdisc add dev bond0 handle 1 root multiq
1712 # tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip \
1713 dst 192.168.1.100 action skbedit queue_mapping 2
1715 These commands tell the kernel to attach a multiqueue queue discipline to the
1716 bond0 interface and filter traffic enqueued to it, such that packets with a dst
1717 ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1718 This value is then passed into the driver, causing the normal output path
1719 selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1721 Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver
1722 that normal output policy selection should take place. One benefit to simply
1723 leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1724 driver that is now present. This awareness allows tc filters to be placed on
1725 slave devices as well as bond devices and the bonding driver will simply act as
1726 a pass-through for selecting output queues on the slave device rather than
1727 output port selection.
1729 This feature first appeared in bonding driver version 3.7.0 and support for
1730 output slave selection was limited to round-robin and active-backup modes.
1732 3.7 Configuring LACP for 802.3ad mode in a more secure way
1733 ----------------------------------------------------------
1735 When using 802.3ad bonding mode, the Actor (host) and Partner (switch)
1736 exchange LACPDUs. These LACPDUs cannot be sniffed, because they are
1737 destined to link local mac addresses (which switches/bridges are not
1738 supposed to forward). However, most of the values are easily predictable
1739 or are simply the machine's MAC address (which is trivially known to all
1740 other hosts in the same L2). This implies that other machines in the L2
1741 domain can spoof LACPDU packets from other hosts to the switch and potentially
1742 cause mayhem by joining (from the point of view of the switch) another
1743 machine's aggregate, thus receiving a portion of that hosts incoming
1744 traffic and / or spoofing traffic from that machine themselves (potentially
1745 even successfully terminating some portion of flows). Though this is not
1746 a likely scenario, one could avoid this possibility by simply configuring
1747 few bonding parameters:
1749 (a) ad_actor_system : You can set a random mac-address that can be used for
1750 these LACPDU exchanges. The value can not be either NULL or Multicast.
1751 Also it's preferable to set the local-admin bit. Following shell code
1752 generates a random mac-address as described above::
1754 # sys_mac_addr=$(printf '%02x:%02x:%02x:%02x:%02x:%02x' \
1755 $(( (RANDOM & 0xFE) | 0x02 )) \
1756 $(( RANDOM & 0xFF )) \
1757 $(( RANDOM & 0xFF )) \
1758 $(( RANDOM & 0xFF )) \
1759 $(( RANDOM & 0xFF )) \
1760 $(( RANDOM & 0xFF )))
1761 # echo $sys_mac_addr > /sys/class/net/bond0/bonding/ad_actor_system
1763 (b) ad_actor_sys_prio : Randomize the system priority. The default value
1764 is 65535, but system can take the value from 1 - 65535. Following shell
1765 code generates random priority and sets it::
1767 # sys_prio=$(( 1 + RANDOM + RANDOM ))
1768 # echo $sys_prio > /sys/class/net/bond0/bonding/ad_actor_sys_prio
1770 (c) ad_user_port_key : Use the user portion of the port-key. The default
1771 keeps this empty. These are the upper 10 bits of the port-key and value
1772 ranges from 0 - 1023. Following shell code generates these 10 bits and
1775 # usr_port_key=$(( RANDOM & 0x3FF ))
1776 # echo $usr_port_key > /sys/class/net/bond0/bonding/ad_user_port_key
1779 4 Querying Bonding Configuration
1780 =================================
1782 4.1 Bonding Configuration
1783 -------------------------
1785 Each bonding device has a read-only file residing in the
1786 /proc/net/bonding directory. The file contents include information
1787 about the bonding configuration, options and state of each slave.
1789 For example, the contents of /proc/net/bonding/bond0 after the
1790 driver is loaded with parameters of mode=0 and miimon=1000 is
1791 generally as follows::
1793 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1794 Bonding Mode: load balancing (round-robin)
1795 Currently Active Slave: eth0
1797 MII Polling Interval (ms): 1000
1801 Slave Interface: eth1
1803 Link Failure Count: 1
1805 Slave Interface: eth0
1807 Link Failure Count: 1
1809 The precise format and contents will change depending upon the
1810 bonding configuration, state, and version of the bonding driver.
1812 4.2 Network configuration
1813 -------------------------
1815 The network configuration can be inspected using the ifconfig
1816 command. Bonding devices will have the MASTER flag set; Bonding slave
1817 devices will have the SLAVE flag set. The ifconfig output does not
1818 contain information on which slaves are associated with which masters.
1820 In the example below, the bond0 interface is the master
1821 (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1822 bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1823 TLB and ALB that require a unique MAC address for each slave::
1826 bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1827 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1828 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1829 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1830 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1831 collisions:0 txqueuelen:0
1833 eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1834 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1835 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1836 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1837 collisions:0 txqueuelen:100
1838 Interrupt:10 Base address:0x1080
1840 eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1841 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1842 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1843 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1844 collisions:0 txqueuelen:100
1845 Interrupt:9 Base address:0x1400
1847 5. Switch Configuration
1848 =======================
1850 For this section, "switch" refers to whatever system the
1851 bonded devices are directly connected to (i.e., where the other end of
1852 the cable plugs into). This may be an actual dedicated switch device,
1853 or it may be another regular system (e.g., another computer running
1856 The active-backup, balance-tlb and balance-alb modes do not
1857 require any specific configuration of the switch.
1859 The 802.3ad mode requires that the switch have the appropriate
1860 ports configured as an 802.3ad aggregation. The precise method used
1861 to configure this varies from switch to switch, but, for example, a
1862 Cisco 3550 series switch requires that the appropriate ports first be
1863 grouped together in a single etherchannel instance, then that
1864 etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1865 standard EtherChannel).
1867 The balance-rr, balance-xor and broadcast modes generally
1868 require that the switch have the appropriate ports grouped together.
1869 The nomenclature for such a group differs between switches, it may be
1870 called an "etherchannel" (as in the Cisco example, above), a "trunk
1871 group" or some other similar variation. For these modes, each switch
1872 will also have its own configuration options for the switch's transmit
1873 policy to the bond. Typical choices include XOR of either the MAC or
1874 IP addresses. The transmit policy of the two peers does not need to
1875 match. For these three modes, the bonding mode really selects a
1876 transmit policy for an EtherChannel group; all three will interoperate
1877 with another EtherChannel group.
1880 6. 802.1q VLAN Support
1881 ======================
1883 It is possible to configure VLAN devices over a bond interface
1884 using the 8021q driver. However, only packets coming from the 8021q
1885 driver and passing through bonding will be tagged by default. Self
1886 generated packets, for example, bonding's learning packets or ARP
1887 packets generated by either ALB mode or the ARP monitor mechanism, are
1888 tagged internally by bonding itself. As a result, bonding must
1889 "learn" the VLAN IDs configured above it, and use those IDs to tag
1890 self generated packets.
1892 For reasons of simplicity, and to support the use of adapters
1893 that can do VLAN hardware acceleration offloading, the bonding
1894 interface declares itself as fully hardware offloading capable, it gets
1895 the add_vid/kill_vid notifications to gather the necessary
1896 information, and it propagates those actions to the slaves. In case
1897 of mixed adapter types, hardware accelerated tagged packets that
1898 should go through an adapter that is not offloading capable are
1899 "un-accelerated" by the bonding driver so the VLAN tag sits in the
1902 VLAN interfaces *must* be added on top of a bonding interface
1903 only after enslaving at least one slave. The bonding interface has a
1904 hardware address of 00:00:00:00:00:00 until the first slave is added.
1905 If the VLAN interface is created prior to the first enslavement, it
1906 would pick up the all-zeroes hardware address. Once the first slave
1907 is attached to the bond, the bond device itself will pick up the
1908 slave's hardware address, which is then available for the VLAN device.
1910 Also, be aware that a similar problem can occur if all slaves
1911 are released from a bond that still has one or more VLAN interfaces on
1912 top of it. When a new slave is added, the bonding interface will
1913 obtain its hardware address from the first slave, which might not
1914 match the hardware address of the VLAN interfaces (which was
1915 ultimately copied from an earlier slave).
1917 There are two methods to insure that the VLAN device operates
1918 with the correct hardware address if all slaves are removed from a
1921 1. Remove all VLAN interfaces then recreate them
1923 2. Set the bonding interface's hardware address so that it
1924 matches the hardware address of the VLAN interfaces.
1926 Note that changing a VLAN interface's HW address would set the
1927 underlying device -- i.e. the bonding interface -- to promiscuous
1928 mode, which might not be what you want.
1934 The bonding driver at present supports two schemes for
1935 monitoring a slave device's link state: the ARP monitor and the MII
1938 At the present time, due to implementation restrictions in the
1939 bonding driver itself, it is not possible to enable both ARP and MII
1940 monitoring simultaneously.
1942 7.1 ARP Monitor Operation
1943 -------------------------
1945 The ARP monitor operates as its name suggests: it sends ARP
1946 queries to one or more designated peer systems on the network, and
1947 uses the response as an indication that the link is operating. This
1948 gives some assurance that traffic is actually flowing to and from one
1949 or more peers on the local network.
1951 The ARP monitor relies on the device driver itself to verify
1952 that traffic is flowing. In particular, the driver must keep up to
1953 date the last receive time, dev->last_rx. Drivers that use NETIF_F_LLTX
1954 flag must also update netdev_queue->trans_start. If they do not, then the
1955 ARP monitor will immediately fail any slaves using that driver, and
1956 those slaves will stay down. If networking monitoring (tcpdump, etc)
1957 shows the ARP requests and replies on the network, then it may be that
1958 your device driver is not updating last_rx and trans_start.
1960 7.2 Configuring Multiple ARP Targets
1961 ------------------------------------
1963 While ARP monitoring can be done with just one target, it can
1964 be useful in a High Availability setup to have several targets to
1965 monitor. In the case of just one target, the target itself may go
1966 down or have a problem making it unresponsive to ARP requests. Having
1967 an additional target (or several) increases the reliability of the ARP
1970 Multiple ARP targets must be separated by commas as follows::
1972 # example options for ARP monitoring with three targets
1974 options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1976 For just a single target the options would resemble::
1978 # example options for ARP monitoring with one target
1980 options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1983 7.3 MII Monitor Operation
1984 -------------------------
1986 The MII monitor monitors only the carrier state of the local
1987 network interface. It accomplishes this in one of three ways: by
1988 depending upon the device driver to maintain its carrier state, by
1989 querying the device's MII registers, or by making an ethtool query to
1992 If the use_carrier module parameter is 1 (the default value),
1993 then the MII monitor will rely on the driver for carrier state
1994 information (via the netif_carrier subsystem). As explained in the
1995 use_carrier parameter information, above, if the MII monitor fails to
1996 detect carrier loss on the device (e.g., when the cable is physically
1997 disconnected), it may be that the driver does not support
2000 If use_carrier is 0, then the MII monitor will first query the
2001 device's (via ioctl) MII registers and check the link state. If that
2002 request fails (not just that it returns carrier down), then the MII
2003 monitor will make an ethtool ETHTOOL_GLINK request to attempt to obtain
2004 the same information. If both methods fail (i.e., the driver either
2005 does not support or had some error in processing both the MII register
2006 and ethtool requests), then the MII monitor will assume the link is
2009 8. Potential Sources of Trouble
2010 ===============================
2012 8.1 Adventures in Routing
2013 -------------------------
2015 When bonding is configured, it is important that the slave
2016 devices not have routes that supersede routes of the master (or,
2017 generally, not have routes at all). For example, suppose the bonding
2018 device bond0 has two slaves, eth0 and eth1, and the routing table is
2021 Kernel IP routing table
2022 Destination Gateway Genmask Flags MSS Window irtt Iface
2023 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
2024 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
2025 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
2026 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
2028 This routing configuration will likely still update the
2029 receive/transmit times in the driver (needed by the ARP monitor), but
2030 may bypass the bonding driver (because outgoing traffic to, in this
2031 case, another host on network 10 would use eth0 or eth1 before bond0).
2033 The ARP monitor (and ARP itself) may become confused by this
2034 configuration, because ARP requests (generated by the ARP monitor)
2035 will be sent on one interface (bond0), but the corresponding reply
2036 will arrive on a different interface (eth0). This reply looks to ARP
2037 as an unsolicited ARP reply (because ARP matches replies on an
2038 interface basis), and is discarded. The MII monitor is not affected
2039 by the state of the routing table.
2041 The solution here is simply to insure that slaves do not have
2042 routes of their own, and if for some reason they must, those routes do
2043 not supersede routes of their master. This should generally be the
2044 case, but unusual configurations or errant manual or automatic static
2045 route additions may cause trouble.
2047 8.2 Ethernet Device Renaming
2048 ----------------------------
2050 On systems with network configuration scripts that do not
2051 associate physical devices directly with network interface names (so
2052 that the same physical device always has the same "ethX" name), it may
2053 be necessary to add some special logic to config files in
2056 For example, given a modules.conf containing the following::
2059 options bond0 mode=some-mode miimon=50
2065 If neither eth0 and eth1 are slaves to bond0, then when the
2066 bond0 interface comes up, the devices may end up reordered. This
2067 happens because bonding is loaded first, then its slave device's
2068 drivers are loaded next. Since no other drivers have been loaded,
2069 when the e1000 driver loads, it will receive eth0 and eth1 for its
2070 devices, but the bonding configuration tries to enslave eth2 and eth3
2071 (which may later be assigned to the tg3 devices).
2073 Adding the following::
2075 add above bonding e1000 tg3
2077 causes modprobe to load e1000 then tg3, in that order, when
2078 bonding is loaded. This command is fully documented in the
2079 modules.conf manual page.
2081 On systems utilizing modprobe an equivalent problem can occur.
2082 In this case, the following can be added to config files in
2083 /etc/modprobe.d/ as::
2085 softdep bonding pre: tg3 e1000
2087 This will load tg3 and e1000 modules before loading the bonding one.
2088 Full documentation on this can be found in the modprobe.d and modprobe
2091 8.3. Painfully Slow Or No Failed Link Detection By Miimon
2092 ---------------------------------------------------------
2094 By default, bonding enables the use_carrier option, which
2095 instructs bonding to trust the driver to maintain carrier state.
2097 As discussed in the options section, above, some drivers do
2098 not support the netif_carrier_on/_off link state tracking system.
2099 With use_carrier enabled, bonding will always see these links as up,
2100 regardless of their actual state.
2102 Additionally, other drivers do support netif_carrier, but do
2103 not maintain it in real time, e.g., only polling the link state at
2104 some fixed interval. In this case, miimon will detect failures, but
2105 only after some long period of time has expired. If it appears that
2106 miimon is very slow in detecting link failures, try specifying
2107 use_carrier=0 to see if that improves the failure detection time. If
2108 it does, then it may be that the driver checks the carrier state at a
2109 fixed interval, but does not cache the MII register values (so the
2110 use_carrier=0 method of querying the registers directly works). If
2111 use_carrier=0 does not improve the failover, then the driver may cache
2112 the registers, or the problem may be elsewhere.
2114 Also, remember that miimon only checks for the device's
2115 carrier state. It has no way to determine the state of devices on or
2116 beyond other ports of a switch, or if a switch is refusing to pass
2117 traffic while still maintaining carrier on.
2122 If running SNMP agents, the bonding driver should be loaded
2123 before any network drivers participating in a bond. This requirement
2124 is due to the interface index (ipAdEntIfIndex) being associated to
2125 the first interface found with a given IP address. That is, there is
2126 only one ipAdEntIfIndex for each IP address. For example, if eth0 and
2127 eth1 are slaves of bond0 and the driver for eth0 is loaded before the
2128 bonding driver, the interface for the IP address will be associated
2129 with the eth0 interface. This configuration is shown below, the IP
2130 address 192.168.1.1 has an interface index of 2 which indexes to eth0
2131 in the ifDescr table (ifDescr.2).
2135 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2136 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
2137 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
2138 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
2139 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
2140 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
2141 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
2142 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2143 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
2144 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2146 This problem is avoided by loading the bonding driver before
2147 any network drivers participating in a bond. Below is an example of
2148 loading the bonding driver first, the IP address 192.168.1.1 is
2149 correctly associated with ifDescr.2.
2151 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2152 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
2153 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
2154 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
2155 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
2156 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
2157 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
2158 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2159 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
2160 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2162 While some distributions may not report the interface name in
2163 ifDescr, the association between the IP address and IfIndex remains
2164 and SNMP functions such as Interface_Scan_Next will report that
2167 10. Promiscuous mode
2168 ====================
2170 When running network monitoring tools, e.g., tcpdump, it is
2171 common to enable promiscuous mode on the device, so that all traffic
2172 is seen (instead of seeing only traffic destined for the local host).
2173 The bonding driver handles promiscuous mode changes to the bonding
2174 master device (e.g., bond0), and propagates the setting to the slave
2177 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
2178 the promiscuous mode setting is propagated to all slaves.
2180 For the active-backup, balance-tlb and balance-alb modes, the
2181 promiscuous mode setting is propagated only to the active slave.
2183 For balance-tlb mode, the active slave is the slave currently
2184 receiving inbound traffic.
2186 For balance-alb mode, the active slave is the slave used as a
2187 "primary." This slave is used for mode-specific control traffic, for
2188 sending to peers that are unassigned or if the load is unbalanced.
2190 For the active-backup, balance-tlb and balance-alb modes, when
2191 the active slave changes (e.g., due to a link failure), the
2192 promiscuous setting will be propagated to the new active slave.
2194 11. Configuring Bonding for High Availability
2195 =============================================
2197 High Availability refers to configurations that provide
2198 maximum network availability by having redundant or backup devices,
2199 links or switches between the host and the rest of the world. The
2200 goal is to provide the maximum availability of network connectivity
2201 (i.e., the network always works), even though other configurations
2202 could provide higher throughput.
2204 11.1 High Availability in a Single Switch Topology
2205 --------------------------------------------------
2207 If two hosts (or a host and a single switch) are directly
2208 connected via multiple physical links, then there is no availability
2209 penalty to optimizing for maximum bandwidth. In this case, there is
2210 only one switch (or peer), so if it fails, there is no alternative
2211 access to fail over to. Additionally, the bonding load balance modes
2212 support link monitoring of their members, so if individual links fail,
2213 the load will be rebalanced across the remaining devices.
2215 See Section 12, "Configuring Bonding for Maximum Throughput"
2216 for information on configuring bonding with one peer device.
2218 11.2 High Availability in a Multiple Switch Topology
2219 ----------------------------------------------------
2221 With multiple switches, the configuration of bonding and the
2222 network changes dramatically. In multiple switch topologies, there is
2223 a trade off between network availability and usable bandwidth.
2225 Below is a sample network, configured to maximize the
2226 availability of the network::
2230 +-----+----+ +-----+----+
2231 | |port2 ISL port2| |
2232 | switch A +--------------------------+ switch B |
2234 +-----+----+ +-----++---+
2237 +-------------+ host1 +---------------+
2240 In this configuration, there is a link between the two
2241 switches (ISL, or inter switch link), and multiple ports connecting to
2242 the outside world ("port3" on each switch). There is no technical
2243 reason that this could not be extended to a third switch.
2245 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
2246 -------------------------------------------------------------
2248 In a topology such as the example above, the active-backup and
2249 broadcast modes are the only useful bonding modes when optimizing for
2250 availability; the other modes require all links to terminate on the
2251 same peer for them to behave rationally.
2254 This is generally the preferred mode, particularly if
2255 the switches have an ISL and play together well. If the
2256 network configuration is such that one switch is specifically
2257 a backup switch (e.g., has lower capacity, higher cost, etc),
2258 then the primary option can be used to insure that the
2259 preferred link is always used when it is available.
2262 This mode is really a special purpose mode, and is suitable
2263 only for very specific needs. For example, if the two
2264 switches are not connected (no ISL), and the networks beyond
2265 them are totally independent. In this case, if it is
2266 necessary for some specific one-way traffic to reach both
2267 independent networks, then the broadcast mode may be suitable.
2269 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
2270 ----------------------------------------------------------------
2272 The choice of link monitoring ultimately depends upon your
2273 switch. If the switch can reliably fail ports in response to other
2274 failures, then either the MII or ARP monitors should work. For
2275 example, in the above example, if the "port3" link fails at the remote
2276 end, the MII monitor has no direct means to detect this. The ARP
2277 monitor could be configured with a target at the remote end of port3,
2278 thus detecting that failure without switch support.
2280 In general, however, in a multiple switch topology, the ARP
2281 monitor can provide a higher level of reliability in detecting end to
2282 end connectivity failures (which may be caused by the failure of any
2283 individual component to pass traffic for any reason). Additionally,
2284 the ARP monitor should be configured with multiple targets (at least
2285 one for each switch in the network). This will insure that,
2286 regardless of which switch is active, the ARP monitor has a suitable
2289 Note, also, that of late many switches now support a functionality
2290 generally referred to as "trunk failover." This is a feature of the
2291 switch that causes the link state of a particular switch port to be set
2292 down (or up) when the state of another switch port goes down (or up).
2293 Its purpose is to propagate link failures from logically "exterior" ports
2294 to the logically "interior" ports that bonding is able to monitor via
2295 miimon. Availability and configuration for trunk failover varies by
2296 switch, but this can be a viable alternative to the ARP monitor when using
2299 12. Configuring Bonding for Maximum Throughput
2300 ==============================================
2302 12.1 Maximizing Throughput in a Single Switch Topology
2303 ------------------------------------------------------
2305 In a single switch configuration, the best method to maximize
2306 throughput depends upon the application and network environment. The
2307 various load balancing modes each have strengths and weaknesses in
2308 different environments, as detailed below.
2310 For this discussion, we will break down the topologies into
2311 two categories. Depending upon the destination of most traffic, we
2312 categorize them into either "gatewayed" or "local" configurations.
2314 In a gatewayed configuration, the "switch" is acting primarily
2315 as a router, and the majority of traffic passes through this router to
2316 other networks. An example would be the following::
2319 +----------+ +----------+
2320 | |eth0 port1| | to other networks
2321 | Host A +---------------------+ router +------------------->
2322 | +---------------------+ | Hosts B and C are out
2323 | |eth1 port2| | here somewhere
2324 +----------+ +----------+
2326 The router may be a dedicated router device, or another host
2327 acting as a gateway. For our discussion, the important point is that
2328 the majority of traffic from Host A will pass through the router to
2329 some other network before reaching its final destination.
2331 In a gatewayed network configuration, although Host A may
2332 communicate with many other systems, all of its traffic will be sent
2333 and received via one other peer on the local network, the router.
2335 Note that the case of two systems connected directly via
2336 multiple physical links is, for purposes of configuring bonding, the
2337 same as a gatewayed configuration. In that case, it happens that all
2338 traffic is destined for the "gateway" itself, not some other network
2341 In a local configuration, the "switch" is acting primarily as
2342 a switch, and the majority of traffic passes through this switch to
2343 reach other stations on the same network. An example would be the
2346 +----------+ +----------+ +--------+
2347 | |eth0 port1| +-------+ Host B |
2348 | Host A +------------+ switch |port3 +--------+
2349 | +------------+ | +--------+
2350 | |eth1 port2| +------------------+ Host C |
2351 +----------+ +----------+port4 +--------+
2354 Again, the switch may be a dedicated switch device, or another
2355 host acting as a gateway. For our discussion, the important point is
2356 that the majority of traffic from Host A is destined for other hosts
2357 on the same local network (Hosts B and C in the above example).
2359 In summary, in a gatewayed configuration, traffic to and from
2360 the bonded device will be to the same MAC level peer on the network
2361 (the gateway itself, i.e., the router), regardless of its final
2362 destination. In a local configuration, traffic flows directly to and
2363 from the final destinations, thus, each destination (Host B, Host C)
2364 will be addressed directly by their individual MAC addresses.
2366 This distinction between a gatewayed and a local network
2367 configuration is important because many of the load balancing modes
2368 available use the MAC addresses of the local network source and
2369 destination to make load balancing decisions. The behavior of each
2370 mode is described below.
2373 12.1.1 MT Bonding Mode Selection for Single Switch Topology
2374 -----------------------------------------------------------
2376 This configuration is the easiest to set up and to understand,
2377 although you will have to decide which bonding mode best suits your
2378 needs. The trade offs for each mode are detailed below:
2381 This mode is the only mode that will permit a single
2382 TCP/IP connection to stripe traffic across multiple
2383 interfaces. It is therefore the only mode that will allow a
2384 single TCP/IP stream to utilize more than one interface's
2385 worth of throughput. This comes at a cost, however: the
2386 striping generally results in peer systems receiving packets out
2387 of order, causing TCP/IP's congestion control system to kick
2388 in, often by retransmitting segments.
2390 It is possible to adjust TCP/IP's congestion limits by
2391 altering the net.ipv4.tcp_reordering sysctl parameter. The
2392 usual default value is 3. But keep in mind TCP stack is able
2393 to automatically increase this when it detects reorders.
2395 Note that the fraction of packets that will be delivered out of
2396 order is highly variable, and is unlikely to be zero. The level
2397 of reordering depends upon a variety of factors, including the
2398 networking interfaces, the switch, and the topology of the
2399 configuration. Speaking in general terms, higher speed network
2400 cards produce more reordering (due to factors such as packet
2401 coalescing), and a "many to many" topology will reorder at a
2402 higher rate than a "many slow to one fast" configuration.
2404 Many switches do not support any modes that stripe traffic
2405 (instead choosing a port based upon IP or MAC level addresses);
2406 for those devices, traffic for a particular connection flowing
2407 through the switch to a balance-rr bond will not utilize greater
2408 than one interface's worth of bandwidth.
2410 If you are utilizing protocols other than TCP/IP, UDP for
2411 example, and your application can tolerate out of order
2412 delivery, then this mode can allow for single stream datagram
2413 performance that scales near linearly as interfaces are added
2416 This mode requires the switch to have the appropriate ports
2417 configured for "etherchannel" or "trunking."
2420 There is not much advantage in this network topology to
2421 the active-backup mode, as the inactive backup devices are all
2422 connected to the same peer as the primary. In this case, a
2423 load balancing mode (with link monitoring) will provide the
2424 same level of network availability, but with increased
2425 available bandwidth. On the plus side, active-backup mode
2426 does not require any configuration of the switch, so it may
2427 have value if the hardware available does not support any of
2428 the load balance modes.
2431 This mode will limit traffic such that packets destined
2432 for specific peers will always be sent over the same
2433 interface. Since the destination is determined by the MAC
2434 addresses involved, this mode works best in a "local" network
2435 configuration (as described above), with destinations all on
2436 the same local network. This mode is likely to be suboptimal
2437 if all your traffic is passed through a single router (i.e., a
2438 "gatewayed" network configuration, as described above).
2440 As with balance-rr, the switch ports need to be configured for
2441 "etherchannel" or "trunking."
2444 Like active-backup, there is not much advantage to this
2445 mode in this type of network topology.
2448 This mode can be a good choice for this type of network
2449 topology. The 802.3ad mode is an IEEE standard, so all peers
2450 that implement 802.3ad should interoperate well. The 802.3ad
2451 protocol includes automatic configuration of the aggregates,
2452 so minimal manual configuration of the switch is needed
2453 (typically only to designate that some set of devices is
2454 available for 802.3ad). The 802.3ad standard also mandates
2455 that frames be delivered in order (within certain limits), so
2456 in general single connections will not see misordering of
2457 packets. The 802.3ad mode does have some drawbacks: the
2458 standard mandates that all devices in the aggregate operate at
2459 the same speed and duplex. Also, as with all bonding load
2460 balance modes other than balance-rr, no single connection will
2461 be able to utilize more than a single interface's worth of
2464 Additionally, the linux bonding 802.3ad implementation
2465 distributes traffic by peer (using an XOR of MAC addresses
2466 and packet type ID), so in a "gatewayed" configuration, all
2467 outgoing traffic will generally use the same device. Incoming
2468 traffic may also end up on a single device, but that is
2469 dependent upon the balancing policy of the peer's 802.3ad
2470 implementation. In a "local" configuration, traffic will be
2471 distributed across the devices in the bond.
2473 Finally, the 802.3ad mode mandates the use of the MII monitor,
2474 therefore, the ARP monitor is not available in this mode.
2477 The balance-tlb mode balances outgoing traffic by peer.
2478 Since the balancing is done according to MAC address, in a
2479 "gatewayed" configuration (as described above), this mode will
2480 send all traffic across a single device. However, in a
2481 "local" network configuration, this mode balances multiple
2482 local network peers across devices in a vaguely intelligent
2483 manner (not a simple XOR as in balance-xor or 802.3ad mode),
2484 so that mathematically unlucky MAC addresses (i.e., ones that
2485 XOR to the same value) will not all "bunch up" on a single
2488 Unlike 802.3ad, interfaces may be of differing speeds, and no
2489 special switch configuration is required. On the down side,
2490 in this mode all incoming traffic arrives over a single
2491 interface, this mode requires certain ethtool support in the
2492 network device driver of the slave interfaces, and the ARP
2493 monitor is not available.
2496 This mode is everything that balance-tlb is, and more.
2497 It has all of the features (and restrictions) of balance-tlb,
2498 and will also balance incoming traffic from local network
2499 peers (as described in the Bonding Module Options section,
2502 The only additional down side to this mode is that the network
2503 device driver must support changing the hardware address while
2506 12.1.2 MT Link Monitoring for Single Switch Topology
2507 ----------------------------------------------------
2509 The choice of link monitoring may largely depend upon which
2510 mode you choose to use. The more advanced load balancing modes do not
2511 support the use of the ARP monitor, and are thus restricted to using
2512 the MII monitor (which does not provide as high a level of end to end
2513 assurance as the ARP monitor).
2515 12.2 Maximum Throughput in a Multiple Switch Topology
2516 -----------------------------------------------------
2518 Multiple switches may be utilized to optimize for throughput
2519 when they are configured in parallel as part of an isolated network
2520 between two or more systems, for example::
2526 +--------+ | +---------+
2528 +------+---+ +-----+----+ +-----+----+
2529 | Switch A | | Switch B | | Switch C |
2530 +------+---+ +-----+----+ +-----+----+
2532 +--------+ | +---------+
2538 In this configuration, the switches are isolated from one
2539 another. One reason to employ a topology such as this is for an
2540 isolated network with many hosts (a cluster configured for high
2541 performance, for example), using multiple smaller switches can be more
2542 cost effective than a single larger switch, e.g., on a network with 24
2543 hosts, three 24 port switches can be significantly less expensive than
2544 a single 72 port switch.
2546 If access beyond the network is required, an individual host
2547 can be equipped with an additional network device connected to an
2548 external network; this host then additionally acts as a gateway.
2550 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2551 -------------------------------------------------------------
2553 In actual practice, the bonding mode typically employed in
2554 configurations of this type is balance-rr. Historically, in this
2555 network configuration, the usual caveats about out of order packet
2556 delivery are mitigated by the use of network adapters that do not do
2557 any kind of packet coalescing (via the use of NAPI, or because the
2558 device itself does not generate interrupts until some number of
2559 packets has arrived). When employed in this fashion, the balance-rr
2560 mode allows individual connections between two hosts to effectively
2561 utilize greater than one interface's bandwidth.
2563 12.2.2 MT Link Monitoring for Multiple Switch Topology
2564 ------------------------------------------------------
2566 Again, in actual practice, the MII monitor is most often used
2567 in this configuration, as performance is given preference over
2568 availability. The ARP monitor will function in this topology, but its
2569 advantages over the MII monitor are mitigated by the volume of probes
2570 needed as the number of systems involved grows (remember that each
2571 host in the network is configured with bonding).
2573 13. Switch Behavior Issues
2574 ==========================
2576 13.1 Link Establishment and Failover Delays
2577 -------------------------------------------
2579 Some switches exhibit undesirable behavior with regard to the
2580 timing of link up and down reporting by the switch.
2582 First, when a link comes up, some switches may indicate that
2583 the link is up (carrier available), but not pass traffic over the
2584 interface for some period of time. This delay is typically due to
2585 some type of autonegotiation or routing protocol, but may also occur
2586 during switch initialization (e.g., during recovery after a switch
2587 failure). If you find this to be a problem, specify an appropriate
2588 value to the updelay bonding module option to delay the use of the
2589 relevant interface(s).
2591 Second, some switches may "bounce" the link state one or more
2592 times while a link is changing state. This occurs most commonly while
2593 the switch is initializing. Again, an appropriate updelay value may
2596 Note that when a bonding interface has no active links, the
2597 driver will immediately reuse the first link that goes up, even if the
2598 updelay parameter has been specified (the updelay is ignored in this
2599 case). If there are slave interfaces waiting for the updelay timeout
2600 to expire, the interface that first went into that state will be
2601 immediately reused. This reduces down time of the network if the
2602 value of updelay has been overestimated, and since this occurs only in
2603 cases with no connectivity, there is no additional penalty for
2604 ignoring the updelay.
2606 In addition to the concerns about switch timings, if your
2607 switches take a long time to go into backup mode, it may be desirable
2608 to not activate a backup interface immediately after a link goes down.
2609 Failover may be delayed via the downdelay bonding module option.
2611 13.2 Duplicated Incoming Packets
2612 --------------------------------
2614 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2615 suppress duplicate packets, which should largely eliminate this problem.
2616 The following description is kept for reference.
2618 It is not uncommon to observe a short burst of duplicated
2619 traffic when the bonding device is first used, or after it has been
2620 idle for some period of time. This is most easily observed by issuing
2621 a "ping" to some other host on the network, and noticing that the
2622 output from ping flags duplicates (typically one per slave).
2624 For example, on a bond in active-backup mode with five slaves
2625 all connected to one switch, the output may appear as follows::
2628 PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
2629 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
2630 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2631 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2632 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2633 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2634 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
2635 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
2636 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2638 This is not due to an error in the bonding driver, rather, it
2639 is a side effect of how many switches update their MAC forwarding
2640 tables. Initially, the switch does not associate the MAC address in
2641 the packet with a particular switch port, and so it may send the
2642 traffic to all ports until its MAC forwarding table is updated. Since
2643 the interfaces attached to the bond may occupy multiple ports on a
2644 single switch, when the switch (temporarily) floods the traffic to all
2645 ports, the bond device receives multiple copies of the same packet
2646 (one per slave device).
2648 The duplicated packet behavior is switch dependent, some
2649 switches exhibit this, and some do not. On switches that display this
2650 behavior, it can be induced by clearing the MAC forwarding table (on
2651 most Cisco switches, the privileged command "clear mac address-table
2652 dynamic" will accomplish this).
2654 14. Hardware Specific Considerations
2655 ====================================
2657 This section contains additional information for configuring
2658 bonding on specific hardware platforms, or for interfacing bonding
2659 with particular switches or other devices.
2661 14.1 IBM BladeCenter
2662 --------------------
2664 This applies to the JS20 and similar systems.
2666 On the JS20 blades, the bonding driver supports only
2667 balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2668 largely due to the network topology inside the BladeCenter, detailed
2671 JS20 network adapter information
2672 --------------------------------
2674 All JS20s come with two Broadcom Gigabit Ethernet ports
2675 integrated on the planar (that's "motherboard" in IBM-speak). In the
2676 BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2677 I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2678 An add-on Broadcom daughter card can be installed on a JS20 to provide
2679 two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2680 wired to I/O Modules 3 and 4, respectively.
2682 Each I/O Module may contain either a switch or a passthrough
2683 module (which allows ports to be directly connected to an external
2684 switch). Some bonding modes require a specific BladeCenter internal
2685 network topology in order to function; these are detailed below.
2687 Additional BladeCenter-specific networking information can be
2688 found in two IBM Redbooks (www.ibm.com/redbooks):
2690 - "IBM eServer BladeCenter Networking Options"
2691 - "IBM eServer BladeCenter Layer 2-7 Network Switching"
2693 BladeCenter networking configuration
2694 ------------------------------------
2696 Because a BladeCenter can be configured in a very large number
2697 of ways, this discussion will be confined to describing basic
2700 Normally, Ethernet Switch Modules (ESMs) are used in I/O
2701 modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2702 JS20 will be connected to different internal switches (in the
2703 respective I/O modules).
2705 A passthrough module (OPM or CPM, optical or copper,
2706 passthrough module) connects the I/O module directly to an external
2707 switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2708 interfaces of a JS20 can be redirected to the outside world and
2709 connected to a common external switch.
2711 Depending upon the mix of ESMs and PMs, the network will
2712 appear to bonding as either a single switch topology (all PMs) or as a
2713 multiple switch topology (one or more ESMs, zero or more PMs). It is
2714 also possible to connect ESMs together, resulting in a configuration
2715 much like the example in "High Availability in a Multiple Switch
2718 Requirements for specific modes
2719 -------------------------------
2721 The balance-rr mode requires the use of passthrough modules
2722 for devices in the bond, all connected to an common external switch.
2723 That switch must be configured for "etherchannel" or "trunking" on the
2724 appropriate ports, as is usual for balance-rr.
2726 The balance-alb and balance-tlb modes will function with
2727 either switch modules or passthrough modules (or a mix). The only
2728 specific requirement for these modes is that all network interfaces
2729 must be able to reach all destinations for traffic sent over the
2730 bonding device (i.e., the network must converge at some point outside
2733 The active-backup mode has no additional requirements.
2735 Link monitoring issues
2736 ----------------------
2738 When an Ethernet Switch Module is in place, only the ARP
2739 monitor will reliably detect link loss to an external switch. This is
2740 nothing unusual, but examination of the BladeCenter cabinet would
2741 suggest that the "external" network ports are the ethernet ports for
2742 the system, when it fact there is a switch between these "external"
2743 ports and the devices on the JS20 system itself. The MII monitor is
2744 only able to detect link failures between the ESM and the JS20 system.
2746 When a passthrough module is in place, the MII monitor does
2747 detect failures to the "external" port, which is then directly
2748 connected to the JS20 system.
2753 The Serial Over LAN (SoL) link is established over the primary
2754 ethernet (eth0) only, therefore, any loss of link to eth0 will result
2755 in losing your SoL connection. It will not fail over with other
2756 network traffic, as the SoL system is beyond the control of the
2759 It may be desirable to disable spanning tree on the switch
2760 (either the internal Ethernet Switch Module, or an external switch) to
2761 avoid fail-over delay issues when using bonding.
2764 15. Frequently Asked Questions
2765 ==============================
2770 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2771 The new driver was designed to be SMP safe from the start.
2773 2. What type of cards will work with it?
2774 -----------------------------------------
2776 Any Ethernet type cards (you can even mix cards - a Intel
2777 EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2778 devices need not be of the same speed.
2780 Starting with version 3.2.1, bonding also supports Infiniband
2781 slaves in active-backup mode.
2783 3. How many bonding devices can I have?
2784 ----------------------------------------
2788 4. How many slaves can a bonding device have?
2789 ----------------------------------------------
2791 This is limited only by the number of network interfaces Linux
2792 supports and/or the number of network cards you can place in your
2795 5. What happens when a slave link dies?
2796 ----------------------------------------
2798 If link monitoring is enabled, then the failing device will be
2799 disabled. The active-backup mode will fail over to a backup link, and
2800 other modes will ignore the failed link. The link will continue to be
2801 monitored, and should it recover, it will rejoin the bond (in whatever
2802 manner is appropriate for the mode). See the sections on High
2803 Availability and the documentation for each mode for additional
2806 Link monitoring can be enabled via either the miimon or
2807 arp_interval parameters (described in the module parameters section,
2808 above). In general, miimon monitors the carrier state as sensed by
2809 the underlying network device, and the arp monitor (arp_interval)
2810 monitors connectivity to another host on the local network.
2812 If no link monitoring is configured, the bonding driver will
2813 be unable to detect link failures, and will assume that all links are
2814 always available. This will likely result in lost packets, and a
2815 resulting degradation of performance. The precise performance loss
2816 depends upon the bonding mode and network configuration.
2818 6. Can bonding be used for High Availability?
2819 ----------------------------------------------
2821 Yes. See the section on High Availability for details.
2823 7. Which switches/systems does it work with?
2824 ---------------------------------------------
2826 The full answer to this depends upon the desired mode.
2828 In the basic balance modes (balance-rr and balance-xor), it
2829 works with any system that supports etherchannel (also called
2830 trunking). Most managed switches currently available have such
2831 support, and many unmanaged switches as well.
2833 The advanced balance modes (balance-tlb and balance-alb) do
2834 not have special switch requirements, but do need device drivers that
2835 support specific features (described in the appropriate section under
2836 module parameters, above).
2838 In 802.3ad mode, it works with systems that support IEEE
2839 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2840 switches currently available support 802.3ad.
2842 The active-backup mode should work with any Layer-II switch.
2844 8. Where does a bonding device get its MAC address from?
2845 ---------------------------------------------------------
2847 When using slave devices that have fixed MAC addresses, or when
2848 the fail_over_mac option is enabled, the bonding device's MAC address is
2849 the MAC address of the active slave.
2851 For other configurations, if not explicitly configured (with
2852 ifconfig or ip link), the MAC address of the bonding device is taken from
2853 its first slave device. This MAC address is then passed to all following
2854 slaves and remains persistent (even if the first slave is removed) until
2855 the bonding device is brought down or reconfigured.
2857 If you wish to change the MAC address, you can set it with
2858 ifconfig or ip link::
2860 # ifconfig bond0 hw ether 00:11:22:33:44:55
2862 # ip link set bond0 address 66:77:88:99:aa:bb
2864 The MAC address can be also changed by bringing down/up the
2865 device and then changing its slaves (or their order)::
2867 # ifconfig bond0 down ; modprobe -r bonding
2868 # ifconfig bond0 .... up
2869 # ifenslave bond0 eth...
2871 This method will automatically take the address from the next
2872 slave that is added.
2874 To restore your slaves' MAC addresses, you need to detach them
2875 from the bond (``ifenslave -d bond0 eth0``). The bonding driver will
2876 then restore the MAC addresses that the slaves had before they were
2879 16. Resources and Links
2880 =======================
2882 The latest version of the bonding driver can be found in the latest
2883 version of the linux kernel, found on http://kernel.org
2885 The latest version of this document can be found in the latest kernel
2886 source (named Documentation/networking/bonding.rst).
2888 Discussions regarding the development of the bonding driver take place
2889 on the main Linux network mailing list, hosted at vger.kernel.org. The list
2892 netdev@vger.kernel.org
2894 The administrative interface (to subscribe or unsubscribe) can
2897 http://vger.kernel.org/vger-lists.html#netdev