1 .. SPDX-License-Identifier: GPL-2.0
3 ===================================
4 Linux Ethernet Bonding Driver HOWTO
5 ===================================
7 Latest update: 27 April 2011
9 Initial release: Thomas Davis <tadavis at lbl.gov>
11 Corrections, HA extensions: 2000/10/03-15:
13 - Willy Tarreau <willy at meta-x.org>
14 - Constantine Gavrilov <const-g at xpert.com>
15 - Chad N. Tindel <ctindel at ieee dot org>
16 - Janice Girouard <girouard at us dot ibm dot com>
17 - Jay Vosburgh <fubar at us dot ibm dot com>
19 Reorganized and updated Feb 2005 by Jay Vosburgh
20 Added Sysfs information: 2006/04/24
22 - Mitch Williams <mitch.a.williams at intel.com>
27 The Linux bonding driver provides a method for aggregating
28 multiple network interfaces into a single logical "bonded" interface.
29 The behavior of the bonded interfaces depends upon the mode; generally
30 speaking, modes provide either hot standby or load balancing services.
31 Additionally, link integrity monitoring may be performed.
33 The bonding driver originally came from Donald Becker's
34 beowulf patches for kernel 2.0. It has changed quite a bit since, and
35 the original tools from extreme-linux and beowulf sites will not work
36 with this version of the driver.
38 For new versions of the driver, updated userspace tools, and
39 who to ask for help, please follow the links at the end of this file.
43 1. Bonding Driver Installation
45 2. Bonding Driver Options
47 3. Configuring Bonding Devices
48 3.1 Configuration with Sysconfig Support
49 3.1.1 Using DHCP with Sysconfig
50 3.1.2 Configuring Multiple Bonds with Sysconfig
51 3.2 Configuration with Initscripts Support
52 3.2.1 Using DHCP with Initscripts
53 3.2.2 Configuring Multiple Bonds with Initscripts
54 3.3 Configuring Bonding Manually with Ifenslave
55 3.3.1 Configuring Multiple Bonds Manually
56 3.4 Configuring Bonding Manually via Sysfs
57 3.5 Configuration with Interfaces Support
58 3.6 Overriding Configuration for Special Cases
59 3.7 Configuring LACP for 802.3ad mode in a more secure way
61 4. Querying Bonding Configuration
62 4.1 Bonding Configuration
63 4.2 Network Configuration
65 5. Switch Configuration
67 6. 802.1q VLAN Support
70 7.1 ARP Monitor Operation
71 7.2 Configuring Multiple ARP Targets
72 7.3 MII Monitor Operation
74 8. Potential Trouble Sources
75 8.1 Adventures in Routing
76 8.2 Ethernet Device Renaming
77 8.3 Painfully Slow Or No Failed Link Detection By Miimon
83 11. Configuring Bonding for High Availability
84 11.1 High Availability in a Single Switch Topology
85 11.2 High Availability in a Multiple Switch Topology
86 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
87 11.2.2 HA Link Monitoring for Multiple Switch Topology
89 12. Configuring Bonding for Maximum Throughput
90 12.1 Maximum Throughput in a Single Switch Topology
91 12.1.1 MT Bonding Mode Selection for Single Switch Topology
92 12.1.2 MT Link Monitoring for Single Switch Topology
93 12.2 Maximum Throughput in a Multiple Switch Topology
94 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
95 12.2.2 MT Link Monitoring for Multiple Switch Topology
97 13. Switch Behavior Issues
98 13.1 Link Establishment and Failover Delays
99 13.2 Duplicated Incoming Packets
101 14. Hardware Specific Considerations
104 15. Frequently Asked Questions
106 16. Resources and Links
109 1. Bonding Driver Installation
110 ==============================
112 Most popular distro kernels ship with the bonding driver
113 already available as a module. If your distro does not, or you
114 have need to compile bonding from source (e.g., configuring and
115 installing a mainline kernel from kernel.org), you'll need to perform
118 1.1 Configure and build the kernel with bonding
119 -----------------------------------------------
121 The current version of the bonding driver is available in the
122 drivers/net/bonding subdirectory of the most recent kernel source
123 (which is available on http://kernel.org). Most users "rolling their
124 own" will want to use the most recent kernel from kernel.org.
126 Configure kernel with "make menuconfig" (or "make xconfig" or
127 "make config"), then select "Bonding driver support" in the "Network
128 device support" section. It is recommended that you configure the
129 driver as module since it is currently the only way to pass parameters
130 to the driver or configure more than one bonding device.
132 Build and install the new kernel and modules.
134 1.2 Bonding Control Utility
135 ---------------------------
137 It is recommended to configure bonding via iproute2 (netlink)
138 or sysfs, the old ifenslave control utility is obsolete.
140 2. Bonding Driver Options
141 =========================
143 Options for the bonding driver are supplied as parameters to the
144 bonding module at load time, or are specified via sysfs.
146 Module options may be given as command line arguments to the
147 insmod or modprobe command, but are usually specified in either the
148 ``/etc/modprobe.d/*.conf`` configuration files, or in a distro-specific
149 configuration file (some of which are detailed in the next section).
151 Details on bonding support for sysfs is provided in the
152 "Configuring Bonding Manually via Sysfs" section, below.
154 The available bonding driver parameters are listed below. If a
155 parameter is not specified the default value is used. When initially
156 configuring a bond, it is recommended "tail -f /var/log/messages" be
157 run in a separate window to watch for bonding driver error messages.
159 It is critical that either the miimon or arp_interval and
160 arp_ip_target parameters be specified, otherwise serious network
161 degradation will occur during link failures. Very few devices do not
162 support at least miimon, so there is really no reason not to use it.
164 Options with textual values will accept either the text name
165 or, for backwards compatibility, the option value. E.g.,
166 "mode=802.3ad" and "mode=4" set the same mode.
168 The parameters are as follows:
172 Specifies the new active slave for modes that support it
173 (active-backup, balance-alb and balance-tlb). Possible values
174 are the name of any currently enslaved interface, or an empty
175 string. If a name is given, the slave and its link must be up in order
176 to be selected as the new active slave. If an empty string is
177 specified, the current active slave is cleared, and a new active
178 slave is selected automatically.
180 Note that this is only available through the sysfs interface. No module
181 parameter by this name exists.
183 The normal value of this option is the name of the currently
184 active slave, or the empty string if there is no active slave or
185 the current mode does not use an active slave.
189 In an AD system, this specifies the system priority. The allowed range
190 is 1 - 65535. If the value is not specified, it takes 65535 as the
193 This parameter has effect only in 802.3ad mode and is available through
198 In an AD system, this specifies the mac-address for the actor in
199 protocol packet exchanges (LACPDUs). The value cannot be a multicast
200 address. If the all-zeroes MAC is specified, bonding will internally
201 use the MAC of the bond itself. It is preferred to have the
202 local-admin bit set for this mac but driver does not enforce it. If
203 the value is not given then system defaults to using the masters'
204 mac address as actors' system address.
206 This parameter has effect only in 802.3ad mode and is available through
211 Specifies the 802.3ad aggregation selection logic to use. The
212 possible values and their effects are:
216 The active aggregator is chosen by largest aggregate
219 Reselection of the active aggregator occurs only when all
220 slaves of the active aggregator are down or the active
221 aggregator has no slaves.
223 This is the default value.
227 The active aggregator is chosen by largest aggregate
228 bandwidth. Reselection occurs if:
230 - A slave is added to or removed from the bond
232 - Any slave's link state changes
234 - Any slave's 802.3ad association state changes
236 - The bond's administrative state changes to up
240 The active aggregator is chosen by the largest number of
241 ports (slaves). Reselection occurs as described under the
242 "bandwidth" setting, above.
244 The bandwidth and count selection policies permit failover of
245 802.3ad aggregations when partial failure of the active aggregator
246 occurs. This keeps the aggregator with the highest availability
247 (either in bandwidth or in number of ports) active at all times.
249 This option was added in bonding version 3.4.0.
253 In an AD system, the port-key has three parts as shown below -
263 This defines the upper 10 bits of the port key. The values can be
264 from 0 - 1023. If not given, the system defaults to 0.
266 This parameter has effect only in 802.3ad mode and is available through
271 Specifies that duplicate frames (received on inactive ports) should be
272 dropped (0) or delivered (1).
274 Normally, bonding will drop duplicate frames (received on inactive
275 ports), which is desirable for most users. But there are some times
276 it is nice to allow duplicate frames to be delivered.
278 The default value is 0 (drop duplicate frames received on inactive
283 Specifies the ARP link monitoring frequency in milliseconds.
285 The ARP monitor works by periodically checking the slave
286 devices to determine whether they have sent or received
287 traffic recently (the precise criteria depends upon the
288 bonding mode, and the state of the slave). Regular traffic is
289 generated via ARP probes issued for the addresses specified by
290 the arp_ip_target option.
292 This behavior can be modified by the arp_validate option,
295 If ARP monitoring is used in an etherchannel compatible mode
296 (modes 0 and 2), the switch should be configured in a mode
297 that evenly distributes packets across all links. If the
298 switch is configured to distribute the packets in an XOR
299 fashion, all replies from the ARP targets will be received on
300 the same link which could cause the other team members to
301 fail. ARP monitoring should not be used in conjunction with
302 miimon. A value of 0 disables ARP monitoring. The default
307 Specifies the IP addresses to use as ARP monitoring peers when
308 arp_interval is > 0. These are the targets of the ARP request
309 sent to determine the health of the link to the targets.
310 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
311 addresses must be separated by a comma. At least one IP
312 address must be given for ARP monitoring to function. The
313 maximum number of targets that can be specified is 16. The
314 default value is no IP addresses.
318 Specifies whether or not ARP probes and replies should be
319 validated in any mode that supports arp monitoring, or whether
320 non-ARP traffic should be filtered (disregarded) for link
327 No validation or filtering is performed.
331 Validation is performed only for the active slave.
335 Validation is performed only for backup slaves.
339 Validation is performed for all slaves.
343 Filtering is applied to all slaves. No validation is
348 Filtering is applied to all slaves, validation is performed
349 only for the active slave.
353 Filtering is applied to all slaves, validation is performed
354 only for backup slaves.
358 Enabling validation causes the ARP monitor to examine the incoming
359 ARP requests and replies, and only consider a slave to be up if it
360 is receiving the appropriate ARP traffic.
362 For an active slave, the validation checks ARP replies to confirm
363 that they were generated by an arp_ip_target. Since backup slaves
364 do not typically receive these replies, the validation performed
365 for backup slaves is on the broadcast ARP request sent out via the
366 active slave. It is possible that some switch or network
367 configurations may result in situations wherein the backup slaves
368 do not receive the ARP requests; in such a situation, validation
369 of backup slaves must be disabled.
371 The validation of ARP requests on backup slaves is mainly helping
372 bonding to decide which slaves are more likely to work in case of
373 the active slave failure, it doesn't really guarantee that the
374 backup slave will work if it's selected as the next active slave.
376 Validation is useful in network configurations in which multiple
377 bonding hosts are concurrently issuing ARPs to one or more targets
378 beyond a common switch. Should the link between the switch and
379 target fail (but not the switch itself), the probe traffic
380 generated by the multiple bonding instances will fool the standard
381 ARP monitor into considering the links as still up. Use of
382 validation can resolve this, as the ARP monitor will only consider
383 ARP requests and replies associated with its own instance of
388 Enabling filtering causes the ARP monitor to only use incoming ARP
389 packets for link availability purposes. Arriving packets that are
390 not ARPs are delivered normally, but do not count when determining
391 if a slave is available.
393 Filtering operates by only considering the reception of ARP
394 packets (any ARP packet, regardless of source or destination) when
395 determining if a slave has received traffic for link availability
398 Filtering is useful in network configurations in which significant
399 levels of third party broadcast traffic would fool the standard
400 ARP monitor into considering the links as still up. Use of
401 filtering can resolve this, as only ARP traffic is considered for
402 link availability purposes.
404 This option was added in bonding version 3.1.0.
408 Specifies the quantity of arp_ip_targets that must be reachable
409 in order for the ARP monitor to consider a slave as being up.
410 This option affects only active-backup mode for slaves with
411 arp_validation enabled.
417 consider the slave up only when any of the arp_ip_targets
422 consider the slave up only when all of the arp_ip_targets
427 Specifies the time, in milliseconds, to wait before disabling
428 a slave after a link failure has been detected. This option
429 is only valid for the miimon link monitor. The downdelay
430 value should be a multiple of the miimon value; if not, it
431 will be rounded down to the nearest multiple. The default
436 Specifies whether active-backup mode should set all slaves to
437 the same MAC address at enslavement (the traditional
438 behavior), or, when enabled, perform special handling of the
439 bond's MAC address in accordance with the selected policy.
445 This setting disables fail_over_mac, and causes
446 bonding to set all slaves of an active-backup bond to
447 the same MAC address at enslavement time. This is the
452 The "active" fail_over_mac policy indicates that the
453 MAC address of the bond should always be the MAC
454 address of the currently active slave. The MAC
455 address of the slaves is not changed; instead, the MAC
456 address of the bond changes during a failover.
458 This policy is useful for devices that cannot ever
459 alter their MAC address, or for devices that refuse
460 incoming broadcasts with their own source MAC (which
461 interferes with the ARP monitor).
463 The down side of this policy is that every device on
464 the network must be updated via gratuitous ARP,
465 vs. just updating a switch or set of switches (which
466 often takes place for any traffic, not just ARP
467 traffic, if the switch snoops incoming traffic to
468 update its tables) for the traditional method. If the
469 gratuitous ARP is lost, communication may be
472 When this policy is used in conjunction with the mii
473 monitor, devices which assert link up prior to being
474 able to actually transmit and receive are particularly
475 susceptible to loss of the gratuitous ARP, and an
476 appropriate updelay setting may be required.
480 The "follow" fail_over_mac policy causes the MAC
481 address of the bond to be selected normally (normally
482 the MAC address of the first slave added to the bond).
483 However, the second and subsequent slaves are not set
484 to this MAC address while they are in a backup role; a
485 slave is programmed with the bond's MAC address at
486 failover time (and the formerly active slave receives
487 the newly active slave's MAC address).
489 This policy is useful for multiport devices that
490 either become confused or incur a performance penalty
491 when multiple ports are programmed with the same MAC
495 The default policy is none, unless the first slave cannot
496 change its MAC address, in which case the active policy is
499 This option may be modified via sysfs only when no slaves are
502 This option was added in bonding version 3.2.0. The "follow"
503 policy was added in bonding version 3.3.0.
507 Option specifying the rate in which we'll ask our link partner
508 to transmit LACPDU packets in 802.3ad mode. Possible values
512 Request partner to transmit LACPDUs every 30 seconds
515 Request partner to transmit LACPDUs every 1 second
521 Specifies the number of bonding devices to create for this
522 instance of the bonding driver. E.g., if max_bonds is 3, and
523 the bonding driver is not already loaded, then bond0, bond1
524 and bond2 will be created. The default value is 1. Specifying
525 a value of 0 will load bonding, but will not create any devices.
529 Specifies the MII link monitoring frequency in milliseconds.
530 This determines how often the link state of each slave is
531 inspected for link failures. A value of zero disables MII
532 link monitoring. A value of 100 is a good starting point.
533 The use_carrier option, below, affects how the link state is
534 determined. See the High Availability section for additional
535 information. The default value is 0.
539 Specifies the minimum number of links that must be active before
540 asserting carrier. It is similar to the Cisco EtherChannel min-links
541 feature. This allows setting the minimum number of member ports that
542 must be up (link-up state) before marking the bond device as up
543 (carrier on). This is useful for situations where higher level services
544 such as clustering want to ensure a minimum number of low bandwidth
545 links are active before switchover. This option only affect 802.3ad
548 The default value is 0. This will cause carrier to be asserted (for
549 802.3ad mode) whenever there is an active aggregator, regardless of the
550 number of available links in that aggregator. Note that, because an
551 aggregator cannot be active without at least one available link,
552 setting this option to 0 or to 1 has the exact same effect.
556 Specifies one of the bonding policies. The default is
557 balance-rr (round robin). Possible values are:
561 Round-robin policy: Transmit packets in sequential
562 order from the first available slave through the
563 last. This mode provides load balancing and fault
568 Active-backup policy: Only one slave in the bond is
569 active. A different slave becomes active if, and only
570 if, the active slave fails. The bond's MAC address is
571 externally visible on only one port (network adapter)
572 to avoid confusing the switch.
574 In bonding version 2.6.2 or later, when a failover
575 occurs in active-backup mode, bonding will issue one
576 or more gratuitous ARPs on the newly active slave.
577 One gratuitous ARP is issued for the bonding master
578 interface and each VLAN interfaces configured above
579 it, provided that the interface has at least one IP
580 address configured. Gratuitous ARPs issued for VLAN
581 interfaces are tagged with the appropriate VLAN id.
583 This mode provides fault tolerance. The primary
584 option, documented below, affects the behavior of this
589 XOR policy: Transmit based on the selected transmit
590 hash policy. The default policy is a simple [(source
591 MAC address XOR'd with destination MAC address XOR
592 packet type ID) modulo slave count]. Alternate transmit
593 policies may be selected via the xmit_hash_policy option,
596 This mode provides load balancing and fault tolerance.
600 Broadcast policy: transmits everything on all slave
601 interfaces. This mode provides fault tolerance.
605 IEEE 802.3ad Dynamic link aggregation. Creates
606 aggregation groups that share the same speed and
607 duplex settings. Utilizes all slaves in the active
608 aggregator according to the 802.3ad specification.
610 Slave selection for outgoing traffic is done according
611 to the transmit hash policy, which may be changed from
612 the default simple XOR policy via the xmit_hash_policy
613 option, documented below. Note that not all transmit
614 policies may be 802.3ad compliant, particularly in
615 regards to the packet mis-ordering requirements of
616 section 43.2.4 of the 802.3ad standard. Differing
617 peer implementations will have varying tolerances for
622 1. Ethtool support in the base drivers for retrieving
623 the speed and duplex of each slave.
625 2. A switch that supports IEEE 802.3ad Dynamic link
628 Most switches will require some type of configuration
629 to enable 802.3ad mode.
633 Adaptive transmit load balancing: channel bonding that
634 does not require any special switch support.
636 In tlb_dynamic_lb=1 mode; the outgoing traffic is
637 distributed according to the current load (computed
638 relative to the speed) on each slave.
640 In tlb_dynamic_lb=0 mode; the load balancing based on
641 current load is disabled and the load is distributed
642 only using the hash distribution.
644 Incoming traffic is received by the current slave.
645 If the receiving slave fails, another slave takes over
646 the MAC address of the failed receiving slave.
650 Ethtool support in the base drivers for retrieving the
655 Adaptive load balancing: includes balance-tlb plus
656 receive load balancing (rlb) for IPV4 traffic, and
657 does not require any special switch support. The
658 receive load balancing is achieved by ARP negotiation.
659 The bonding driver intercepts the ARP Replies sent by
660 the local system on their way out and overwrites the
661 source hardware address with the unique hardware
662 address of one of the slaves in the bond such that
663 different peers use different hardware addresses for
666 Receive traffic from connections created by the server
667 is also balanced. When the local system sends an ARP
668 Request the bonding driver copies and saves the peer's
669 IP information from the ARP packet. When the ARP
670 Reply arrives from the peer, its hardware address is
671 retrieved and the bonding driver initiates an ARP
672 reply to this peer assigning it to one of the slaves
673 in the bond. A problematic outcome of using ARP
674 negotiation for balancing is that each time that an
675 ARP request is broadcast it uses the hardware address
676 of the bond. Hence, peers learn the hardware address
677 of the bond and the balancing of receive traffic
678 collapses to the current slave. This is handled by
679 sending updates (ARP Replies) to all the peers with
680 their individually assigned hardware address such that
681 the traffic is redistributed. Receive traffic is also
682 redistributed when a new slave is added to the bond
683 and when an inactive slave is re-activated. The
684 receive load is distributed sequentially (round robin)
685 among the group of highest speed slaves in the bond.
687 When a link is reconnected or a new slave joins the
688 bond the receive traffic is redistributed among all
689 active slaves in the bond by initiating ARP Replies
690 with the selected MAC address to each of the
691 clients. The updelay parameter (detailed below) must
692 be set to a value equal or greater than the switch's
693 forwarding delay so that the ARP Replies sent to the
694 peers will not be blocked by the switch.
698 1. Ethtool support in the base drivers for retrieving
699 the speed of each slave.
701 2. Base driver support for setting the hardware
702 address of a device while it is open. This is
703 required so that there will always be one slave in the
704 team using the bond hardware address (the
705 curr_active_slave) while having a unique hardware
706 address for each slave in the bond. If the
707 curr_active_slave fails its hardware address is
708 swapped with the new curr_active_slave that was
714 Specify the number of peer notifications (gratuitous ARPs and
715 unsolicited IPv6 Neighbor Advertisements) to be issued after a
716 failover event. As soon as the link is up on the new slave
717 (possibly immediately) a peer notification is sent on the
718 bonding device and each VLAN sub-device. This is repeated at
719 the rate specified by peer_notif_delay if the number is
722 The valid range is 0 - 255; the default value is 1. These options
723 affect only the active-backup mode. These options were added for
724 bonding versions 3.3.0 and 3.4.0 respectively.
726 From Linux 3.0 and bonding version 3.7.1, these notifications
727 are generated by the ipv4 and ipv6 code and the numbers of
728 repetitions cannot be set independently.
732 Specify the number of packets to transmit through a slave before
733 moving to the next one. When set to 0 then a slave is chosen at
736 The valid range is 0 - 65535; the default value is 1. This option
737 has effect only in balance-rr mode.
741 Specify the delay, in milliseconds, between each peer
742 notification (gratuitous ARP and unsolicited IPv6 Neighbor
743 Advertisement) when they are issued after a failover event.
744 This delay should be a multiple of the link monitor interval
745 (arp_interval or miimon, whichever is active). The default
746 value is 0 which means to match the value of the link monitor
751 A string (eth0, eth2, etc) specifying which slave is the
752 primary device. The specified device will always be the
753 active slave while it is available. Only when the primary is
754 off-line will alternate devices be used. This is useful when
755 one slave is preferred over another, e.g., when one slave has
756 higher throughput than another.
758 The primary option is only valid for active-backup(1),
759 balance-tlb (5) and balance-alb (6) mode.
763 Specifies the reselection policy for the primary slave. This
764 affects how the primary slave is chosen to become the active slave
765 when failure of the active slave or recovery of the primary slave
766 occurs. This option is designed to prevent flip-flopping between
767 the primary slave and other slaves. Possible values are:
769 always or 0 (default)
771 The primary slave becomes the active slave whenever it
776 The primary slave becomes the active slave when it comes
777 back up, if the speed and duplex of the primary slave is
778 better than the speed and duplex of the current active
783 The primary slave becomes the active slave only if the
784 current active slave fails and the primary slave is up.
786 The primary_reselect setting is ignored in two cases:
788 If no slaves are active, the first slave to recover is
789 made the active slave.
791 When initially enslaved, the primary slave is always made
794 Changing the primary_reselect policy via sysfs will cause an
795 immediate selection of the best active slave according to the new
796 policy. This may or may not result in a change of the active
797 slave, depending upon the circumstances.
799 This option was added for bonding version 3.6.0.
803 Specifies if dynamic shuffling of flows is enabled in tlb
804 mode. The value has no effect on any other modes.
806 The default behavior of tlb mode is to shuffle active flows across
807 slaves based on the load in that interval. This gives nice lb
808 characteristics but can cause packet reordering. If re-ordering is
809 a concern use this variable to disable flow shuffling and rely on
810 load balancing provided solely by the hash distribution.
811 xmit-hash-policy can be used to select the appropriate hashing for
814 The sysfs entry can be used to change the setting per bond device
815 and the initial value is derived from the module parameter. The
816 sysfs entry is allowed to be changed only if the bond device is
819 The default value is "1" that enables flow shuffling while value "0"
820 disables it. This option was added in bonding driver 3.7.1
825 Specifies the time, in milliseconds, to wait before enabling a
826 slave after a link recovery has been detected. This option is
827 only valid for the miimon link monitor. The updelay value
828 should be a multiple of the miimon value; if not, it will be
829 rounded down to the nearest multiple. The default value is 0.
833 Specifies whether or not miimon should use MII or ETHTOOL
834 ioctls vs. netif_carrier_ok() to determine the link
835 status. The MII or ETHTOOL ioctls are less efficient and
836 utilize a deprecated calling sequence within the kernel. The
837 netif_carrier_ok() relies on the device driver to maintain its
838 state with netif_carrier_on/off; at this writing, most, but
839 not all, device drivers support this facility.
841 If bonding insists that the link is up when it should not be,
842 it may be that your network device driver does not support
843 netif_carrier_on/off. The default state for netif_carrier is
844 "carrier on," so if a driver does not support netif_carrier,
845 it will appear as if the link is always up. In this case,
846 setting use_carrier to 0 will cause bonding to revert to the
847 MII / ETHTOOL ioctl method to determine the link state.
849 A value of 1 enables the use of netif_carrier_ok(), a value of
850 0 will use the deprecated MII / ETHTOOL ioctls. The default
855 Selects the transmit hash policy to use for slave selection in
856 balance-xor, 802.3ad, and tlb modes. Possible values are:
860 Uses XOR of hardware MAC addresses and packet type ID
861 field to generate the hash. The formula is
863 hash = source MAC XOR destination MAC XOR packet type ID
864 slave number = hash modulo slave count
866 This algorithm will place all traffic to a particular
867 network peer on the same slave.
869 This algorithm is 802.3ad compliant.
873 This policy uses a combination of layer2 and layer3
874 protocol information to generate the hash.
876 Uses XOR of hardware MAC addresses and IP addresses to
877 generate the hash. The formula is
879 hash = source MAC XOR destination MAC XOR packet type ID
880 hash = hash XOR source IP XOR destination IP
881 hash = hash XOR (hash RSHIFT 16)
882 hash = hash XOR (hash RSHIFT 8)
883 And then hash is reduced modulo slave count.
885 If the protocol is IPv6 then the source and destination
886 addresses are first hashed using ipv6_addr_hash.
888 This algorithm will place all traffic to a particular
889 network peer on the same slave. For non-IP traffic,
890 the formula is the same as for the layer2 transmit
893 This policy is intended to provide a more balanced
894 distribution of traffic than layer2 alone, especially
895 in environments where a layer3 gateway device is
896 required to reach most destinations.
898 This algorithm is 802.3ad compliant.
902 This policy uses upper layer protocol information,
903 when available, to generate the hash. This allows for
904 traffic to a particular network peer to span multiple
905 slaves, although a single connection will not span
908 The formula for unfragmented TCP and UDP packets is
910 hash = source port, destination port (as in the header)
911 hash = hash XOR source IP XOR destination IP
912 hash = hash XOR (hash RSHIFT 16)
913 hash = hash XOR (hash RSHIFT 8)
914 And then hash is reduced modulo slave count.
916 If the protocol is IPv6 then the source and destination
917 addresses are first hashed using ipv6_addr_hash.
919 For fragmented TCP or UDP packets and all other IPv4 and
920 IPv6 protocol traffic, the source and destination port
921 information is omitted. For non-IP traffic, the
922 formula is the same as for the layer2 transmit hash
925 This algorithm is not fully 802.3ad compliant. A
926 single TCP or UDP conversation containing both
927 fragmented and unfragmented packets will see packets
928 striped across two interfaces. This may result in out
929 of order delivery. Most traffic types will not meet
930 this criteria, as TCP rarely fragments traffic, and
931 most UDP traffic is not involved in extended
932 conversations. Other implementations of 802.3ad may
933 or may not tolerate this noncompliance.
937 This policy uses the same formula as layer2+3 but it
938 relies on skb_flow_dissect to obtain the header fields
939 which might result in the use of inner headers if an
940 encapsulation protocol is used. For example this will
941 improve the performance for tunnel users because the
942 packets will be distributed according to the encapsulated
947 This policy uses the same formula as layer3+4 but it
948 relies on skb_flow_dissect to obtain the header fields
949 which might result in the use of inner headers if an
950 encapsulation protocol is used. For example this will
951 improve the performance for tunnel users because the
952 packets will be distributed according to the encapsulated
955 The default value is layer2. This option was added in bonding
956 version 2.6.3. In earlier versions of bonding, this parameter
957 does not exist, and the layer2 policy is the only policy. The
958 layer2+3 value was added for bonding version 3.2.2.
962 Specifies the number of IGMP membership reports to be issued after
963 a failover event. One membership report is issued immediately after
964 the failover, subsequent packets are sent in each 200ms interval.
966 The valid range is 0 - 255; the default value is 1. A value of 0
967 prevents the IGMP membership report from being issued in response
968 to the failover event.
970 This option is useful for bonding modes balance-rr (0), active-backup
971 (1), balance-tlb (5) and balance-alb (6), in which a failover can
972 switch the IGMP traffic from one slave to another. Therefore a fresh
973 IGMP report must be issued to cause the switch to forward the incoming
974 IGMP traffic over the newly selected slave.
976 This option was added for bonding version 3.7.0.
980 Specifies the number of seconds between instances where the bonding
981 driver sends learning packets to each slaves peer switch.
983 The valid range is 1 - 0x7fffffff; the default value is 1. This Option
984 has effect only in balance-tlb and balance-alb modes.
986 3. Configuring Bonding Devices
987 ==============================
989 You can configure bonding using either your distro's network
990 initialization scripts, or manually using either iproute2 or the
991 sysfs interface. Distros generally use one of three packages for the
992 network initialization scripts: initscripts, sysconfig or interfaces.
993 Recent versions of these packages have support for bonding, while older
996 We will first describe the options for configuring bonding for
997 distros using versions of initscripts, sysconfig and interfaces with full
998 or partial support for bonding, then provide information on enabling
999 bonding without support from the network initialization scripts (i.e.,
1000 older versions of initscripts or sysconfig).
1002 If you're unsure whether your distro uses sysconfig,
1003 initscripts or interfaces, or don't know if it's new enough, have no fear.
1004 Determining this is fairly straightforward.
1006 First, look for a file called interfaces in /etc/network directory.
1007 If this file is present in your system, then your system use interfaces. See
1008 Configuration with Interfaces Support.
1010 Else, issue the command::
1012 $ rpm -qf /sbin/ifup
1014 It will respond with a line of text starting with either
1015 "initscripts" or "sysconfig," followed by some numbers. This is the
1016 package that provides your network initialization scripts.
1018 Next, to determine if your installation supports bonding,
1021 $ grep ifenslave /sbin/ifup
1023 If this returns any matches, then your initscripts or
1024 sysconfig has support for bonding.
1026 3.1 Configuration with Sysconfig Support
1027 ----------------------------------------
1029 This section applies to distros using a version of sysconfig
1030 with bonding support, for example, SuSE Linux Enterprise Server 9.
1032 SuSE SLES 9's networking configuration system does support
1033 bonding, however, at this writing, the YaST system configuration
1034 front end does not provide any means to work with bonding devices.
1035 Bonding devices can be managed by hand, however, as follows.
1037 First, if they have not already been configured, configure the
1038 slave devices. On SLES 9, this is most easily done by running the
1039 yast2 sysconfig configuration utility. The goal is for to create an
1040 ifcfg-id file for each slave device. The simplest way to accomplish
1041 this is to configure the devices for DHCP (this is only to get the
1042 file ifcfg-id file created; see below for some issues with DHCP). The
1043 name of the configuration file for each device will be of the form::
1045 ifcfg-id-xx:xx:xx:xx:xx:xx
1047 Where the "xx" portion will be replaced with the digits from
1048 the device's permanent MAC address.
1050 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
1051 created, it is necessary to edit the configuration files for the slave
1052 devices (the MAC addresses correspond to those of the slave devices).
1053 Before editing, the file will contain multiple lines, and will look
1054 something like this::
1059 UNIQUE='XNzu.WeZGOGF+4wE'
1060 _nm_name='bus-pci-0001:61:01.0'
1062 Change the BOOTPROTO and STARTMODE lines to the following::
1067 Do not alter the UNIQUE or _nm_name lines. Remove any other
1068 lines (USERCTL, etc).
1070 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
1071 it's time to create the configuration file for the bonding device
1072 itself. This file is named ifcfg-bondX, where X is the number of the
1073 bonding device to create, starting at 0. The first such file is
1074 ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
1075 network configuration system will correctly start multiple instances
1078 The contents of the ifcfg-bondX file is as follows::
1081 BROADCAST="10.0.2.255"
1083 NETMASK="255.255.0.0"
1087 BONDING_MASTER="yes"
1088 BONDING_MODULE_OPTS="mode=active-backup miimon=100"
1089 BONDING_SLAVE0="eth0"
1090 BONDING_SLAVE1="bus-pci-0000:06:08.1"
1092 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
1093 values with the appropriate values for your network.
1095 The STARTMODE specifies when the device is brought online.
1096 The possible values are:
1098 ======== ======================================================
1099 onboot The device is started at boot time. If you're not
1100 sure, this is probably what you want.
1102 manual The device is started only when ifup is called
1103 manually. Bonding devices may be configured this
1104 way if you do not wish them to start automatically
1105 at boot for some reason.
1107 hotplug The device is started by a hotplug event. This is not
1108 a valid choice for a bonding device.
1110 off or The device configuration is ignored.
1112 ======== ======================================================
1114 The line BONDING_MASTER='yes' indicates that the device is a
1115 bonding master device. The only useful value is "yes."
1117 The contents of BONDING_MODULE_OPTS are supplied to the
1118 instance of the bonding module for this device. Specify the options
1119 for the bonding mode, link monitoring, and so on here. Do not include
1120 the max_bonds bonding parameter; this will confuse the configuration
1121 system if you have multiple bonding devices.
1123 Finally, supply one BONDING_SLAVEn="slave device" for each
1124 slave. where "n" is an increasing value, one for each slave. The
1125 "slave device" is either an interface name, e.g., "eth0", or a device
1126 specifier for the network device. The interface name is easier to
1127 find, but the ethN names are subject to change at boot time if, e.g.,
1128 a device early in the sequence has failed. The device specifiers
1129 (bus-pci-0000:06:08.1 in the example above) specify the physical
1130 network device, and will not change unless the device's bus location
1131 changes (for example, it is moved from one PCI slot to another). The
1132 example above uses one of each type for demonstration purposes; most
1133 configurations will choose one or the other for all slave devices.
1135 When all configuration files have been modified or created,
1136 networking must be restarted for the configuration changes to take
1137 effect. This can be accomplished via the following::
1139 # /etc/init.d/network restart
1141 Note that the network control script (/sbin/ifdown) will
1142 remove the bonding module as part of the network shutdown processing,
1143 so it is not necessary to remove the module by hand if, e.g., the
1144 module parameters have changed.
1146 Also, at this writing, YaST/YaST2 will not manage bonding
1147 devices (they do not show bonding interfaces on its list of network
1148 devices). It is necessary to edit the configuration file by hand to
1149 change the bonding configuration.
1151 Additional general options and details of the ifcfg file
1152 format can be found in an example ifcfg template file::
1154 /etc/sysconfig/network/ifcfg.template
1156 Note that the template does not document the various ``BONDING_*``
1157 settings described above, but does describe many of the other options.
1159 3.1.1 Using DHCP with Sysconfig
1160 -------------------------------
1162 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1163 will cause it to query DHCP for its IP address information. At this
1164 writing, this does not function for bonding devices; the scripts
1165 attempt to obtain the device address from DHCP prior to adding any of
1166 the slave devices. Without active slaves, the DHCP requests are not
1167 sent to the network.
1169 3.1.2 Configuring Multiple Bonds with Sysconfig
1170 -----------------------------------------------
1172 The sysconfig network initialization system is capable of
1173 handling multiple bonding devices. All that is necessary is for each
1174 bonding instance to have an appropriately configured ifcfg-bondX file
1175 (as described above). Do not specify the "max_bonds" parameter to any
1176 instance of bonding, as this will confuse sysconfig. If you require
1177 multiple bonding devices with identical parameters, create multiple
1180 Because the sysconfig scripts supply the bonding module
1181 options in the ifcfg-bondX file, it is not necessary to add them to
1182 the system ``/etc/modules.d/*.conf`` configuration files.
1184 3.2 Configuration with Initscripts Support
1185 ------------------------------------------
1187 This section applies to distros using a recent version of
1188 initscripts with bonding support, for example, Red Hat Enterprise Linux
1189 version 3 or later, Fedora, etc. On these systems, the network
1190 initialization scripts have knowledge of bonding, and can be configured to
1191 control bonding devices. Note that older versions of the initscripts
1192 package have lower levels of support for bonding; this will be noted where
1195 These distros will not automatically load the network adapter
1196 driver unless the ethX device is configured with an IP address.
1197 Because of this constraint, users must manually configure a
1198 network-script file for all physical adapters that will be members of
1199 a bondX link. Network script files are located in the directory:
1201 /etc/sysconfig/network-scripts
1203 The file name must be prefixed with "ifcfg-eth" and suffixed
1204 with the adapter's physical adapter number. For example, the script
1205 for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1206 Place the following text in the file::
1215 The DEVICE= line will be different for every ethX device and
1216 must correspond with the name of the file, i.e., ifcfg-eth1 must have
1217 a device line of DEVICE=eth1. The setting of the MASTER= line will
1218 also depend on the final bonding interface name chosen for your bond.
1219 As with other network devices, these typically start at 0, and go up
1220 one for each device, i.e., the first bonding instance is bond0, the
1221 second is bond1, and so on.
1223 Next, create a bond network script. The file name for this
1224 script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1225 the number of the bond. For bond0 the file is named "ifcfg-bond0",
1226 for bond1 it is named "ifcfg-bond1", and so on. Within that file,
1227 place the following text::
1231 NETMASK=255.255.255.0
1233 BROADCAST=192.168.1.255
1238 Be sure to change the networking specific lines (IPADDR,
1239 NETMASK, NETWORK and BROADCAST) to match your network configuration.
1241 For later versions of initscripts, such as that found with Fedora
1242 7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1243 and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
1244 file, e.g. a line of the format::
1246 BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
1248 will configure the bond with the specified options. The options
1249 specified in BONDING_OPTS are identical to the bonding module parameters
1250 except for the arp_ip_target field when using versions of initscripts older
1251 than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
1252 using older versions each target should be included as a separate option and
1253 should be preceded by a '+' to indicate it should be added to the list of
1254 queried targets, e.g.,::
1256 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1258 is the proper syntax to specify multiple targets. When specifying
1259 options via BONDING_OPTS, it is not necessary to edit
1260 ``/etc/modprobe.d/*.conf``.
1262 For even older versions of initscripts that do not support
1263 BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1264 your distro) to load the bonding module with your desired options when the
1265 bond0 interface is brought up. The following lines in /etc/modprobe.d/*.conf
1266 will load the bonding module, and select its options:
1269 options bond0 mode=balance-alb miimon=100
1271 Replace the sample parameters with the appropriate set of
1272 options for your configuration.
1274 Finally run "/etc/rc.d/init.d/network restart" as root. This
1275 will restart the networking subsystem and your bond link should be now
1278 3.2.1 Using DHCP with Initscripts
1279 ---------------------------------
1281 Recent versions of initscripts (the versions supplied with Fedora
1282 Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1283 work) have support for assigning IP information to bonding devices via
1286 To configure bonding for DHCP, configure it as described
1287 above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1288 and add a line consisting of "TYPE=Bonding". Note that the TYPE value
1291 3.2.2 Configuring Multiple Bonds with Initscripts
1292 -------------------------------------------------
1294 Initscripts packages that are included with Fedora 7 and Red Hat
1295 Enterprise Linux 5 support multiple bonding interfaces by simply
1296 specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1297 number of the bond. This support requires sysfs support in the kernel,
1298 and a bonding driver of version 3.0.0 or later. Other configurations may
1299 not support this method for specifying multiple bonding interfaces; for
1300 those instances, see the "Configuring Multiple Bonds Manually" section,
1303 3.3 Configuring Bonding Manually with iproute2
1304 -----------------------------------------------
1306 This section applies to distros whose network initialization
1307 scripts (the sysconfig or initscripts package) do not have specific
1308 knowledge of bonding. One such distro is SuSE Linux Enterprise Server
1311 The general method for these systems is to place the bonding
1312 module parameters into a config file in /etc/modprobe.d/ (as
1313 appropriate for the installed distro), then add modprobe and/or
1314 `ip link` commands to the system's global init script. The name of
1315 the global init script differs; for sysconfig, it is
1316 /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1318 For example, if you wanted to make a simple bond of two e100
1319 devices (presumed to be eth0 and eth1), and have it persist across
1320 reboots, edit the appropriate file (/etc/init.d/boot.local or
1321 /etc/rc.d/rc.local), and add the following::
1323 modprobe bonding mode=balance-alb miimon=100
1325 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1326 ip link set eth0 master bond0
1327 ip link set eth1 master bond0
1329 Replace the example bonding module parameters and bond0
1330 network configuration (IP address, netmask, etc) with the appropriate
1331 values for your configuration.
1333 Unfortunately, this method will not provide support for the
1334 ifup and ifdown scripts on the bond devices. To reload the bonding
1335 configuration, it is necessary to run the initialization script, e.g.,::
1337 # /etc/init.d/boot.local
1341 # /etc/rc.d/rc.local
1343 It may be desirable in such a case to create a separate script
1344 which only initializes the bonding configuration, then call that
1345 separate script from within boot.local. This allows for bonding to be
1346 enabled without re-running the entire global init script.
1348 To shut down the bonding devices, it is necessary to first
1349 mark the bonding device itself as being down, then remove the
1350 appropriate device driver modules. For our example above, you can do
1353 # ifconfig bond0 down
1357 Again, for convenience, it may be desirable to create a script
1358 with these commands.
1361 3.3.1 Configuring Multiple Bonds Manually
1362 -----------------------------------------
1364 This section contains information on configuring multiple
1365 bonding devices with differing options for those systems whose network
1366 initialization scripts lack support for configuring multiple bonds.
1368 If you require multiple bonding devices, but all with the same
1369 options, you may wish to use the "max_bonds" module parameter,
1372 To create multiple bonding devices with differing options, it is
1373 preferable to use bonding parameters exported by sysfs, documented in the
1376 For versions of bonding without sysfs support, the only means to
1377 provide multiple instances of bonding with differing options is to load
1378 the bonding driver multiple times. Note that current versions of the
1379 sysconfig network initialization scripts handle this automatically; if
1380 your distro uses these scripts, no special action is needed. See the
1381 section Configuring Bonding Devices, above, if you're not sure about your
1382 network initialization scripts.
1384 To load multiple instances of the module, it is necessary to
1385 specify a different name for each instance (the module loading system
1386 requires that every loaded module, even multiple instances of the same
1387 module, have a unique name). This is accomplished by supplying multiple
1388 sets of bonding options in ``/etc/modprobe.d/*.conf``, for example::
1391 options bond0 -o bond0 mode=balance-rr miimon=100
1394 options bond1 -o bond1 mode=balance-alb miimon=50
1396 will load the bonding module two times. The first instance is
1397 named "bond0" and creates the bond0 device in balance-rr mode with an
1398 miimon of 100. The second instance is named "bond1" and creates the
1399 bond1 device in balance-alb mode with an miimon of 50.
1401 In some circumstances (typically with older distributions),
1402 the above does not work, and the second bonding instance never sees
1403 its options. In that case, the second options line can be substituted
1406 install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1407 mode=balance-alb miimon=50
1409 This may be repeated any number of times, specifying a new and
1410 unique name in place of bond1 for each subsequent instance.
1412 It has been observed that some Red Hat supplied kernels are unable
1413 to rename modules at load time (the "-o bond1" part). Attempts to pass
1414 that option to modprobe will produce an "Operation not permitted" error.
1415 This has been reported on some Fedora Core kernels, and has been seen on
1416 RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1417 to configure multiple bonds with differing parameters (as they are older
1418 kernels, and also lack sysfs support).
1420 3.4 Configuring Bonding Manually via Sysfs
1421 ------------------------------------------
1423 Starting with version 3.0.0, Channel Bonding may be configured
1424 via the sysfs interface. This interface allows dynamic configuration
1425 of all bonds in the system without unloading the module. It also
1426 allows for adding and removing bonds at runtime. Ifenslave is no
1427 longer required, though it is still supported.
1429 Use of the sysfs interface allows you to use multiple bonds
1430 with different configurations without having to reload the module.
1431 It also allows you to use multiple, differently configured bonds when
1432 bonding is compiled into the kernel.
1434 You must have the sysfs filesystem mounted to configure
1435 bonding this way. The examples in this document assume that you
1436 are using the standard mount point for sysfs, e.g. /sys. If your
1437 sysfs filesystem is mounted elsewhere, you will need to adjust the
1438 example paths accordingly.
1440 Creating and Destroying Bonds
1441 -----------------------------
1442 To add a new bond foo::
1444 # echo +foo > /sys/class/net/bonding_masters
1446 To remove an existing bond bar::
1448 # echo -bar > /sys/class/net/bonding_masters
1450 To show all existing bonds::
1452 # cat /sys/class/net/bonding_masters
1456 due to 4K size limitation of sysfs files, this list may be
1457 truncated if you have more than a few hundred bonds. This is unlikely
1458 to occur under normal operating conditions.
1460 Adding and Removing Slaves
1461 --------------------------
1462 Interfaces may be enslaved to a bond using the file
1463 /sys/class/net/<bond>/bonding/slaves. The semantics for this file
1464 are the same as for the bonding_masters file.
1466 To enslave interface eth0 to bond bond0::
1469 # echo +eth0 > /sys/class/net/bond0/bonding/slaves
1471 To free slave eth0 from bond bond0::
1473 # echo -eth0 > /sys/class/net/bond0/bonding/slaves
1475 When an interface is enslaved to a bond, symlinks between the
1476 two are created in the sysfs filesystem. In this case, you would get
1477 /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1478 /sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1480 This means that you can tell quickly whether or not an
1481 interface is enslaved by looking for the master symlink. Thus:
1482 # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1483 will free eth0 from whatever bond it is enslaved to, regardless of
1484 the name of the bond interface.
1486 Changing a Bond's Configuration
1487 -------------------------------
1488 Each bond may be configured individually by manipulating the
1489 files located in /sys/class/net/<bond name>/bonding
1491 The names of these files correspond directly with the command-
1492 line parameters described elsewhere in this file, and, with the
1493 exception of arp_ip_target, they accept the same values. To see the
1494 current setting, simply cat the appropriate file.
1496 A few examples will be given here; for specific usage
1497 guidelines for each parameter, see the appropriate section in this
1500 To configure bond0 for balance-alb mode::
1502 # ifconfig bond0 down
1503 # echo 6 > /sys/class/net/bond0/bonding/mode
1505 # echo balance-alb > /sys/class/net/bond0/bonding/mode
1509 The bond interface must be down before the mode can be changed.
1511 To enable MII monitoring on bond0 with a 1 second interval::
1513 # echo 1000 > /sys/class/net/bond0/bonding/miimon
1517 If ARP monitoring is enabled, it will disabled when MII
1518 monitoring is enabled, and vice-versa.
1520 To add ARP targets::
1522 # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1523 # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1527 up to 16 target addresses may be specified.
1529 To remove an ARP target::
1531 # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1533 To configure the interval between learning packet transmits::
1535 # echo 12 > /sys/class/net/bond0/bonding/lp_interval
1539 the lp_interval is the number of seconds between instances where
1540 the bonding driver sends learning packets to each slaves peer switch. The
1541 default interval is 1 second.
1543 Example Configuration
1544 ---------------------
1545 We begin with the same example that is shown in section 3.3,
1546 executed with sysfs, and without using ifenslave.
1548 To make a simple bond of two e100 devices (presumed to be eth0
1549 and eth1), and have it persist across reboots, edit the appropriate
1550 file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1555 echo balance-alb > /sys/class/net/bond0/bonding/mode
1556 ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1557 echo 100 > /sys/class/net/bond0/bonding/miimon
1558 echo +eth0 > /sys/class/net/bond0/bonding/slaves
1559 echo +eth1 > /sys/class/net/bond0/bonding/slaves
1561 To add a second bond, with two e1000 interfaces in
1562 active-backup mode, using ARP monitoring, add the following lines to
1566 echo +bond1 > /sys/class/net/bonding_masters
1567 echo active-backup > /sys/class/net/bond1/bonding/mode
1568 ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1569 echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1570 echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1571 echo +eth2 > /sys/class/net/bond1/bonding/slaves
1572 echo +eth3 > /sys/class/net/bond1/bonding/slaves
1574 3.5 Configuration with Interfaces Support
1575 -----------------------------------------
1577 This section applies to distros which use /etc/network/interfaces file
1578 to describe network interface configuration, most notably Debian and it's
1581 The ifup and ifdown commands on Debian don't support bonding out of
1582 the box. The ifenslave-2.6 package should be installed to provide bonding
1583 support. Once installed, this package will provide ``bond-*`` options
1584 to be used into /etc/network/interfaces.
1586 Note that ifenslave-2.6 package will load the bonding module and use
1587 the ifenslave command when appropriate.
1589 Example Configurations
1590 ----------------------
1592 In /etc/network/interfaces, the following stanza will configure bond0, in
1593 active-backup mode, with eth0 and eth1 as slaves::
1596 iface bond0 inet dhcp
1597 bond-slaves eth0 eth1
1598 bond-mode active-backup
1600 bond-primary eth0 eth1
1602 If the above configuration doesn't work, you might have a system using
1603 upstart for system startup. This is most notably true for recent
1604 Ubuntu versions. The following stanza in /etc/network/interfaces will
1605 produce the same result on those systems::
1608 iface bond0 inet dhcp
1610 bond-mode active-backup
1614 iface eth0 inet manual
1616 bond-primary eth0 eth1
1619 iface eth1 inet manual
1621 bond-primary eth0 eth1
1623 For a full list of ``bond-*`` supported options in /etc/network/interfaces and
1624 some more advanced examples tailored to you particular distros, see the files in
1625 /usr/share/doc/ifenslave-2.6.
1627 3.6 Overriding Configuration for Special Cases
1628 ----------------------------------------------
1630 When using the bonding driver, the physical port which transmits a frame is
1631 typically selected by the bonding driver, and is not relevant to the user or
1632 system administrator. The output port is simply selected using the policies of
1633 the selected bonding mode. On occasion however, it is helpful to direct certain
1634 classes of traffic to certain physical interfaces on output to implement
1635 slightly more complex policies. For example, to reach a web server over a
1636 bonded interface in which eth0 connects to a private network, while eth1
1637 connects via a public network, it may be desirous to bias the bond to send said
1638 traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1639 can safely be sent over either interface. Such configurations may be achieved
1640 using the traffic control utilities inherent in linux.
1642 By default the bonding driver is multiqueue aware and 16 queues are created
1643 when the driver initializes (see Documentation/networking/multiqueue.rst
1644 for details). If more or less queues are desired the module parameter
1645 tx_queues can be used to change this value. There is no sysfs parameter
1646 available as the allocation is done at module init time.
1648 The output of the file /proc/net/bonding/bondX has changed so the output Queue
1649 ID is now printed for each slave::
1651 Bonding Mode: fault-tolerance (active-backup)
1653 Currently Active Slave: eth0
1655 MII Polling Interval (ms): 0
1659 Slave Interface: eth0
1661 Link Failure Count: 0
1662 Permanent HW addr: 00:1a:a0:12:8f:cb
1665 Slave Interface: eth1
1667 Link Failure Count: 0
1668 Permanent HW addr: 00:1a:a0:12:8f:cc
1671 The queue_id for a slave can be set using the command::
1673 # echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1675 Any interface that needs a queue_id set should set it with multiple calls
1676 like the one above until proper priorities are set for all interfaces. On
1677 distributions that allow configuration via initscripts, multiple 'queue_id'
1678 arguments can be added to BONDING_OPTS to set all needed slave queues.
1680 These queue id's can be used in conjunction with the tc utility to configure
1681 a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1682 slave devices. For instance, say we wanted, in the above configuration to
1683 force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1684 device. The following commands would accomplish this::
1686 # tc qdisc add dev bond0 handle 1 root multiq
1688 # tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip \
1689 dst 192.168.1.100 action skbedit queue_mapping 2
1691 These commands tell the kernel to attach a multiqueue queue discipline to the
1692 bond0 interface and filter traffic enqueued to it, such that packets with a dst
1693 ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1694 This value is then passed into the driver, causing the normal output path
1695 selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1697 Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver
1698 that normal output policy selection should take place. One benefit to simply
1699 leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1700 driver that is now present. This awareness allows tc filters to be placed on
1701 slave devices as well as bond devices and the bonding driver will simply act as
1702 a pass-through for selecting output queues on the slave device rather than
1703 output port selection.
1705 This feature first appeared in bonding driver version 3.7.0 and support for
1706 output slave selection was limited to round-robin and active-backup modes.
1708 3.7 Configuring LACP for 802.3ad mode in a more secure way
1709 ----------------------------------------------------------
1711 When using 802.3ad bonding mode, the Actor (host) and Partner (switch)
1712 exchange LACPDUs. These LACPDUs cannot be sniffed, because they are
1713 destined to link local mac addresses (which switches/bridges are not
1714 supposed to forward). However, most of the values are easily predictable
1715 or are simply the machine's MAC address (which is trivially known to all
1716 other hosts in the same L2). This implies that other machines in the L2
1717 domain can spoof LACPDU packets from other hosts to the switch and potentially
1718 cause mayhem by joining (from the point of view of the switch) another
1719 machine's aggregate, thus receiving a portion of that hosts incoming
1720 traffic and / or spoofing traffic from that machine themselves (potentially
1721 even successfully terminating some portion of flows). Though this is not
1722 a likely scenario, one could avoid this possibility by simply configuring
1723 few bonding parameters:
1725 (a) ad_actor_system : You can set a random mac-address that can be used for
1726 these LACPDU exchanges. The value can not be either NULL or Multicast.
1727 Also it's preferable to set the local-admin bit. Following shell code
1728 generates a random mac-address as described above::
1730 # sys_mac_addr=$(printf '%02x:%02x:%02x:%02x:%02x:%02x' \
1731 $(( (RANDOM & 0xFE) | 0x02 )) \
1732 $(( RANDOM & 0xFF )) \
1733 $(( RANDOM & 0xFF )) \
1734 $(( RANDOM & 0xFF )) \
1735 $(( RANDOM & 0xFF )) \
1736 $(( RANDOM & 0xFF )))
1737 # echo $sys_mac_addr > /sys/class/net/bond0/bonding/ad_actor_system
1739 (b) ad_actor_sys_prio : Randomize the system priority. The default value
1740 is 65535, but system can take the value from 1 - 65535. Following shell
1741 code generates random priority and sets it::
1743 # sys_prio=$(( 1 + RANDOM + RANDOM ))
1744 # echo $sys_prio > /sys/class/net/bond0/bonding/ad_actor_sys_prio
1746 (c) ad_user_port_key : Use the user portion of the port-key. The default
1747 keeps this empty. These are the upper 10 bits of the port-key and value
1748 ranges from 0 - 1023. Following shell code generates these 10 bits and
1751 # usr_port_key=$(( RANDOM & 0x3FF ))
1752 # echo $usr_port_key > /sys/class/net/bond0/bonding/ad_user_port_key
1755 4 Querying Bonding Configuration
1756 =================================
1758 4.1 Bonding Configuration
1759 -------------------------
1761 Each bonding device has a read-only file residing in the
1762 /proc/net/bonding directory. The file contents include information
1763 about the bonding configuration, options and state of each slave.
1765 For example, the contents of /proc/net/bonding/bond0 after the
1766 driver is loaded with parameters of mode=0 and miimon=1000 is
1767 generally as follows::
1769 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1770 Bonding Mode: load balancing (round-robin)
1771 Currently Active Slave: eth0
1773 MII Polling Interval (ms): 1000
1777 Slave Interface: eth1
1779 Link Failure Count: 1
1781 Slave Interface: eth0
1783 Link Failure Count: 1
1785 The precise format and contents will change depending upon the
1786 bonding configuration, state, and version of the bonding driver.
1788 4.2 Network configuration
1789 -------------------------
1791 The network configuration can be inspected using the ifconfig
1792 command. Bonding devices will have the MASTER flag set; Bonding slave
1793 devices will have the SLAVE flag set. The ifconfig output does not
1794 contain information on which slaves are associated with which masters.
1796 In the example below, the bond0 interface is the master
1797 (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1798 bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1799 TLB and ALB that require a unique MAC address for each slave::
1802 bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1803 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1804 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1805 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1806 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1807 collisions:0 txqueuelen:0
1809 eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1810 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1811 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1812 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1813 collisions:0 txqueuelen:100
1814 Interrupt:10 Base address:0x1080
1816 eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1817 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1818 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1819 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1820 collisions:0 txqueuelen:100
1821 Interrupt:9 Base address:0x1400
1823 5. Switch Configuration
1824 =======================
1826 For this section, "switch" refers to whatever system the
1827 bonded devices are directly connected to (i.e., where the other end of
1828 the cable plugs into). This may be an actual dedicated switch device,
1829 or it may be another regular system (e.g., another computer running
1832 The active-backup, balance-tlb and balance-alb modes do not
1833 require any specific configuration of the switch.
1835 The 802.3ad mode requires that the switch have the appropriate
1836 ports configured as an 802.3ad aggregation. The precise method used
1837 to configure this varies from switch to switch, but, for example, a
1838 Cisco 3550 series switch requires that the appropriate ports first be
1839 grouped together in a single etherchannel instance, then that
1840 etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1841 standard EtherChannel).
1843 The balance-rr, balance-xor and broadcast modes generally
1844 require that the switch have the appropriate ports grouped together.
1845 The nomenclature for such a group differs between switches, it may be
1846 called an "etherchannel" (as in the Cisco example, above), a "trunk
1847 group" or some other similar variation. For these modes, each switch
1848 will also have its own configuration options for the switch's transmit
1849 policy to the bond. Typical choices include XOR of either the MAC or
1850 IP addresses. The transmit policy of the two peers does not need to
1851 match. For these three modes, the bonding mode really selects a
1852 transmit policy for an EtherChannel group; all three will interoperate
1853 with another EtherChannel group.
1856 6. 802.1q VLAN Support
1857 ======================
1859 It is possible to configure VLAN devices over a bond interface
1860 using the 8021q driver. However, only packets coming from the 8021q
1861 driver and passing through bonding will be tagged by default. Self
1862 generated packets, for example, bonding's learning packets or ARP
1863 packets generated by either ALB mode or the ARP monitor mechanism, are
1864 tagged internally by bonding itself. As a result, bonding must
1865 "learn" the VLAN IDs configured above it, and use those IDs to tag
1866 self generated packets.
1868 For reasons of simplicity, and to support the use of adapters
1869 that can do VLAN hardware acceleration offloading, the bonding
1870 interface declares itself as fully hardware offloading capable, it gets
1871 the add_vid/kill_vid notifications to gather the necessary
1872 information, and it propagates those actions to the slaves. In case
1873 of mixed adapter types, hardware accelerated tagged packets that
1874 should go through an adapter that is not offloading capable are
1875 "un-accelerated" by the bonding driver so the VLAN tag sits in the
1878 VLAN interfaces *must* be added on top of a bonding interface
1879 only after enslaving at least one slave. The bonding interface has a
1880 hardware address of 00:00:00:00:00:00 until the first slave is added.
1881 If the VLAN interface is created prior to the first enslavement, it
1882 would pick up the all-zeroes hardware address. Once the first slave
1883 is attached to the bond, the bond device itself will pick up the
1884 slave's hardware address, which is then available for the VLAN device.
1886 Also, be aware that a similar problem can occur if all slaves
1887 are released from a bond that still has one or more VLAN interfaces on
1888 top of it. When a new slave is added, the bonding interface will
1889 obtain its hardware address from the first slave, which might not
1890 match the hardware address of the VLAN interfaces (which was
1891 ultimately copied from an earlier slave).
1893 There are two methods to insure that the VLAN device operates
1894 with the correct hardware address if all slaves are removed from a
1897 1. Remove all VLAN interfaces then recreate them
1899 2. Set the bonding interface's hardware address so that it
1900 matches the hardware address of the VLAN interfaces.
1902 Note that changing a VLAN interface's HW address would set the
1903 underlying device -- i.e. the bonding interface -- to promiscuous
1904 mode, which might not be what you want.
1910 The bonding driver at present supports two schemes for
1911 monitoring a slave device's link state: the ARP monitor and the MII
1914 At the present time, due to implementation restrictions in the
1915 bonding driver itself, it is not possible to enable both ARP and MII
1916 monitoring simultaneously.
1918 7.1 ARP Monitor Operation
1919 -------------------------
1921 The ARP monitor operates as its name suggests: it sends ARP
1922 queries to one or more designated peer systems on the network, and
1923 uses the response as an indication that the link is operating. This
1924 gives some assurance that traffic is actually flowing to and from one
1925 or more peers on the local network.
1927 The ARP monitor relies on the device driver itself to verify
1928 that traffic is flowing. In particular, the driver must keep up to
1929 date the last receive time, dev->last_rx. Drivers that use NETIF_F_LLTX
1930 flag must also update netdev_queue->trans_start. If they do not, then the
1931 ARP monitor will immediately fail any slaves using that driver, and
1932 those slaves will stay down. If networking monitoring (tcpdump, etc)
1933 shows the ARP requests and replies on the network, then it may be that
1934 your device driver is not updating last_rx and trans_start.
1936 7.2 Configuring Multiple ARP Targets
1937 ------------------------------------
1939 While ARP monitoring can be done with just one target, it can
1940 be useful in a High Availability setup to have several targets to
1941 monitor. In the case of just one target, the target itself may go
1942 down or have a problem making it unresponsive to ARP requests. Having
1943 an additional target (or several) increases the reliability of the ARP
1946 Multiple ARP targets must be separated by commas as follows::
1948 # example options for ARP monitoring with three targets
1950 options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1952 For just a single target the options would resemble::
1954 # example options for ARP monitoring with one target
1956 options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1959 7.3 MII Monitor Operation
1960 -------------------------
1962 The MII monitor monitors only the carrier state of the local
1963 network interface. It accomplishes this in one of three ways: by
1964 depending upon the device driver to maintain its carrier state, by
1965 querying the device's MII registers, or by making an ethtool query to
1968 If the use_carrier module parameter is 1 (the default value),
1969 then the MII monitor will rely on the driver for carrier state
1970 information (via the netif_carrier subsystem). As explained in the
1971 use_carrier parameter information, above, if the MII monitor fails to
1972 detect carrier loss on the device (e.g., when the cable is physically
1973 disconnected), it may be that the driver does not support
1976 If use_carrier is 0, then the MII monitor will first query the
1977 device's (via ioctl) MII registers and check the link state. If that
1978 request fails (not just that it returns carrier down), then the MII
1979 monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1980 the same information. If both methods fail (i.e., the driver either
1981 does not support or had some error in processing both the MII register
1982 and ethtool requests), then the MII monitor will assume the link is
1985 8. Potential Sources of Trouble
1986 ===============================
1988 8.1 Adventures in Routing
1989 -------------------------
1991 When bonding is configured, it is important that the slave
1992 devices not have routes that supersede routes of the master (or,
1993 generally, not have routes at all). For example, suppose the bonding
1994 device bond0 has two slaves, eth0 and eth1, and the routing table is
1997 Kernel IP routing table
1998 Destination Gateway Genmask Flags MSS Window irtt Iface
1999 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
2000 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
2001 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
2002 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
2004 This routing configuration will likely still update the
2005 receive/transmit times in the driver (needed by the ARP monitor), but
2006 may bypass the bonding driver (because outgoing traffic to, in this
2007 case, another host on network 10 would use eth0 or eth1 before bond0).
2009 The ARP monitor (and ARP itself) may become confused by this
2010 configuration, because ARP requests (generated by the ARP monitor)
2011 will be sent on one interface (bond0), but the corresponding reply
2012 will arrive on a different interface (eth0). This reply looks to ARP
2013 as an unsolicited ARP reply (because ARP matches replies on an
2014 interface basis), and is discarded. The MII monitor is not affected
2015 by the state of the routing table.
2017 The solution here is simply to insure that slaves do not have
2018 routes of their own, and if for some reason they must, those routes do
2019 not supersede routes of their master. This should generally be the
2020 case, but unusual configurations or errant manual or automatic static
2021 route additions may cause trouble.
2023 8.2 Ethernet Device Renaming
2024 ----------------------------
2026 On systems with network configuration scripts that do not
2027 associate physical devices directly with network interface names (so
2028 that the same physical device always has the same "ethX" name), it may
2029 be necessary to add some special logic to config files in
2032 For example, given a modules.conf containing the following::
2035 options bond0 mode=some-mode miimon=50
2041 If neither eth0 and eth1 are slaves to bond0, then when the
2042 bond0 interface comes up, the devices may end up reordered. This
2043 happens because bonding is loaded first, then its slave device's
2044 drivers are loaded next. Since no other drivers have been loaded,
2045 when the e1000 driver loads, it will receive eth0 and eth1 for its
2046 devices, but the bonding configuration tries to enslave eth2 and eth3
2047 (which may later be assigned to the tg3 devices).
2049 Adding the following::
2051 add above bonding e1000 tg3
2053 causes modprobe to load e1000 then tg3, in that order, when
2054 bonding is loaded. This command is fully documented in the
2055 modules.conf manual page.
2057 On systems utilizing modprobe an equivalent problem can occur.
2058 In this case, the following can be added to config files in
2059 /etc/modprobe.d/ as::
2061 softdep bonding pre: tg3 e1000
2063 This will load tg3 and e1000 modules before loading the bonding one.
2064 Full documentation on this can be found in the modprobe.d and modprobe
2067 8.3. Painfully Slow Or No Failed Link Detection By Miimon
2068 ---------------------------------------------------------
2070 By default, bonding enables the use_carrier option, which
2071 instructs bonding to trust the driver to maintain carrier state.
2073 As discussed in the options section, above, some drivers do
2074 not support the netif_carrier_on/_off link state tracking system.
2075 With use_carrier enabled, bonding will always see these links as up,
2076 regardless of their actual state.
2078 Additionally, other drivers do support netif_carrier, but do
2079 not maintain it in real time, e.g., only polling the link state at
2080 some fixed interval. In this case, miimon will detect failures, but
2081 only after some long period of time has expired. If it appears that
2082 miimon is very slow in detecting link failures, try specifying
2083 use_carrier=0 to see if that improves the failure detection time. If
2084 it does, then it may be that the driver checks the carrier state at a
2085 fixed interval, but does not cache the MII register values (so the
2086 use_carrier=0 method of querying the registers directly works). If
2087 use_carrier=0 does not improve the failover, then the driver may cache
2088 the registers, or the problem may be elsewhere.
2090 Also, remember that miimon only checks for the device's
2091 carrier state. It has no way to determine the state of devices on or
2092 beyond other ports of a switch, or if a switch is refusing to pass
2093 traffic while still maintaining carrier on.
2098 If running SNMP agents, the bonding driver should be loaded
2099 before any network drivers participating in a bond. This requirement
2100 is due to the interface index (ipAdEntIfIndex) being associated to
2101 the first interface found with a given IP address. That is, there is
2102 only one ipAdEntIfIndex for each IP address. For example, if eth0 and
2103 eth1 are slaves of bond0 and the driver for eth0 is loaded before the
2104 bonding driver, the interface for the IP address will be associated
2105 with the eth0 interface. This configuration is shown below, the IP
2106 address 192.168.1.1 has an interface index of 2 which indexes to eth0
2107 in the ifDescr table (ifDescr.2).
2111 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2112 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
2113 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
2114 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
2115 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
2116 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
2117 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
2118 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2119 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
2120 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2122 This problem is avoided by loading the bonding driver before
2123 any network drivers participating in a bond. Below is an example of
2124 loading the bonding driver first, the IP address 192.168.1.1 is
2125 correctly associated with ifDescr.2.
2127 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2128 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
2129 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
2130 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
2131 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
2132 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
2133 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
2134 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2135 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
2136 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2138 While some distributions may not report the interface name in
2139 ifDescr, the association between the IP address and IfIndex remains
2140 and SNMP functions such as Interface_Scan_Next will report that
2143 10. Promiscuous mode
2144 ====================
2146 When running network monitoring tools, e.g., tcpdump, it is
2147 common to enable promiscuous mode on the device, so that all traffic
2148 is seen (instead of seeing only traffic destined for the local host).
2149 The bonding driver handles promiscuous mode changes to the bonding
2150 master device (e.g., bond0), and propagates the setting to the slave
2153 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
2154 the promiscuous mode setting is propagated to all slaves.
2156 For the active-backup, balance-tlb and balance-alb modes, the
2157 promiscuous mode setting is propagated only to the active slave.
2159 For balance-tlb mode, the active slave is the slave currently
2160 receiving inbound traffic.
2162 For balance-alb mode, the active slave is the slave used as a
2163 "primary." This slave is used for mode-specific control traffic, for
2164 sending to peers that are unassigned or if the load is unbalanced.
2166 For the active-backup, balance-tlb and balance-alb modes, when
2167 the active slave changes (e.g., due to a link failure), the
2168 promiscuous setting will be propagated to the new active slave.
2170 11. Configuring Bonding for High Availability
2171 =============================================
2173 High Availability refers to configurations that provide
2174 maximum network availability by having redundant or backup devices,
2175 links or switches between the host and the rest of the world. The
2176 goal is to provide the maximum availability of network connectivity
2177 (i.e., the network always works), even though other configurations
2178 could provide higher throughput.
2180 11.1 High Availability in a Single Switch Topology
2181 --------------------------------------------------
2183 If two hosts (or a host and a single switch) are directly
2184 connected via multiple physical links, then there is no availability
2185 penalty to optimizing for maximum bandwidth. In this case, there is
2186 only one switch (or peer), so if it fails, there is no alternative
2187 access to fail over to. Additionally, the bonding load balance modes
2188 support link monitoring of their members, so if individual links fail,
2189 the load will be rebalanced across the remaining devices.
2191 See Section 12, "Configuring Bonding for Maximum Throughput"
2192 for information on configuring bonding with one peer device.
2194 11.2 High Availability in a Multiple Switch Topology
2195 ----------------------------------------------------
2197 With multiple switches, the configuration of bonding and the
2198 network changes dramatically. In multiple switch topologies, there is
2199 a trade off between network availability and usable bandwidth.
2201 Below is a sample network, configured to maximize the
2202 availability of the network::
2206 +-----+----+ +-----+----+
2207 | |port2 ISL port2| |
2208 | switch A +--------------------------+ switch B |
2210 +-----+----+ +-----++---+
2213 +-------------+ host1 +---------------+
2216 In this configuration, there is a link between the two
2217 switches (ISL, or inter switch link), and multiple ports connecting to
2218 the outside world ("port3" on each switch). There is no technical
2219 reason that this could not be extended to a third switch.
2221 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
2222 -------------------------------------------------------------
2224 In a topology such as the example above, the active-backup and
2225 broadcast modes are the only useful bonding modes when optimizing for
2226 availability; the other modes require all links to terminate on the
2227 same peer for them to behave rationally.
2230 This is generally the preferred mode, particularly if
2231 the switches have an ISL and play together well. If the
2232 network configuration is such that one switch is specifically
2233 a backup switch (e.g., has lower capacity, higher cost, etc),
2234 then the primary option can be used to insure that the
2235 preferred link is always used when it is available.
2238 This mode is really a special purpose mode, and is suitable
2239 only for very specific needs. For example, if the two
2240 switches are not connected (no ISL), and the networks beyond
2241 them are totally independent. In this case, if it is
2242 necessary for some specific one-way traffic to reach both
2243 independent networks, then the broadcast mode may be suitable.
2245 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
2246 ----------------------------------------------------------------
2248 The choice of link monitoring ultimately depends upon your
2249 switch. If the switch can reliably fail ports in response to other
2250 failures, then either the MII or ARP monitors should work. For
2251 example, in the above example, if the "port3" link fails at the remote
2252 end, the MII monitor has no direct means to detect this. The ARP
2253 monitor could be configured with a target at the remote end of port3,
2254 thus detecting that failure without switch support.
2256 In general, however, in a multiple switch topology, the ARP
2257 monitor can provide a higher level of reliability in detecting end to
2258 end connectivity failures (which may be caused by the failure of any
2259 individual component to pass traffic for any reason). Additionally,
2260 the ARP monitor should be configured with multiple targets (at least
2261 one for each switch in the network). This will insure that,
2262 regardless of which switch is active, the ARP monitor has a suitable
2265 Note, also, that of late many switches now support a functionality
2266 generally referred to as "trunk failover." This is a feature of the
2267 switch that causes the link state of a particular switch port to be set
2268 down (or up) when the state of another switch port goes down (or up).
2269 Its purpose is to propagate link failures from logically "exterior" ports
2270 to the logically "interior" ports that bonding is able to monitor via
2271 miimon. Availability and configuration for trunk failover varies by
2272 switch, but this can be a viable alternative to the ARP monitor when using
2275 12. Configuring Bonding for Maximum Throughput
2276 ==============================================
2278 12.1 Maximizing Throughput in a Single Switch Topology
2279 ------------------------------------------------------
2281 In a single switch configuration, the best method to maximize
2282 throughput depends upon the application and network environment. The
2283 various load balancing modes each have strengths and weaknesses in
2284 different environments, as detailed below.
2286 For this discussion, we will break down the topologies into
2287 two categories. Depending upon the destination of most traffic, we
2288 categorize them into either "gatewayed" or "local" configurations.
2290 In a gatewayed configuration, the "switch" is acting primarily
2291 as a router, and the majority of traffic passes through this router to
2292 other networks. An example would be the following::
2295 +----------+ +----------+
2296 | |eth0 port1| | to other networks
2297 | Host A +---------------------+ router +------------------->
2298 | +---------------------+ | Hosts B and C are out
2299 | |eth1 port2| | here somewhere
2300 +----------+ +----------+
2302 The router may be a dedicated router device, or another host
2303 acting as a gateway. For our discussion, the important point is that
2304 the majority of traffic from Host A will pass through the router to
2305 some other network before reaching its final destination.
2307 In a gatewayed network configuration, although Host A may
2308 communicate with many other systems, all of its traffic will be sent
2309 and received via one other peer on the local network, the router.
2311 Note that the case of two systems connected directly via
2312 multiple physical links is, for purposes of configuring bonding, the
2313 same as a gatewayed configuration. In that case, it happens that all
2314 traffic is destined for the "gateway" itself, not some other network
2317 In a local configuration, the "switch" is acting primarily as
2318 a switch, and the majority of traffic passes through this switch to
2319 reach other stations on the same network. An example would be the
2322 +----------+ +----------+ +--------+
2323 | |eth0 port1| +-------+ Host B |
2324 | Host A +------------+ switch |port3 +--------+
2325 | +------------+ | +--------+
2326 | |eth1 port2| +------------------+ Host C |
2327 +----------+ +----------+port4 +--------+
2330 Again, the switch may be a dedicated switch device, or another
2331 host acting as a gateway. For our discussion, the important point is
2332 that the majority of traffic from Host A is destined for other hosts
2333 on the same local network (Hosts B and C in the above example).
2335 In summary, in a gatewayed configuration, traffic to and from
2336 the bonded device will be to the same MAC level peer on the network
2337 (the gateway itself, i.e., the router), regardless of its final
2338 destination. In a local configuration, traffic flows directly to and
2339 from the final destinations, thus, each destination (Host B, Host C)
2340 will be addressed directly by their individual MAC addresses.
2342 This distinction between a gatewayed and a local network
2343 configuration is important because many of the load balancing modes
2344 available use the MAC addresses of the local network source and
2345 destination to make load balancing decisions. The behavior of each
2346 mode is described below.
2349 12.1.1 MT Bonding Mode Selection for Single Switch Topology
2350 -----------------------------------------------------------
2352 This configuration is the easiest to set up and to understand,
2353 although you will have to decide which bonding mode best suits your
2354 needs. The trade offs for each mode are detailed below:
2357 This mode is the only mode that will permit a single
2358 TCP/IP connection to stripe traffic across multiple
2359 interfaces. It is therefore the only mode that will allow a
2360 single TCP/IP stream to utilize more than one interface's
2361 worth of throughput. This comes at a cost, however: the
2362 striping generally results in peer systems receiving packets out
2363 of order, causing TCP/IP's congestion control system to kick
2364 in, often by retransmitting segments.
2366 It is possible to adjust TCP/IP's congestion limits by
2367 altering the net.ipv4.tcp_reordering sysctl parameter. The
2368 usual default value is 3. But keep in mind TCP stack is able
2369 to automatically increase this when it detects reorders.
2371 Note that the fraction of packets that will be delivered out of
2372 order is highly variable, and is unlikely to be zero. The level
2373 of reordering depends upon a variety of factors, including the
2374 networking interfaces, the switch, and the topology of the
2375 configuration. Speaking in general terms, higher speed network
2376 cards produce more reordering (due to factors such as packet
2377 coalescing), and a "many to many" topology will reorder at a
2378 higher rate than a "many slow to one fast" configuration.
2380 Many switches do not support any modes that stripe traffic
2381 (instead choosing a port based upon IP or MAC level addresses);
2382 for those devices, traffic for a particular connection flowing
2383 through the switch to a balance-rr bond will not utilize greater
2384 than one interface's worth of bandwidth.
2386 If you are utilizing protocols other than TCP/IP, UDP for
2387 example, and your application can tolerate out of order
2388 delivery, then this mode can allow for single stream datagram
2389 performance that scales near linearly as interfaces are added
2392 This mode requires the switch to have the appropriate ports
2393 configured for "etherchannel" or "trunking."
2396 There is not much advantage in this network topology to
2397 the active-backup mode, as the inactive backup devices are all
2398 connected to the same peer as the primary. In this case, a
2399 load balancing mode (with link monitoring) will provide the
2400 same level of network availability, but with increased
2401 available bandwidth. On the plus side, active-backup mode
2402 does not require any configuration of the switch, so it may
2403 have value if the hardware available does not support any of
2404 the load balance modes.
2407 This mode will limit traffic such that packets destined
2408 for specific peers will always be sent over the same
2409 interface. Since the destination is determined by the MAC
2410 addresses involved, this mode works best in a "local" network
2411 configuration (as described above), with destinations all on
2412 the same local network. This mode is likely to be suboptimal
2413 if all your traffic is passed through a single router (i.e., a
2414 "gatewayed" network configuration, as described above).
2416 As with balance-rr, the switch ports need to be configured for
2417 "etherchannel" or "trunking."
2420 Like active-backup, there is not much advantage to this
2421 mode in this type of network topology.
2424 This mode can be a good choice for this type of network
2425 topology. The 802.3ad mode is an IEEE standard, so all peers
2426 that implement 802.3ad should interoperate well. The 802.3ad
2427 protocol includes automatic configuration of the aggregates,
2428 so minimal manual configuration of the switch is needed
2429 (typically only to designate that some set of devices is
2430 available for 802.3ad). The 802.3ad standard also mandates
2431 that frames be delivered in order (within certain limits), so
2432 in general single connections will not see misordering of
2433 packets. The 802.3ad mode does have some drawbacks: the
2434 standard mandates that all devices in the aggregate operate at
2435 the same speed and duplex. Also, as with all bonding load
2436 balance modes other than balance-rr, no single connection will
2437 be able to utilize more than a single interface's worth of
2440 Additionally, the linux bonding 802.3ad implementation
2441 distributes traffic by peer (using an XOR of MAC addresses
2442 and packet type ID), so in a "gatewayed" configuration, all
2443 outgoing traffic will generally use the same device. Incoming
2444 traffic may also end up on a single device, but that is
2445 dependent upon the balancing policy of the peer's 802.3ad
2446 implementation. In a "local" configuration, traffic will be
2447 distributed across the devices in the bond.
2449 Finally, the 802.3ad mode mandates the use of the MII monitor,
2450 therefore, the ARP monitor is not available in this mode.
2453 The balance-tlb mode balances outgoing traffic by peer.
2454 Since the balancing is done according to MAC address, in a
2455 "gatewayed" configuration (as described above), this mode will
2456 send all traffic across a single device. However, in a
2457 "local" network configuration, this mode balances multiple
2458 local network peers across devices in a vaguely intelligent
2459 manner (not a simple XOR as in balance-xor or 802.3ad mode),
2460 so that mathematically unlucky MAC addresses (i.e., ones that
2461 XOR to the same value) will not all "bunch up" on a single
2464 Unlike 802.3ad, interfaces may be of differing speeds, and no
2465 special switch configuration is required. On the down side,
2466 in this mode all incoming traffic arrives over a single
2467 interface, this mode requires certain ethtool support in the
2468 network device driver of the slave interfaces, and the ARP
2469 monitor is not available.
2472 This mode is everything that balance-tlb is, and more.
2473 It has all of the features (and restrictions) of balance-tlb,
2474 and will also balance incoming traffic from local network
2475 peers (as described in the Bonding Module Options section,
2478 The only additional down side to this mode is that the network
2479 device driver must support changing the hardware address while
2482 12.1.2 MT Link Monitoring for Single Switch Topology
2483 ----------------------------------------------------
2485 The choice of link monitoring may largely depend upon which
2486 mode you choose to use. The more advanced load balancing modes do not
2487 support the use of the ARP monitor, and are thus restricted to using
2488 the MII monitor (which does not provide as high a level of end to end
2489 assurance as the ARP monitor).
2491 12.2 Maximum Throughput in a Multiple Switch Topology
2492 -----------------------------------------------------
2494 Multiple switches may be utilized to optimize for throughput
2495 when they are configured in parallel as part of an isolated network
2496 between two or more systems, for example::
2502 +--------+ | +---------+
2504 +------+---+ +-----+----+ +-----+----+
2505 | Switch A | | Switch B | | Switch C |
2506 +------+---+ +-----+----+ +-----+----+
2508 +--------+ | +---------+
2514 In this configuration, the switches are isolated from one
2515 another. One reason to employ a topology such as this is for an
2516 isolated network with many hosts (a cluster configured for high
2517 performance, for example), using multiple smaller switches can be more
2518 cost effective than a single larger switch, e.g., on a network with 24
2519 hosts, three 24 port switches can be significantly less expensive than
2520 a single 72 port switch.
2522 If access beyond the network is required, an individual host
2523 can be equipped with an additional network device connected to an
2524 external network; this host then additionally acts as a gateway.
2526 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2527 -------------------------------------------------------------
2529 In actual practice, the bonding mode typically employed in
2530 configurations of this type is balance-rr. Historically, in this
2531 network configuration, the usual caveats about out of order packet
2532 delivery are mitigated by the use of network adapters that do not do
2533 any kind of packet coalescing (via the use of NAPI, or because the
2534 device itself does not generate interrupts until some number of
2535 packets has arrived). When employed in this fashion, the balance-rr
2536 mode allows individual connections between two hosts to effectively
2537 utilize greater than one interface's bandwidth.
2539 12.2.2 MT Link Monitoring for Multiple Switch Topology
2540 ------------------------------------------------------
2542 Again, in actual practice, the MII monitor is most often used
2543 in this configuration, as performance is given preference over
2544 availability. The ARP monitor will function in this topology, but its
2545 advantages over the MII monitor are mitigated by the volume of probes
2546 needed as the number of systems involved grows (remember that each
2547 host in the network is configured with bonding).
2549 13. Switch Behavior Issues
2550 ==========================
2552 13.1 Link Establishment and Failover Delays
2553 -------------------------------------------
2555 Some switches exhibit undesirable behavior with regard to the
2556 timing of link up and down reporting by the switch.
2558 First, when a link comes up, some switches may indicate that
2559 the link is up (carrier available), but not pass traffic over the
2560 interface for some period of time. This delay is typically due to
2561 some type of autonegotiation or routing protocol, but may also occur
2562 during switch initialization (e.g., during recovery after a switch
2563 failure). If you find this to be a problem, specify an appropriate
2564 value to the updelay bonding module option to delay the use of the
2565 relevant interface(s).
2567 Second, some switches may "bounce" the link state one or more
2568 times while a link is changing state. This occurs most commonly while
2569 the switch is initializing. Again, an appropriate updelay value may
2572 Note that when a bonding interface has no active links, the
2573 driver will immediately reuse the first link that goes up, even if the
2574 updelay parameter has been specified (the updelay is ignored in this
2575 case). If there are slave interfaces waiting for the updelay timeout
2576 to expire, the interface that first went into that state will be
2577 immediately reused. This reduces down time of the network if the
2578 value of updelay has been overestimated, and since this occurs only in
2579 cases with no connectivity, there is no additional penalty for
2580 ignoring the updelay.
2582 In addition to the concerns about switch timings, if your
2583 switches take a long time to go into backup mode, it may be desirable
2584 to not activate a backup interface immediately after a link goes down.
2585 Failover may be delayed via the downdelay bonding module option.
2587 13.2 Duplicated Incoming Packets
2588 --------------------------------
2590 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2591 suppress duplicate packets, which should largely eliminate this problem.
2592 The following description is kept for reference.
2594 It is not uncommon to observe a short burst of duplicated
2595 traffic when the bonding device is first used, or after it has been
2596 idle for some period of time. This is most easily observed by issuing
2597 a "ping" to some other host on the network, and noticing that the
2598 output from ping flags duplicates (typically one per slave).
2600 For example, on a bond in active-backup mode with five slaves
2601 all connected to one switch, the output may appear as follows::
2604 PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
2605 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
2606 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2607 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2608 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2609 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2610 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
2611 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
2612 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2614 This is not due to an error in the bonding driver, rather, it
2615 is a side effect of how many switches update their MAC forwarding
2616 tables. Initially, the switch does not associate the MAC address in
2617 the packet with a particular switch port, and so it may send the
2618 traffic to all ports until its MAC forwarding table is updated. Since
2619 the interfaces attached to the bond may occupy multiple ports on a
2620 single switch, when the switch (temporarily) floods the traffic to all
2621 ports, the bond device receives multiple copies of the same packet
2622 (one per slave device).
2624 The duplicated packet behavior is switch dependent, some
2625 switches exhibit this, and some do not. On switches that display this
2626 behavior, it can be induced by clearing the MAC forwarding table (on
2627 most Cisco switches, the privileged command "clear mac address-table
2628 dynamic" will accomplish this).
2630 14. Hardware Specific Considerations
2631 ====================================
2633 This section contains additional information for configuring
2634 bonding on specific hardware platforms, or for interfacing bonding
2635 with particular switches or other devices.
2637 14.1 IBM BladeCenter
2638 --------------------
2640 This applies to the JS20 and similar systems.
2642 On the JS20 blades, the bonding driver supports only
2643 balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2644 largely due to the network topology inside the BladeCenter, detailed
2647 JS20 network adapter information
2648 --------------------------------
2650 All JS20s come with two Broadcom Gigabit Ethernet ports
2651 integrated on the planar (that's "motherboard" in IBM-speak). In the
2652 BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2653 I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2654 An add-on Broadcom daughter card can be installed on a JS20 to provide
2655 two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2656 wired to I/O Modules 3 and 4, respectively.
2658 Each I/O Module may contain either a switch or a passthrough
2659 module (which allows ports to be directly connected to an external
2660 switch). Some bonding modes require a specific BladeCenter internal
2661 network topology in order to function; these are detailed below.
2663 Additional BladeCenter-specific networking information can be
2664 found in two IBM Redbooks (www.ibm.com/redbooks):
2666 - "IBM eServer BladeCenter Networking Options"
2667 - "IBM eServer BladeCenter Layer 2-7 Network Switching"
2669 BladeCenter networking configuration
2670 ------------------------------------
2672 Because a BladeCenter can be configured in a very large number
2673 of ways, this discussion will be confined to describing basic
2676 Normally, Ethernet Switch Modules (ESMs) are used in I/O
2677 modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2678 JS20 will be connected to different internal switches (in the
2679 respective I/O modules).
2681 A passthrough module (OPM or CPM, optical or copper,
2682 passthrough module) connects the I/O module directly to an external
2683 switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2684 interfaces of a JS20 can be redirected to the outside world and
2685 connected to a common external switch.
2687 Depending upon the mix of ESMs and PMs, the network will
2688 appear to bonding as either a single switch topology (all PMs) or as a
2689 multiple switch topology (one or more ESMs, zero or more PMs). It is
2690 also possible to connect ESMs together, resulting in a configuration
2691 much like the example in "High Availability in a Multiple Switch
2694 Requirements for specific modes
2695 -------------------------------
2697 The balance-rr mode requires the use of passthrough modules
2698 for devices in the bond, all connected to an common external switch.
2699 That switch must be configured for "etherchannel" or "trunking" on the
2700 appropriate ports, as is usual for balance-rr.
2702 The balance-alb and balance-tlb modes will function with
2703 either switch modules or passthrough modules (or a mix). The only
2704 specific requirement for these modes is that all network interfaces
2705 must be able to reach all destinations for traffic sent over the
2706 bonding device (i.e., the network must converge at some point outside
2709 The active-backup mode has no additional requirements.
2711 Link monitoring issues
2712 ----------------------
2714 When an Ethernet Switch Module is in place, only the ARP
2715 monitor will reliably detect link loss to an external switch. This is
2716 nothing unusual, but examination of the BladeCenter cabinet would
2717 suggest that the "external" network ports are the ethernet ports for
2718 the system, when it fact there is a switch between these "external"
2719 ports and the devices on the JS20 system itself. The MII monitor is
2720 only able to detect link failures between the ESM and the JS20 system.
2722 When a passthrough module is in place, the MII monitor does
2723 detect failures to the "external" port, which is then directly
2724 connected to the JS20 system.
2729 The Serial Over LAN (SoL) link is established over the primary
2730 ethernet (eth0) only, therefore, any loss of link to eth0 will result
2731 in losing your SoL connection. It will not fail over with other
2732 network traffic, as the SoL system is beyond the control of the
2735 It may be desirable to disable spanning tree on the switch
2736 (either the internal Ethernet Switch Module, or an external switch) to
2737 avoid fail-over delay issues when using bonding.
2740 15. Frequently Asked Questions
2741 ==============================
2746 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2747 The new driver was designed to be SMP safe from the start.
2749 2. What type of cards will work with it?
2750 -----------------------------------------
2752 Any Ethernet type cards (you can even mix cards - a Intel
2753 EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2754 devices need not be of the same speed.
2756 Starting with version 3.2.1, bonding also supports Infiniband
2757 slaves in active-backup mode.
2759 3. How many bonding devices can I have?
2760 ----------------------------------------
2764 4. How many slaves can a bonding device have?
2765 ----------------------------------------------
2767 This is limited only by the number of network interfaces Linux
2768 supports and/or the number of network cards you can place in your
2771 5. What happens when a slave link dies?
2772 ----------------------------------------
2774 If link monitoring is enabled, then the failing device will be
2775 disabled. The active-backup mode will fail over to a backup link, and
2776 other modes will ignore the failed link. The link will continue to be
2777 monitored, and should it recover, it will rejoin the bond (in whatever
2778 manner is appropriate for the mode). See the sections on High
2779 Availability and the documentation for each mode for additional
2782 Link monitoring can be enabled via either the miimon or
2783 arp_interval parameters (described in the module parameters section,
2784 above). In general, miimon monitors the carrier state as sensed by
2785 the underlying network device, and the arp monitor (arp_interval)
2786 monitors connectivity to another host on the local network.
2788 If no link monitoring is configured, the bonding driver will
2789 be unable to detect link failures, and will assume that all links are
2790 always available. This will likely result in lost packets, and a
2791 resulting degradation of performance. The precise performance loss
2792 depends upon the bonding mode and network configuration.
2794 6. Can bonding be used for High Availability?
2795 ----------------------------------------------
2797 Yes. See the section on High Availability for details.
2799 7. Which switches/systems does it work with?
2800 ---------------------------------------------
2802 The full answer to this depends upon the desired mode.
2804 In the basic balance modes (balance-rr and balance-xor), it
2805 works with any system that supports etherchannel (also called
2806 trunking). Most managed switches currently available have such
2807 support, and many unmanaged switches as well.
2809 The advanced balance modes (balance-tlb and balance-alb) do
2810 not have special switch requirements, but do need device drivers that
2811 support specific features (described in the appropriate section under
2812 module parameters, above).
2814 In 802.3ad mode, it works with systems that support IEEE
2815 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2816 switches currently available support 802.3ad.
2818 The active-backup mode should work with any Layer-II switch.
2820 8. Where does a bonding device get its MAC address from?
2821 ---------------------------------------------------------
2823 When using slave devices that have fixed MAC addresses, or when
2824 the fail_over_mac option is enabled, the bonding device's MAC address is
2825 the MAC address of the active slave.
2827 For other configurations, if not explicitly configured (with
2828 ifconfig or ip link), the MAC address of the bonding device is taken from
2829 its first slave device. This MAC address is then passed to all following
2830 slaves and remains persistent (even if the first slave is removed) until
2831 the bonding device is brought down or reconfigured.
2833 If you wish to change the MAC address, you can set it with
2834 ifconfig or ip link::
2836 # ifconfig bond0 hw ether 00:11:22:33:44:55
2838 # ip link set bond0 address 66:77:88:99:aa:bb
2840 The MAC address can be also changed by bringing down/up the
2841 device and then changing its slaves (or their order)::
2843 # ifconfig bond0 down ; modprobe -r bonding
2844 # ifconfig bond0 .... up
2845 # ifenslave bond0 eth...
2847 This method will automatically take the address from the next
2848 slave that is added.
2850 To restore your slaves' MAC addresses, you need to detach them
2851 from the bond (``ifenslave -d bond0 eth0``). The bonding driver will
2852 then restore the MAC addresses that the slaves had before they were
2855 16. Resources and Links
2856 =======================
2858 The latest version of the bonding driver can be found in the latest
2859 version of the linux kernel, found on http://kernel.org
2861 The latest version of this document can be found in the latest kernel
2862 source (named Documentation/networking/bonding.rst).
2864 Discussions regarding the development of the bonding driver take place
2865 on the main Linux network mailing list, hosted at vger.kernel.org. The list
2868 netdev@vger.kernel.org
2870 The administrative interface (to subscribe or unsubscribe) can
2873 http://vger.kernel.org/vger-lists.html#netdev