About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Documentation / networking / bonding.txt




Custom Search

Based on kernel version 3.15.4. Page generated on 2014-07-07 09:03 EST.

1	
2			Linux Ethernet Bonding Driver HOWTO
3	
4			Latest update: 27 April 2011
5	
6	Initial release : Thomas Davis <tadavis at lbl.gov>
7	Corrections, HA extensions : 2000/10/03-15 :
8	  - Willy Tarreau <willy at meta-x.org>
9	  - Constantine Gavrilov <const-g at xpert.com>
10	  - Chad N. Tindel <ctindel at ieee dot org>
11	  - Janice Girouard <girouard at us dot ibm dot com>
12	  - Jay Vosburgh <fubar at us dot ibm dot com>
13	
14	Reorganized and updated Feb 2005 by Jay Vosburgh
15	Added Sysfs information: 2006/04/24
16	  - Mitch Williams <mitch.a.williams at intel.com>
17	
18	Introduction
19	============
20	
21		The Linux bonding driver provides a method for aggregating
22	multiple network interfaces into a single logical "bonded" interface.
23	The behavior of the bonded interfaces depends upon the mode; generally
24	speaking, modes provide either hot standby or load balancing services.
25	Additionally, link integrity monitoring may be performed.
26		
27		The bonding driver originally came from Donald Becker's
28	beowulf patches for kernel 2.0. It has changed quite a bit since, and
29	the original tools from extreme-linux and beowulf sites will not work
30	with this version of the driver.
31	
32		For new versions of the driver, updated userspace tools, and
33	who to ask for help, please follow the links at the end of this file.
34	
35	Table of Contents
36	=================
37	
38	1. Bonding Driver Installation
39	
40	2. Bonding Driver Options
41	
42	3. Configuring Bonding Devices
43	3.1	Configuration with Sysconfig Support
44	3.1.1		Using DHCP with Sysconfig
45	3.1.2		Configuring Multiple Bonds with Sysconfig
46	3.2	Configuration with Initscripts Support
47	3.2.1		Using DHCP with Initscripts
48	3.2.2		Configuring Multiple Bonds with Initscripts
49	3.3	Configuring Bonding Manually with Ifenslave
50	3.3.1		Configuring Multiple Bonds Manually
51	3.4	Configuring Bonding Manually via Sysfs
52	3.5	Configuration with Interfaces Support
53	3.6	Overriding Configuration for Special Cases
54	
55	4. Querying Bonding Configuration
56	4.1	Bonding Configuration
57	4.2	Network Configuration
58	
59	5. Switch Configuration
60	
61	6. 802.1q VLAN Support
62	
63	7. Link Monitoring
64	7.1	ARP Monitor Operation
65	7.2	Configuring Multiple ARP Targets
66	7.3	MII Monitor Operation
67	
68	8. Potential Trouble Sources
69	8.1	Adventures in Routing
70	8.2	Ethernet Device Renaming
71	8.3	Painfully Slow Or No Failed Link Detection By Miimon
72	
73	9. SNMP agents
74	
75	10. Promiscuous mode
76	
77	11. Configuring Bonding for High Availability
78	11.1	High Availability in a Single Switch Topology
79	11.2	High Availability in a Multiple Switch Topology
80	11.2.1		HA Bonding Mode Selection for Multiple Switch Topology
81	11.2.2		HA Link Monitoring for Multiple Switch Topology
82	
83	12. Configuring Bonding for Maximum Throughput
84	12.1	Maximum Throughput in a Single Switch Topology
85	12.1.1		MT Bonding Mode Selection for Single Switch Topology
86	12.1.2		MT Link Monitoring for Single Switch Topology
87	12.2	Maximum Throughput in a Multiple Switch Topology
88	12.2.1		MT Bonding Mode Selection for Multiple Switch Topology
89	12.2.2		MT Link Monitoring for Multiple Switch Topology
90	
91	13. Switch Behavior Issues
92	13.1	Link Establishment and Failover Delays
93	13.2	Duplicated Incoming Packets
94	
95	14. Hardware Specific Considerations
96	14.1	IBM BladeCenter
97	
98	15. Frequently Asked Questions
99	
100	16. Resources and Links
101	
102	
103	1. Bonding Driver Installation
104	==============================
105	
106		Most popular distro kernels ship with the bonding driver
107	already available as a module. If your distro does not, or you
108	have need to compile bonding from source (e.g., configuring and
109	installing a mainline kernel from kernel.org), you'll need to perform
110	the following steps:
111	
112	1.1 Configure and build the kernel with bonding
113	-----------------------------------------------
114	
115		The current version of the bonding driver is available in the
116	drivers/net/bonding subdirectory of the most recent kernel source
117	(which is available on http://kernel.org).  Most users "rolling their
118	own" will want to use the most recent kernel from kernel.org.
119	
120		Configure kernel with "make menuconfig" (or "make xconfig" or
121	"make config"), then select "Bonding driver support" in the "Network
122	device support" section.  It is recommended that you configure the
123	driver as module since it is currently the only way to pass parameters
124	to the driver or configure more than one bonding device.
125	
126		Build and install the new kernel and modules.
127	
128	1.2 Bonding Control Utility
129	-------------------------------------
130	
131		 It is recommended to configure bonding via iproute2 (netlink)
132	or sysfs, the old ifenslave control utility is obsolete.
133	
134	2. Bonding Driver Options
135	=========================
136	
137		Options for the bonding driver are supplied as parameters to the
138	bonding module at load time, or are specified via sysfs.
139	
140		Module options may be given as command line arguments to the
141	insmod or modprobe command, but are usually specified in either the
142	/etc/modrobe.d/*.conf configuration files, or in a distro-specific
143	configuration file (some of which are detailed in the next section).
144	
145		Details on bonding support for sysfs is provided in the
146	"Configuring Bonding Manually via Sysfs" section, below.
147	
148		The available bonding driver parameters are listed below. If a
149	parameter is not specified the default value is used.  When initially
150	configuring a bond, it is recommended "tail -f /var/log/messages" be
151	run in a separate window to watch for bonding driver error messages.
152	
153		It is critical that either the miimon or arp_interval and
154	arp_ip_target parameters be specified, otherwise serious network
155	degradation will occur during link failures.  Very few devices do not
156	support at least miimon, so there is really no reason not to use it.
157	
158		Options with textual values will accept either the text name
159	or, for backwards compatibility, the option value.  E.g.,
160	"mode=802.3ad" and "mode=4" set the same mode.
161	
162		The parameters are as follows:
163	
164	active_slave
165	
166		Specifies the new active slave for modes that support it
167		(active-backup, balance-alb and balance-tlb).  Possible values
168		are the name of any currently enslaved interface, or an empty
169		string.  If a name is given, the slave and its link must be up in order
170		to be selected as the new active slave.  If an empty string is
171		specified, the current active slave is cleared, and a new active
172		slave is selected automatically.
173	
174		Note that this is only available through the sysfs interface. No module
175		parameter by this name exists.
176	
177		The normal value of this option is the name of the currently
178		active slave, or the empty string if there is no active slave or
179		the current mode does not use an active slave.
180	
181	ad_select
182	
183		Specifies the 802.3ad aggregation selection logic to use.  The
184		possible values and their effects are:
185	
186		stable or 0
187	
188			The active aggregator is chosen by largest aggregate
189			bandwidth.
190	
191			Reselection of the active aggregator occurs only when all
192			slaves of the active aggregator are down or the active
193			aggregator has no slaves.
194	
195			This is the default value.
196	
197		bandwidth or 1
198	
199			The active aggregator is chosen by largest aggregate
200			bandwidth.  Reselection occurs if:
201	
202			- A slave is added to or removed from the bond
203	
204			- Any slave's link state changes
205	
206			- Any slave's 802.3ad association state changes
207	
208			- The bond's administrative state changes to up
209	
210		count or 2
211	
212			The active aggregator is chosen by the largest number of
213			ports (slaves).  Reselection occurs as described under the
214			"bandwidth" setting, above.
215	
216		The bandwidth and count selection policies permit failover of
217		802.3ad aggregations when partial failure of the active aggregator
218		occurs.  This keeps the aggregator with the highest availability
219		(either in bandwidth or in number of ports) active at all times.
220	
221		This option was added in bonding version 3.4.0.
222	
223	all_slaves_active
224	
225		Specifies that duplicate frames (received on inactive ports) should be
226		dropped (0) or delivered (1).
227	
228		Normally, bonding will drop duplicate frames (received on inactive
229		ports), which is desirable for most users. But there are some times
230		it is nice to allow duplicate frames to be delivered.
231	
232		The default value is 0 (drop duplicate frames received on inactive
233		ports).
234	
235	arp_interval
236	
237		Specifies the ARP link monitoring frequency in milliseconds.
238	
239		The ARP monitor works by periodically checking the slave
240		devices to determine whether they have sent or received
241		traffic recently (the precise criteria depends upon the
242		bonding mode, and the state of the slave).  Regular traffic is
243		generated via ARP probes issued for the addresses specified by
244		the arp_ip_target option.
245	
246		This behavior can be modified by the arp_validate option,
247		below.
248	
249		If ARP monitoring is used in an etherchannel compatible mode
250		(modes 0 and 2), the switch should be configured in a mode
251		that evenly distributes packets across all links. If the
252		switch is configured to distribute the packets in an XOR
253		fashion, all replies from the ARP targets will be received on
254		the same link which could cause the other team members to
255		fail.  ARP monitoring should not be used in conjunction with
256		miimon.  A value of 0 disables ARP monitoring.  The default
257		value is 0.
258	
259	arp_ip_target
260	
261		Specifies the IP addresses to use as ARP monitoring peers when
262		arp_interval is > 0.  These are the targets of the ARP request
263		sent to determine the health of the link to the targets.
264		Specify these values in ddd.ddd.ddd.ddd format.  Multiple IP
265		addresses must be separated by a comma.  At least one IP
266		address must be given for ARP monitoring to function.  The
267		maximum number of targets that can be specified is 16.  The
268		default value is no IP addresses.
269	
270	arp_validate
271	
272		Specifies whether or not ARP probes and replies should be
273		validated in any mode that supports arp monitoring, or whether
274		non-ARP traffic should be filtered (disregarded) for link
275		monitoring purposes.
276	
277		Possible values are:
278	
279		none or 0
280	
281			No validation or filtering is performed.
282	
283		active or 1
284	
285			Validation is performed only for the active slave.
286	
287		backup or 2
288	
289			Validation is performed only for backup slaves.
290	
291		all or 3
292	
293			Validation is performed for all slaves.
294	
295		filter or 4
296	
297			Filtering is applied to all slaves. No validation is
298			performed.
299	
300		filter_active or 5
301	
302			Filtering is applied to all slaves, validation is performed
303			only for the active slave.
304	
305		filter_backup or 6
306	
307			Filtering is applied to all slaves, validation is performed
308			only for backup slaves.
309	
310		Validation:
311	
312		Enabling validation causes the ARP monitor to examine the incoming
313		ARP requests and replies, and only consider a slave to be up if it
314		is receiving the appropriate ARP traffic.
315	
316		For an active slave, the validation checks ARP replies to confirm
317		that they were generated by an arp_ip_target.  Since backup slaves
318		do not typically receive these replies, the validation performed
319		for backup slaves is on the broadcast ARP request sent out via the
320		active slave.  It is possible that some switch or network
321		configurations may result in situations wherein the backup slaves
322		do not receive the ARP requests; in such a situation, validation
323		of backup slaves must be disabled.
324	
325		The validation of ARP requests on backup slaves is mainly helping
326		bonding to decide which slaves are more likely to work in case of
327		the active slave failure, it doesn't really guarantee that the
328		backup slave will work if it's selected as the next active slave.
329	
330		Validation is useful in network configurations in which multiple
331		bonding hosts are concurrently issuing ARPs to one or more targets
332		beyond a common switch.  Should the link between the switch and
333		target fail (but not the switch itself), the probe traffic
334		generated by the multiple bonding instances will fool the standard
335		ARP monitor into considering the links as still up.  Use of
336		validation can resolve this, as the ARP monitor will only consider
337		ARP requests and replies associated with its own instance of
338		bonding.
339	
340		Filtering:
341	
342		Enabling filtering causes the ARP monitor to only use incoming ARP
343		packets for link availability purposes.  Arriving packets that are
344		not ARPs are delivered normally, but do not count when determining
345		if a slave is available.
346	
347		Filtering operates by only considering the reception of ARP
348		packets (any ARP packet, regardless of source or destination) when
349		determining if a slave has received traffic for link availability
350		purposes.
351	
352		Filtering is useful in network configurations in which significant
353		levels of third party broadcast traffic would fool the standard
354		ARP monitor into considering the links as still up.  Use of
355		filtering can resolve this, as only ARP traffic is considered for
356		link availability purposes.
357	
358		This option was added in bonding version 3.1.0.
359	
360	arp_all_targets
361	
362		Specifies the quantity of arp_ip_targets that must be reachable
363		in order for the ARP monitor to consider a slave as being up.
364		This option affects only active-backup mode for slaves with
365		arp_validation enabled.
366	
367		Possible values are:
368	
369		any or 0
370	
371			consider the slave up only when any of the arp_ip_targets
372			is reachable
373	
374		all or 1
375	
376			consider the slave up only when all of the arp_ip_targets
377			are reachable
378	
379	downdelay
380	
381		Specifies the time, in milliseconds, to wait before disabling
382		a slave after a link failure has been detected.  This option
383		is only valid for the miimon link monitor.  The downdelay
384		value should be a multiple of the miimon value; if not, it
385		will be rounded down to the nearest multiple.  The default
386		value is 0.
387	
388	fail_over_mac
389	
390		Specifies whether active-backup mode should set all slaves to
391		the same MAC address at enslavement (the traditional
392		behavior), or, when enabled, perform special handling of the
393		bond's MAC address in accordance with the selected policy.
394	
395		Possible values are:
396	
397		none or 0
398	
399			This setting disables fail_over_mac, and causes
400			bonding to set all slaves of an active-backup bond to
401			the same MAC address at enslavement time.  This is the
402			default.
403	
404		active or 1
405	
406			The "active" fail_over_mac policy indicates that the
407			MAC address of the bond should always be the MAC
408			address of the currently active slave.  The MAC
409			address of the slaves is not changed; instead, the MAC
410			address of the bond changes during a failover.
411	
412			This policy is useful for devices that cannot ever
413			alter their MAC address, or for devices that refuse
414			incoming broadcasts with their own source MAC (which
415			interferes with the ARP monitor).
416	
417			The down side of this policy is that every device on
418			the network must be updated via gratuitous ARP,
419			vs. just updating a switch or set of switches (which
420			often takes place for any traffic, not just ARP
421			traffic, if the switch snoops incoming traffic to
422			update its tables) for the traditional method.  If the
423			gratuitous ARP is lost, communication may be
424			disrupted.
425	
426			When this policy is used in conjunction with the mii
427			monitor, devices which assert link up prior to being
428			able to actually transmit and receive are particularly
429			susceptible to loss of the gratuitous ARP, and an
430			appropriate updelay setting may be required.
431	
432		follow or 2
433	
434			The "follow" fail_over_mac policy causes the MAC
435			address of the bond to be selected normally (normally
436			the MAC address of the first slave added to the bond).
437			However, the second and subsequent slaves are not set
438			to this MAC address while they are in a backup role; a
439			slave is programmed with the bond's MAC address at
440			failover time (and the formerly active slave receives
441			the newly active slave's MAC address).
442	
443			This policy is useful for multiport devices that
444			either become confused or incur a performance penalty
445			when multiple ports are programmed with the same MAC
446			address.
447	
448	
449		The default policy is none, unless the first slave cannot
450		change its MAC address, in which case the active policy is
451		selected by default.
452	
453		This option may be modified via sysfs only when no slaves are
454		present in the bond.
455	
456		This option was added in bonding version 3.2.0.  The "follow"
457		policy was added in bonding version 3.3.0.
458	
459	lacp_rate
460	
461		Option specifying the rate in which we'll ask our link partner
462		to transmit LACPDU packets in 802.3ad mode.  Possible values
463		are:
464	
465		slow or 0
466			Request partner to transmit LACPDUs every 30 seconds
467	
468		fast or 1
469			Request partner to transmit LACPDUs every 1 second
470	
471		The default is slow.
472	
473	max_bonds
474	
475		Specifies the number of bonding devices to create for this
476		instance of the bonding driver.  E.g., if max_bonds is 3, and
477		the bonding driver is not already loaded, then bond0, bond1
478		and bond2 will be created.  The default value is 1.  Specifying
479		a value of 0 will load bonding, but will not create any devices.
480	
481	miimon
482	
483		Specifies the MII link monitoring frequency in milliseconds.
484		This determines how often the link state of each slave is
485		inspected for link failures.  A value of zero disables MII
486		link monitoring.  A value of 100 is a good starting point.
487		The use_carrier option, below, affects how the link state is
488		determined.  See the High Availability section for additional
489		information.  The default value is 0.
490	
491	min_links
492	
493		Specifies the minimum number of links that must be active before
494		asserting carrier. It is similar to the Cisco EtherChannel min-links
495		feature. This allows setting the minimum number of member ports that
496		must be up (link-up state) before marking the bond device as up
497		(carrier on). This is useful for situations where higher level services
498		such as clustering want to ensure a minimum number of low bandwidth
499		links are active before switchover. This option only affect 802.3ad
500		mode.
501	
502		The default value is 0. This will cause carrier to be asserted (for
503		802.3ad mode) whenever there is an active aggregator, regardless of the
504		number of available links in that aggregator. Note that, because an
505		aggregator cannot be active without at least one available link,
506		setting this option to 0 or to 1 has the exact same effect.
507	
508	mode
509	
510		Specifies one of the bonding policies. The default is
511		balance-rr (round robin).  Possible values are:
512	
513		balance-rr or 0
514	
515			Round-robin policy: Transmit packets in sequential
516			order from the first available slave through the
517			last.  This mode provides load balancing and fault
518			tolerance.
519	
520		active-backup or 1
521	
522			Active-backup policy: Only one slave in the bond is
523			active.  A different slave becomes active if, and only
524			if, the active slave fails.  The bond's MAC address is
525			externally visible on only one port (network adapter)
526			to avoid confusing the switch.
527	
528			In bonding version 2.6.2 or later, when a failover
529			occurs in active-backup mode, bonding will issue one
530			or more gratuitous ARPs on the newly active slave.
531			One gratuitous ARP is issued for the bonding master
532			interface and each VLAN interfaces configured above
533			it, provided that the interface has at least one IP
534			address configured.  Gratuitous ARPs issued for VLAN
535			interfaces are tagged with the appropriate VLAN id.
536	
537			This mode provides fault tolerance.  The primary
538			option, documented below, affects the behavior of this
539			mode.
540	
541		balance-xor or 2
542	
543			XOR policy: Transmit based on the selected transmit
544			hash policy.  The default policy is a simple [(source
545			MAC address XOR'd with destination MAC address) modulo
546			slave count].  Alternate transmit policies may be
547			selected via the xmit_hash_policy option, described
548			below.
549	
550			This mode provides load balancing and fault tolerance.
551	
552		broadcast or 3
553	
554			Broadcast policy: transmits everything on all slave
555			interfaces.  This mode provides fault tolerance.
556	
557		802.3ad or 4
558	
559			IEEE 802.3ad Dynamic link aggregation.  Creates
560			aggregation groups that share the same speed and
561			duplex settings.  Utilizes all slaves in the active
562			aggregator according to the 802.3ad specification.
563	
564			Slave selection for outgoing traffic is done according
565			to the transmit hash policy, which may be changed from
566			the default simple XOR policy via the xmit_hash_policy
567			option, documented below.  Note that not all transmit
568			policies may be 802.3ad compliant, particularly in
569			regards to the packet mis-ordering requirements of
570			section 43.2.4 of the 802.3ad standard.  Differing
571			peer implementations will have varying tolerances for
572			noncompliance.
573	
574			Prerequisites:
575	
576			1. Ethtool support in the base drivers for retrieving
577			the speed and duplex of each slave.
578	
579			2. A switch that supports IEEE 802.3ad Dynamic link
580			aggregation.
581	
582			Most switches will require some type of configuration
583			to enable 802.3ad mode.
584	
585		balance-tlb or 5
586	
587			Adaptive transmit load balancing: channel bonding that
588			does not require any special switch support.  The
589			outgoing traffic is distributed according to the
590			current load (computed relative to the speed) on each
591			slave.  Incoming traffic is received by the current
592			slave.  If the receiving slave fails, another slave
593			takes over the MAC address of the failed receiving
594			slave.
595	
596			Prerequisite:
597	
598			Ethtool support in the base drivers for retrieving the
599			speed of each slave.
600	
601		balance-alb or 6
602	
603			Adaptive load balancing: includes balance-tlb plus
604			receive load balancing (rlb) for IPV4 traffic, and
605			does not require any special switch support.  The
606			receive load balancing is achieved by ARP negotiation.
607			The bonding driver intercepts the ARP Replies sent by
608			the local system on their way out and overwrites the
609			source hardware address with the unique hardware
610			address of one of the slaves in the bond such that
611			different peers use different hardware addresses for
612			the server.
613	
614			Receive traffic from connections created by the server
615			is also balanced.  When the local system sends an ARP
616			Request the bonding driver copies and saves the peer's
617			IP information from the ARP packet.  When the ARP
618			Reply arrives from the peer, its hardware address is
619			retrieved and the bonding driver initiates an ARP
620			reply to this peer assigning it to one of the slaves
621			in the bond.  A problematic outcome of using ARP
622			negotiation for balancing is that each time that an
623			ARP request is broadcast it uses the hardware address
624			of the bond.  Hence, peers learn the hardware address
625			of the bond and the balancing of receive traffic
626			collapses to the current slave.  This is handled by
627			sending updates (ARP Replies) to all the peers with
628			their individually assigned hardware address such that
629			the traffic is redistributed.  Receive traffic is also
630			redistributed when a new slave is added to the bond
631			and when an inactive slave is re-activated.  The
632			receive load is distributed sequentially (round robin)
633			among the group of highest speed slaves in the bond.
634	
635			When a link is reconnected or a new slave joins the
636			bond the receive traffic is redistributed among all
637			active slaves in the bond by initiating ARP Replies
638			with the selected MAC address to each of the
639			clients. The updelay parameter (detailed below) must
640			be set to a value equal or greater than the switch's
641			forwarding delay so that the ARP Replies sent to the
642			peers will not be blocked by the switch.
643	
644			Prerequisites:
645	
646			1. Ethtool support in the base drivers for retrieving
647			the speed of each slave.
648	
649			2. Base driver support for setting the hardware
650			address of a device while it is open.  This is
651			required so that there will always be one slave in the
652			team using the bond hardware address (the
653			curr_active_slave) while having a unique hardware
654			address for each slave in the bond.  If the
655			curr_active_slave fails its hardware address is
656			swapped with the new curr_active_slave that was
657			chosen.
658	
659	num_grat_arp
660	num_unsol_na
661	
662		Specify the number of peer notifications (gratuitous ARPs and
663		unsolicited IPv6 Neighbor Advertisements) to be issued after a
664		failover event.  As soon as the link is up on the new slave
665		(possibly immediately) a peer notification is sent on the
666		bonding device and each VLAN sub-device.  This is repeated at
667		each link monitor interval (arp_interval or miimon, whichever
668		is active) if the number is greater than 1.
669	
670		The valid range is 0 - 255; the default value is 1.  These options
671		affect only the active-backup mode.  These options were added for
672		bonding versions 3.3.0 and 3.4.0 respectively.
673	
674		From Linux 3.0 and bonding version 3.7.1, these notifications
675		are generated by the ipv4 and ipv6 code and the numbers of
676		repetitions cannot be set independently.
677	
678	packets_per_slave
679	
680		Specify the number of packets to transmit through a slave before
681		moving to the next one. When set to 0 then a slave is chosen at
682		random.
683	
684		The valid range is 0 - 65535; the default value is 1. This option
685		has effect only in balance-rr mode.
686	
687	primary
688	
689		A string (eth0, eth2, etc) specifying which slave is the
690		primary device.  The specified device will always be the
691		active slave while it is available.  Only when the primary is
692		off-line will alternate devices be used.  This is useful when
693		one slave is preferred over another, e.g., when one slave has
694		higher throughput than another.
695	
696		The primary option is only valid for active-backup(1),
697		balance-tlb (5) and balance-alb (6) mode.
698	
699	primary_reselect
700	
701		Specifies the reselection policy for the primary slave.  This
702		affects how the primary slave is chosen to become the active slave
703		when failure of the active slave or recovery of the primary slave
704		occurs.  This option is designed to prevent flip-flopping between
705		the primary slave and other slaves.  Possible values are:
706	
707		always or 0 (default)
708	
709			The primary slave becomes the active slave whenever it
710			comes back up.
711	
712		better or 1
713	
714			The primary slave becomes the active slave when it comes
715			back up, if the speed and duplex of the primary slave is
716			better than the speed and duplex of the current active
717			slave.
718	
719		failure or 2
720	
721			The primary slave becomes the active slave only if the
722			current active slave fails and the primary slave is up.
723	
724		The primary_reselect setting is ignored in two cases:
725	
726			If no slaves are active, the first slave to recover is
727			made the active slave.
728	
729			When initially enslaved, the primary slave is always made
730			the active slave.
731	
732		Changing the primary_reselect policy via sysfs will cause an
733		immediate selection of the best active slave according to the new
734		policy.  This may or may not result in a change of the active
735		slave, depending upon the circumstances.
736	
737		This option was added for bonding version 3.6.0.
738	
739	updelay
740	
741		Specifies the time, in milliseconds, to wait before enabling a
742		slave after a link recovery has been detected.  This option is
743		only valid for the miimon link monitor.  The updelay value
744		should be a multiple of the miimon value; if not, it will be
745		rounded down to the nearest multiple.  The default value is 0.
746	
747	use_carrier
748	
749		Specifies whether or not miimon should use MII or ETHTOOL
750		ioctls vs. netif_carrier_ok() to determine the link
751		status. The MII or ETHTOOL ioctls are less efficient and
752		utilize a deprecated calling sequence within the kernel.  The
753		netif_carrier_ok() relies on the device driver to maintain its
754		state with netif_carrier_on/off; at this writing, most, but
755		not all, device drivers support this facility.
756	
757		If bonding insists that the link is up when it should not be,
758		it may be that your network device driver does not support
759		netif_carrier_on/off.  The default state for netif_carrier is
760		"carrier on," so if a driver does not support netif_carrier,
761		it will appear as if the link is always up.  In this case,
762		setting use_carrier to 0 will cause bonding to revert to the
763		MII / ETHTOOL ioctl method to determine the link state.
764	
765		A value of 1 enables the use of netif_carrier_ok(), a value of
766		0 will use the deprecated MII / ETHTOOL ioctls.  The default
767		value is 1.
768	
769	xmit_hash_policy
770	
771		Selects the transmit hash policy to use for slave selection in
772		balance-xor and 802.3ad modes.  Possible values are:
773	
774		layer2
775	
776			Uses XOR of hardware MAC addresses to generate the
777			hash.  The formula is
778	
779			(source MAC XOR destination MAC) modulo slave count
780	
781			This algorithm will place all traffic to a particular
782			network peer on the same slave.
783	
784			This algorithm is 802.3ad compliant.
785	
786		layer2+3
787	
788			This policy uses a combination of layer2 and layer3
789			protocol information to generate the hash.
790	
791			Uses XOR of hardware MAC addresses and IP addresses to
792			generate the hash.  The formula is
793	
794			hash = source MAC XOR destination MAC
795			hash = hash XOR source IP XOR destination IP
796			hash = hash XOR (hash RSHIFT 16)
797			hash = hash XOR (hash RSHIFT 8)
798			And then hash is reduced modulo slave count.
799	
800			If the protocol is IPv6 then the source and destination
801			addresses are first hashed using ipv6_addr_hash.
802	
803			This algorithm will place all traffic to a particular
804			network peer on the same slave.  For non-IP traffic,
805			the formula is the same as for the layer2 transmit
806			hash policy.
807	
808			This policy is intended to provide a more balanced
809			distribution of traffic than layer2 alone, especially
810			in environments where a layer3 gateway device is
811			required to reach most destinations.
812	
813			This algorithm is 802.3ad compliant.
814	
815		layer3+4
816	
817			This policy uses upper layer protocol information,
818			when available, to generate the hash.  This allows for
819			traffic to a particular network peer to span multiple
820			slaves, although a single connection will not span
821			multiple slaves.
822	
823			The formula for unfragmented TCP and UDP packets is
824	
825			hash = source port, destination port (as in the header)
826			hash = hash XOR source IP XOR destination IP
827			hash = hash XOR (hash RSHIFT 16)
828			hash = hash XOR (hash RSHIFT 8)
829			And then hash is reduced modulo slave count.
830	
831			If the protocol is IPv6 then the source and destination
832			addresses are first hashed using ipv6_addr_hash.
833	
834			For fragmented TCP or UDP packets and all other IPv4 and
835			IPv6 protocol traffic, the source and destination port
836			information is omitted.  For non-IP traffic, the
837			formula is the same as for the layer2 transmit hash
838			policy.
839	
840			This algorithm is not fully 802.3ad compliant.  A
841			single TCP or UDP conversation containing both
842			fragmented and unfragmented packets will see packets
843			striped across two interfaces.  This may result in out
844			of order delivery.  Most traffic types will not meet
845			this criteria, as TCP rarely fragments traffic, and
846			most UDP traffic is not involved in extended
847			conversations.  Other implementations of 802.3ad may
848			or may not tolerate this noncompliance.
849	
850		encap2+3
851	
852			This policy uses the same formula as layer2+3 but it
853			relies on skb_flow_dissect to obtain the header fields
854			which might result in the use of inner headers if an
855			encapsulation protocol is used. For example this will
856			improve the performance for tunnel users because the
857			packets will be distributed according to the encapsulated
858			flows.
859	
860		encap3+4
861	
862			This policy uses the same formula as layer3+4 but it
863			relies on skb_flow_dissect to obtain the header fields
864			which might result in the use of inner headers if an
865			encapsulation protocol is used. For example this will
866			improve the performance for tunnel users because the
867			packets will be distributed according to the encapsulated
868			flows.
869	
870		The default value is layer2.  This option was added in bonding
871		version 2.6.3.  In earlier versions of bonding, this parameter
872		does not exist, and the layer2 policy is the only policy.  The
873		layer2+3 value was added for bonding version 3.2.2.
874	
875	resend_igmp
876	
877		Specifies the number of IGMP membership reports to be issued after
878		a failover event. One membership report is issued immediately after
879		the failover, subsequent packets are sent in each 200ms interval.
880	
881		The valid range is 0 - 255; the default value is 1. A value of 0
882		prevents the IGMP membership report from being issued in response
883		to the failover event.
884	
885		This option is useful for bonding modes balance-rr (0), active-backup
886		(1), balance-tlb (5) and balance-alb (6), in which a failover can
887		switch the IGMP traffic from one slave to another.  Therefore a fresh
888		IGMP report must be issued to cause the switch to forward the incoming
889		IGMP traffic over the newly selected slave.
890	
891		This option was added for bonding version 3.7.0.
892	
893	lp_interval
894	
895		Specifies the number of seconds between instances where the bonding
896		driver sends learning packets to each slaves peer switch.
897	
898		The valid range is 1 - 0x7fffffff; the default value is 1. This Option
899		has effect only in balance-tlb and balance-alb modes.
900	
901	3. Configuring Bonding Devices
902	==============================
903	
904		You can configure bonding using either your distro's network
905	initialization scripts, or manually using either iproute2 or the
906	sysfs interface.  Distros generally use one of three packages for the
907	network initialization scripts: initscripts, sysconfig or interfaces.
908	Recent versions of these packages have support for bonding, while older
909	versions do not.
910	
911		We will first describe the options for configuring bonding for
912	distros using versions of initscripts, sysconfig and interfaces with full
913	or partial support for bonding, then provide information on enabling
914	bonding without support from the network initialization scripts (i.e.,
915	older versions of initscripts or sysconfig).
916	
917		If you're unsure whether your distro uses sysconfig,
918	initscripts or interfaces, or don't know if it's new enough, have no fear.
919	Determining this is fairly straightforward.
920	
921		First, look for a file called interfaces in /etc/network directory.
922	If this file is present in your system, then your system use interfaces. See
923	Configuration with Interfaces Support.
924	
925		Else, issue the command:
926	
927	$ rpm -qf /sbin/ifup
928	
929		It will respond with a line of text starting with either
930	"initscripts" or "sysconfig," followed by some numbers.  This is the
931	package that provides your network initialization scripts.
932	
933		Next, to determine if your installation supports bonding,
934	issue the command:
935	
936	$ grep ifenslave /sbin/ifup
937	
938		If this returns any matches, then your initscripts or
939	sysconfig has support for bonding.
940	
941	3.1 Configuration with Sysconfig Support
942	----------------------------------------
943	
944		This section applies to distros using a version of sysconfig
945	with bonding support, for example, SuSE Linux Enterprise Server 9.
946	
947		SuSE SLES 9's networking configuration system does support
948	bonding, however, at this writing, the YaST system configuration
949	front end does not provide any means to work with bonding devices.
950	Bonding devices can be managed by hand, however, as follows.
951	
952		First, if they have not already been configured, configure the
953	slave devices.  On SLES 9, this is most easily done by running the
954	yast2 sysconfig configuration utility.  The goal is for to create an
955	ifcfg-id file for each slave device.  The simplest way to accomplish
956	this is to configure the devices for DHCP (this is only to get the
957	file ifcfg-id file created; see below for some issues with DHCP).  The
958	name of the configuration file for each device will be of the form:
959	
960	ifcfg-id-xx:xx:xx:xx:xx:xx
961	
962		Where the "xx" portion will be replaced with the digits from
963	the device's permanent MAC address.
964	
965		Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
966	created, it is necessary to edit the configuration files for the slave
967	devices (the MAC addresses correspond to those of the slave devices).
968	Before editing, the file will contain multiple lines, and will look
969	something like this:
970	
971	BOOTPROTO='dhcp'
972	STARTMODE='on'
973	USERCTL='no'
974	UNIQUE='XNzu.WeZGOGF+4wE'
975	_nm_name='bus-pci-0001:61:01.0'
976	
977		Change the BOOTPROTO and STARTMODE lines to the following:
978	
979	BOOTPROTO='none'
980	STARTMODE='off'
981	
982		Do not alter the UNIQUE or _nm_name lines.  Remove any other
983	lines (USERCTL, etc).
984	
985		Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
986	it's time to create the configuration file for the bonding device
987	itself.  This file is named ifcfg-bondX, where X is the number of the
988	bonding device to create, starting at 0.  The first such file is
989	ifcfg-bond0, the second is ifcfg-bond1, and so on.  The sysconfig
990	network configuration system will correctly start multiple instances
991	of bonding.
992	
993		The contents of the ifcfg-bondX file is as follows:
994	
995	BOOTPROTO="static"
996	BROADCAST="10.0.2.255"
997	IPADDR="10.0.2.10"
998	NETMASK="255.255.0.0"
999	NETWORK="10.0.2.0"
1000	REMOTE_IPADDR=""
1001	STARTMODE="onboot"
1002	BONDING_MASTER="yes"
1003	BONDING_MODULE_OPTS="mode=active-backup miimon=100"
1004	BONDING_SLAVE0="eth0"
1005	BONDING_SLAVE1="bus-pci-0000:06:08.1"
1006	
1007		Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
1008	values with the appropriate values for your network.
1009	
1010		The STARTMODE specifies when the device is brought online.
1011	The possible values are:
1012	
1013		onboot:	 The device is started at boot time.  If you're not
1014			 sure, this is probably what you want.
1015	
1016		manual:	 The device is started only when ifup is called
1017			 manually.  Bonding devices may be configured this
1018			 way if you do not wish them to start automatically
1019			 at boot for some reason.
1020	
1021		hotplug: The device is started by a hotplug event.  This is not
1022			 a valid choice for a bonding device.
1023	
1024		off or ignore: The device configuration is ignored.
1025	
1026		The line BONDING_MASTER='yes' indicates that the device is a
1027	bonding master device.  The only useful value is "yes."
1028	
1029		The contents of BONDING_MODULE_OPTS are supplied to the
1030	instance of the bonding module for this device.  Specify the options
1031	for the bonding mode, link monitoring, and so on here.  Do not include
1032	the max_bonds bonding parameter; this will confuse the configuration
1033	system if you have multiple bonding devices.
1034	
1035		Finally, supply one BONDING_SLAVEn="slave device" for each
1036	slave.  where "n" is an increasing value, one for each slave.  The
1037	"slave device" is either an interface name, e.g., "eth0", or a device
1038	specifier for the network device.  The interface name is easier to
1039	find, but the ethN names are subject to change at boot time if, e.g.,
1040	a device early in the sequence has failed.  The device specifiers
1041	(bus-pci-0000:06:08.1 in the example above) specify the physical
1042	network device, and will not change unless the device's bus location
1043	changes (for example, it is moved from one PCI slot to another).  The
1044	example above uses one of each type for demonstration purposes; most
1045	configurations will choose one or the other for all slave devices.
1046	
1047		When all configuration files have been modified or created,
1048	networking must be restarted for the configuration changes to take
1049	effect.  This can be accomplished via the following:
1050	
1051	# /etc/init.d/network restart
1052	
1053		Note that the network control script (/sbin/ifdown) will
1054	remove the bonding module as part of the network shutdown processing,
1055	so it is not necessary to remove the module by hand if, e.g., the
1056	module parameters have changed.
1057	
1058		Also, at this writing, YaST/YaST2 will not manage bonding
1059	devices (they do not show bonding interfaces on its list of network
1060	devices).  It is necessary to edit the configuration file by hand to
1061	change the bonding configuration.
1062	
1063		Additional general options and details of the ifcfg file
1064	format can be found in an example ifcfg template file:
1065	
1066	/etc/sysconfig/network/ifcfg.template
1067	
1068		Note that the template does not document the various BONDING_
1069	settings described above, but does describe many of the other options.
1070	
1071	3.1.1 Using DHCP with Sysconfig
1072	-------------------------------
1073	
1074		Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1075	will cause it to query DHCP for its IP address information.  At this
1076	writing, this does not function for bonding devices; the scripts
1077	attempt to obtain the device address from DHCP prior to adding any of
1078	the slave devices.  Without active slaves, the DHCP requests are not
1079	sent to the network.
1080	
1081	3.1.2 Configuring Multiple Bonds with Sysconfig
1082	-----------------------------------------------
1083	
1084		The sysconfig network initialization system is capable of
1085	handling multiple bonding devices.  All that is necessary is for each
1086	bonding instance to have an appropriately configured ifcfg-bondX file
1087	(as described above).  Do not specify the "max_bonds" parameter to any
1088	instance of bonding, as this will confuse sysconfig.  If you require
1089	multiple bonding devices with identical parameters, create multiple
1090	ifcfg-bondX files.
1091	
1092		Because the sysconfig scripts supply the bonding module
1093	options in the ifcfg-bondX file, it is not necessary to add them to
1094	the system /etc/modules.d/*.conf configuration files.
1095	
1096	3.2 Configuration with Initscripts Support
1097	------------------------------------------
1098	
1099		This section applies to distros using a recent version of
1100	initscripts with bonding support, for example, Red Hat Enterprise Linux
1101	version 3 or later, Fedora, etc.  On these systems, the network
1102	initialization scripts have knowledge of bonding, and can be configured to
1103	control bonding devices.  Note that older versions of the initscripts
1104	package have lower levels of support for bonding; this will be noted where
1105	applicable.
1106	
1107		These distros will not automatically load the network adapter
1108	driver unless the ethX device is configured with an IP address.
1109	Because of this constraint, users must manually configure a
1110	network-script file for all physical adapters that will be members of
1111	a bondX link.  Network script files are located in the directory:
1112	
1113	/etc/sysconfig/network-scripts
1114	
1115		The file name must be prefixed with "ifcfg-eth" and suffixed
1116	with the adapter's physical adapter number.  For example, the script
1117	for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1118	Place the following text in the file:
1119	
1120	DEVICE=eth0
1121	USERCTL=no
1122	ONBOOT=yes
1123	MASTER=bond0
1124	SLAVE=yes
1125	BOOTPROTO=none
1126	
1127		The DEVICE= line will be different for every ethX device and
1128	must correspond with the name of the file, i.e., ifcfg-eth1 must have
1129	a device line of DEVICE=eth1.  The setting of the MASTER= line will
1130	also depend on the final bonding interface name chosen for your bond.
1131	As with other network devices, these typically start at 0, and go up
1132	one for each device, i.e., the first bonding instance is bond0, the
1133	second is bond1, and so on.
1134	
1135		Next, create a bond network script.  The file name for this
1136	script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1137	the number of the bond.  For bond0 the file is named "ifcfg-bond0",
1138	for bond1 it is named "ifcfg-bond1", and so on.  Within that file,
1139	place the following text:
1140	
1141	DEVICE=bond0
1142	IPADDR=192.168.1.1
1143	NETMASK=255.255.255.0
1144	NETWORK=192.168.1.0
1145	BROADCAST=192.168.1.255
1146	ONBOOT=yes
1147	BOOTPROTO=none
1148	USERCTL=no
1149	
1150		Be sure to change the networking specific lines (IPADDR,
1151	NETMASK, NETWORK and BROADCAST) to match your network configuration.
1152	
1153		For later versions of initscripts, such as that found with Fedora
1154	7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1155	and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
1156	file, e.g. a line of the format:
1157	
1158	BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
1159	
1160		will configure the bond with the specified options.  The options
1161	specified in BONDING_OPTS are identical to the bonding module parameters
1162	except for the arp_ip_target field when using versions of initscripts older
1163	than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2).  When
1164	using older versions each target should be included as a separate option and
1165	should be preceded by a '+' to indicate it should be added to the list of
1166	queried targets, e.g.,
1167	
1168		arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1169	
1170		is the proper syntax to specify multiple targets.  When specifying
1171	options via BONDING_OPTS, it is not necessary to edit /etc/modprobe.d/*.conf.
1172	
1173		For even older versions of initscripts that do not support
1174	BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1175	your distro) to load the bonding module with your desired options when the
1176	bond0 interface is brought up.  The following lines in /etc/modprobe.d/*.conf
1177	will load the bonding module, and select its options:
1178	
1179	alias bond0 bonding
1180	options bond0 mode=balance-alb miimon=100
1181	
1182		Replace the sample parameters with the appropriate set of
1183	options for your configuration.
1184	
1185		Finally run "/etc/rc.d/init.d/network restart" as root.  This
1186	will restart the networking subsystem and your bond link should be now
1187	up and running.
1188	
1189	3.2.1 Using DHCP with Initscripts
1190	---------------------------------
1191	
1192		Recent versions of initscripts (the versions supplied with Fedora
1193	Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1194	work) have support for assigning IP information to bonding devices via
1195	DHCP.
1196	
1197		To configure bonding for DHCP, configure it as described
1198	above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1199	and add a line consisting of "TYPE=Bonding".  Note that the TYPE value
1200	is case sensitive.
1201	
1202	3.2.2 Configuring Multiple Bonds with Initscripts
1203	-------------------------------------------------
1204	
1205		Initscripts packages that are included with Fedora 7 and Red Hat
1206	Enterprise Linux 5 support multiple bonding interfaces by simply
1207	specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1208	number of the bond.  This support requires sysfs support in the kernel,
1209	and a bonding driver of version 3.0.0 or later.  Other configurations may
1210	not support this method for specifying multiple bonding interfaces; for
1211	those instances, see the "Configuring Multiple Bonds Manually" section,
1212	below.
1213	
1214	3.3 Configuring Bonding Manually with iproute2
1215	-----------------------------------------------
1216	
1217		This section applies to distros whose network initialization
1218	scripts (the sysconfig or initscripts package) do not have specific
1219	knowledge of bonding.  One such distro is SuSE Linux Enterprise Server
1220	version 8.
1221	
1222		The general method for these systems is to place the bonding
1223	module parameters into a config file in /etc/modprobe.d/ (as
1224	appropriate for the installed distro), then add modprobe and/or
1225	`ip link` commands to the system's global init script.  The name of
1226	the global init script differs; for sysconfig, it is
1227	/etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1228	
1229		For example, if you wanted to make a simple bond of two e100
1230	devices (presumed to be eth0 and eth1), and have it persist across
1231	reboots, edit the appropriate file (/etc/init.d/boot.local or
1232	/etc/rc.d/rc.local), and add the following:
1233	
1234	modprobe bonding mode=balance-alb miimon=100
1235	modprobe e100
1236	ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1237	ip link set eth0 master bond0
1238	ip link set eth1 master bond0
1239	
1240		Replace the example bonding module parameters and bond0
1241	network configuration (IP address, netmask, etc) with the appropriate
1242	values for your configuration.
1243	
1244		Unfortunately, this method will not provide support for the
1245	ifup and ifdown scripts on the bond devices.  To reload the bonding
1246	configuration, it is necessary to run the initialization script, e.g.,
1247	
1248	# /etc/init.d/boot.local
1249	
1250		or
1251	
1252	# /etc/rc.d/rc.local
1253	
1254		It may be desirable in such a case to create a separate script
1255	which only initializes the bonding configuration, then call that
1256	separate script from within boot.local.  This allows for bonding to be
1257	enabled without re-running the entire global init script.
1258	
1259		To shut down the bonding devices, it is necessary to first
1260	mark the bonding device itself as being down, then remove the
1261	appropriate device driver modules.  For our example above, you can do
1262	the following:
1263	
1264	# ifconfig bond0 down
1265	# rmmod bonding
1266	# rmmod e100
1267	
1268		Again, for convenience, it may be desirable to create a script
1269	with these commands.
1270	
1271	
1272	3.3.1 Configuring Multiple Bonds Manually
1273	-----------------------------------------
1274	
1275		This section contains information on configuring multiple
1276	bonding devices with differing options for those systems whose network
1277	initialization scripts lack support for configuring multiple bonds.
1278	
1279		If you require multiple bonding devices, but all with the same
1280	options, you may wish to use the "max_bonds" module parameter,
1281	documented above.
1282	
1283		To create multiple bonding devices with differing options, it is
1284	preferable to use bonding parameters exported by sysfs, documented in the
1285	section below.
1286	
1287		For versions of bonding without sysfs support, the only means to
1288	provide multiple instances of bonding with differing options is to load
1289	the bonding driver multiple times.  Note that current versions of the
1290	sysconfig network initialization scripts handle this automatically; if
1291	your distro uses these scripts, no special action is needed.  See the
1292	section Configuring Bonding Devices, above, if you're not sure about your
1293	network initialization scripts.
1294	
1295		To load multiple instances of the module, it is necessary to
1296	specify a different name for each instance (the module loading system
1297	requires that every loaded module, even multiple instances of the same
1298	module, have a unique name).  This is accomplished by supplying multiple
1299	sets of bonding options in /etc/modprobe.d/*.conf, for example:
1300	
1301	alias bond0 bonding
1302	options bond0 -o bond0 mode=balance-rr miimon=100
1303	
1304	alias bond1 bonding
1305	options bond1 -o bond1 mode=balance-alb miimon=50
1306	
1307		will load the bonding module two times.  The first instance is
1308	named "bond0" and creates the bond0 device in balance-rr mode with an
1309	miimon of 100.  The second instance is named "bond1" and creates the
1310	bond1 device in balance-alb mode with an miimon of 50.
1311	
1312		In some circumstances (typically with older distributions),
1313	the above does not work, and the second bonding instance never sees
1314	its options.  In that case, the second options line can be substituted
1315	as follows:
1316	
1317	install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1318		mode=balance-alb miimon=50
1319	
1320		This may be repeated any number of times, specifying a new and
1321	unique name in place of bond1 for each subsequent instance.
1322	
1323		It has been observed that some Red Hat supplied kernels are unable
1324	to rename modules at load time (the "-o bond1" part).  Attempts to pass
1325	that option to modprobe will produce an "Operation not permitted" error.
1326	This has been reported on some Fedora Core kernels, and has been seen on
1327	RHEL 4 as well.  On kernels exhibiting this problem, it will be impossible
1328	to configure multiple bonds with differing parameters (as they are older
1329	kernels, and also lack sysfs support).
1330	
1331	3.4 Configuring Bonding Manually via Sysfs
1332	------------------------------------------
1333	
1334		Starting with version 3.0.0, Channel Bonding may be configured
1335	via the sysfs interface.  This interface allows dynamic configuration
1336	of all bonds in the system without unloading the module.  It also
1337	allows for adding and removing bonds at runtime.  Ifenslave is no
1338	longer required, though it is still supported.
1339	
1340		Use of the sysfs interface allows you to use multiple bonds
1341	with different configurations without having to reload the module.
1342	It also allows you to use multiple, differently configured bonds when
1343	bonding is compiled into the kernel.
1344	
1345		You must have the sysfs filesystem mounted to configure
1346	bonding this way.  The examples in this document assume that you
1347	are using the standard mount point for sysfs, e.g. /sys.  If your
1348	sysfs filesystem is mounted elsewhere, you will need to adjust the
1349	example paths accordingly.
1350	
1351	Creating and Destroying Bonds
1352	-----------------------------
1353	To add a new bond foo:
1354	# echo +foo > /sys/class/net/bonding_masters
1355	
1356	To remove an existing bond bar:
1357	# echo -bar > /sys/class/net/bonding_masters
1358	
1359	To show all existing bonds:
1360	# cat /sys/class/net/bonding_masters
1361	
1362	NOTE: due to 4K size limitation of sysfs files, this list may be
1363	truncated if you have more than a few hundred bonds.  This is unlikely
1364	to occur under normal operating conditions.
1365	
1366	Adding and Removing Slaves
1367	--------------------------
1368		Interfaces may be enslaved to a bond using the file
1369	/sys/class/net/<bond>/bonding/slaves.  The semantics for this file
1370	are the same as for the bonding_masters file.
1371	
1372	To enslave interface eth0 to bond bond0:
1373	# ifconfig bond0 up
1374	# echo +eth0 > /sys/class/net/bond0/bonding/slaves
1375	
1376	To free slave eth0 from bond bond0:
1377	# echo -eth0 > /sys/class/net/bond0/bonding/slaves
1378	
1379		When an interface is enslaved to a bond, symlinks between the
1380	two are created in the sysfs filesystem.  In this case, you would get
1381	/sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1382	/sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1383	
1384		This means that you can tell quickly whether or not an
1385	interface is enslaved by looking for the master symlink.  Thus:
1386	# echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1387	will free eth0 from whatever bond it is enslaved to, regardless of
1388	the name of the bond interface.
1389	
1390	Changing a Bond's Configuration
1391	-------------------------------
1392		Each bond may be configured individually by manipulating the
1393	files located in /sys/class/net/<bond name>/bonding
1394	
1395		The names of these files correspond directly with the command-
1396	line parameters described elsewhere in this file, and, with the
1397	exception of arp_ip_target, they accept the same values.  To see the
1398	current setting, simply cat the appropriate file.
1399	
1400		A few examples will be given here; for specific usage
1401	guidelines for each parameter, see the appropriate section in this
1402	document.
1403	
1404	To configure bond0 for balance-alb mode:
1405	# ifconfig bond0 down
1406	# echo 6 > /sys/class/net/bond0/bonding/mode
1407	 - or -
1408	# echo balance-alb > /sys/class/net/bond0/bonding/mode
1409		NOTE: The bond interface must be down before the mode can be
1410	changed.
1411	
1412	To enable MII monitoring on bond0 with a 1 second interval:
1413	# echo 1000 > /sys/class/net/bond0/bonding/miimon
1414		NOTE: If ARP monitoring is enabled, it will disabled when MII
1415	monitoring is enabled, and vice-versa.
1416	
1417	To add ARP targets:
1418	# echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1419	# echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1420		NOTE:  up to 16 target addresses may be specified.
1421	
1422	To remove an ARP target:
1423	# echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1424	
1425	To configure the interval between learning packet transmits:
1426	# echo 12 > /sys/class/net/bond0/bonding/lp_interval
1427		NOTE: the lp_inteval is the number of seconds between instances where
1428	the bonding driver sends learning packets to each slaves peer switch.  The
1429	default interval is 1 second.
1430	
1431	Example Configuration
1432	---------------------
1433		We begin with the same example that is shown in section 3.3,
1434	executed with sysfs, and without using ifenslave.
1435	
1436		To make a simple bond of two e100 devices (presumed to be eth0
1437	and eth1), and have it persist across reboots, edit the appropriate
1438	file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1439	following:
1440	
1441	modprobe bonding
1442	modprobe e100
1443	echo balance-alb > /sys/class/net/bond0/bonding/mode
1444	ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1445	echo 100 > /sys/class/net/bond0/bonding/miimon
1446	echo +eth0 > /sys/class/net/bond0/bonding/slaves
1447	echo +eth1 > /sys/class/net/bond0/bonding/slaves
1448	
1449		To add a second bond, with two e1000 interfaces in
1450	active-backup mode, using ARP monitoring, add the following lines to
1451	your init script:
1452	
1453	modprobe e1000
1454	echo +bond1 > /sys/class/net/bonding_masters
1455	echo active-backup > /sys/class/net/bond1/bonding/mode
1456	ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1457	echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1458	echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1459	echo +eth2 > /sys/class/net/bond1/bonding/slaves
1460	echo +eth3 > /sys/class/net/bond1/bonding/slaves
1461	
1462	3.5 Configuration with Interfaces Support
1463	-----------------------------------------
1464	
1465	        This section applies to distros which use /etc/network/interfaces file
1466	to describe network interface configuration, most notably Debian and it's
1467	derivatives.
1468	
1469		The ifup and ifdown commands on Debian don't support bonding out of
1470	the box. The ifenslave-2.6 package should be installed to provide bonding
1471	support.  Once installed, this package will provide bond-* options to be used
1472	into /etc/network/interfaces.
1473	
1474		Note that ifenslave-2.6 package will load the bonding module and use
1475	the ifenslave command when appropriate.
1476	
1477	Example Configurations
1478	----------------------
1479	
1480	In /etc/network/interfaces, the following stanza will configure bond0, in
1481	active-backup mode, with eth0 and eth1 as slaves.
1482	
1483	auto bond0
1484	iface bond0 inet dhcp
1485		bond-slaves eth0 eth1
1486		bond-mode active-backup
1487		bond-miimon 100
1488		bond-primary eth0 eth1
1489	
1490	If the above configuration doesn't work, you might have a system using
1491	upstart for system startup. This is most notably true for recent
1492	Ubuntu versions. The following stanza in /etc/network/interfaces will
1493	produce the same result on those systems.
1494	
1495	auto bond0
1496	iface bond0 inet dhcp
1497		bond-slaves none
1498		bond-mode active-backup
1499		bond-miimon 100
1500	
1501	auto eth0
1502	iface eth0 inet manual
1503		bond-master bond0
1504		bond-primary eth0 eth1
1505	
1506	auto eth1
1507	iface eth1 inet manual
1508		bond-master bond0
1509		bond-primary eth0 eth1
1510	
1511	For a full list of bond-* supported options in /etc/network/interfaces and some
1512	more advanced examples tailored to you particular distros, see the files in
1513	/usr/share/doc/ifenslave-2.6.
1514	
1515	3.6 Overriding Configuration for Special Cases
1516	----------------------------------------------
1517	
1518	When using the bonding driver, the physical port which transmits a frame is
1519	typically selected by the bonding driver, and is not relevant to the user or
1520	system administrator.  The output port is simply selected using the policies of
1521	the selected bonding mode.  On occasion however, it is helpful to direct certain
1522	classes of traffic to certain physical interfaces on output to implement
1523	slightly more complex policies.  For example, to reach a web server over a
1524	bonded interface in which eth0 connects to a private network, while eth1
1525	connects via a public network, it may be desirous to bias the bond to send said
1526	traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1527	can safely be sent over either interface.  Such configurations may be achieved
1528	using the traffic control utilities inherent in linux.
1529	
1530	By default the bonding driver is multiqueue aware and 16 queues are created
1531	when the driver initializes (see Documentation/networking/multiqueue.txt
1532	for details).  If more or less queues are desired the module parameter
1533	tx_queues can be used to change this value.  There is no sysfs parameter
1534	available as the allocation is done at module init time.
1535	
1536	The output of the file /proc/net/bonding/bondX has changed so the output Queue
1537	ID is now printed for each slave:
1538	
1539	Bonding Mode: fault-tolerance (active-backup)
1540	Primary Slave: None
1541	Currently Active Slave: eth0
1542	MII Status: up
1543	MII Polling Interval (ms): 0
1544	Up Delay (ms): 0
1545	Down Delay (ms): 0
1546	
1547	Slave Interface: eth0
1548	MII Status: up
1549	Link Failure Count: 0
1550	Permanent HW addr: 00:1a:a0:12:8f:cb
1551	Slave queue ID: 0
1552	
1553	Slave Interface: eth1
1554	MII Status: up
1555	Link Failure Count: 0
1556	Permanent HW addr: 00:1a:a0:12:8f:cc
1557	Slave queue ID: 2
1558	
1559	The queue_id for a slave can be set using the command:
1560	
1561	# echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1562	
1563	Any interface that needs a queue_id set should set it with multiple calls
1564	like the one above until proper priorities are set for all interfaces.  On
1565	distributions that allow configuration via initscripts, multiple 'queue_id'
1566	arguments can be added to BONDING_OPTS to set all needed slave queues.
1567	
1568	These queue id's can be used in conjunction with the tc utility to configure
1569	a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1570	slave devices.  For instance, say we wanted, in the above configuration to
1571	force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1572	device. The following commands would accomplish this:
1573	
1574	# tc qdisc add dev bond0 handle 1 root multiq
1575	
1576	# tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip dst \
1577		192.168.1.100 action skbedit queue_mapping 2
1578	
1579	These commands tell the kernel to attach a multiqueue queue discipline to the
1580	bond0 interface and filter traffic enqueued to it, such that packets with a dst
1581	ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1582	This value is then passed into the driver, causing the normal output path
1583	selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1584	
1585	Note that qid values begin at 1.  Qid 0 is reserved to initiate to the driver
1586	that normal output policy selection should take place.  One benefit to simply
1587	leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1588	driver that is now present.  This awareness allows tc filters to be placed on
1589	slave devices as well as bond devices and the bonding driver will simply act as
1590	a pass-through for selecting output queues on the slave device rather than 
1591	output port selection.
1592	
1593	This feature first appeared in bonding driver version 3.7.0 and support for
1594	output slave selection was limited to round-robin and active-backup modes.
1595	
1596	4 Querying Bonding Configuration
1597	=================================
1598	
1599	4.1 Bonding Configuration
1600	-------------------------
1601	
1602		Each bonding device has a read-only file residing in the
1603	/proc/net/bonding directory.  The file contents include information
1604	about the bonding configuration, options and state of each slave.
1605	
1606		For example, the contents of /proc/net/bonding/bond0 after the
1607	driver is loaded with parameters of mode=0 and miimon=1000 is
1608	generally as follows:
1609	
1610		Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1611	        Bonding Mode: load balancing (round-robin)
1612	        Currently Active Slave: eth0
1613	        MII Status: up
1614	        MII Polling Interval (ms): 1000
1615	        Up Delay (ms): 0
1616	        Down Delay (ms): 0
1617	
1618	        Slave Interface: eth1
1619	        MII Status: up
1620	        Link Failure Count: 1
1621	
1622	        Slave Interface: eth0
1623	        MII Status: up
1624	        Link Failure Count: 1
1625	
1626		The precise format and contents will change depending upon the
1627	bonding configuration, state, and version of the bonding driver.
1628	
1629	4.2 Network configuration
1630	-------------------------
1631	
1632		The network configuration can be inspected using the ifconfig
1633	command.  Bonding devices will have the MASTER flag set; Bonding slave
1634	devices will have the SLAVE flag set.  The ifconfig output does not
1635	contain information on which slaves are associated with which masters.
1636	
1637		In the example below, the bond0 interface is the master
1638	(MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1639	bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1640	TLB and ALB that require a unique MAC address for each slave.
1641	
1642	# /sbin/ifconfig
1643	bond0     Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1644	          inet addr:XXX.XXX.XXX.YYY  Bcast:XXX.XXX.XXX.255  Mask:255.255.252.0
1645	          UP BROADCAST RUNNING MASTER MULTICAST  MTU:1500  Metric:1
1646	          RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1647	          TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1648	          collisions:0 txqueuelen:0
1649	
1650	eth0      Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1651	          UP BROADCAST RUNNING SLAVE MULTICAST  MTU:1500  Metric:1
1652	          RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1653	          TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1654	          collisions:0 txqueuelen:100
1655	          Interrupt:10 Base address:0x1080
1656	
1657	eth1      Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1658	          UP BROADCAST RUNNING SLAVE MULTICAST  MTU:1500  Metric:1
1659	          RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1660	          TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1661	          collisions:0 txqueuelen:100
1662	          Interrupt:9 Base address:0x1400
1663	
1664	5. Switch Configuration
1665	=======================
1666	
1667		For this section, "switch" refers to whatever system the
1668	bonded devices are directly connected to (i.e., where the other end of
1669	the cable plugs into).  This may be an actual dedicated switch device,
1670	or it may be another regular system (e.g., another computer running
1671	Linux),
1672	
1673		The active-backup, balance-tlb and balance-alb modes do not
1674	require any specific configuration of the switch.
1675	
1676		The 802.3ad mode requires that the switch have the appropriate
1677	ports configured as an 802.3ad aggregation.  The precise method used
1678	to configure this varies from switch to switch, but, for example, a
1679	Cisco 3550 series switch requires that the appropriate ports first be
1680	grouped together in a single etherchannel instance, then that
1681	etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1682	standard EtherChannel).
1683	
1684		The balance-rr, balance-xor and broadcast modes generally
1685	require that the switch have the appropriate ports grouped together.
1686	The nomenclature for such a group differs between switches, it may be
1687	called an "etherchannel" (as in the Cisco example, above), a "trunk
1688	group" or some other similar variation.  For these modes, each switch
1689	will also have its own configuration options for the switch's transmit
1690	policy to the bond.  Typical choices include XOR of either the MAC or
1691	IP addresses.  The transmit policy of the two peers does not need to
1692	match.  For these three modes, the bonding mode really selects a
1693	transmit policy for an EtherChannel group; all three will interoperate
1694	with another EtherChannel group.
1695	
1696	
1697	6. 802.1q VLAN Support
1698	======================
1699	
1700		It is possible to configure VLAN devices over a bond interface
1701	using the 8021q driver.  However, only packets coming from the 8021q
1702	driver and passing through bonding will be tagged by default.  Self
1703	generated packets, for example, bonding's learning packets or ARP
1704	packets generated by either ALB mode or the ARP monitor mechanism, are
1705	tagged internally by bonding itself.  As a result, bonding must
1706	"learn" the VLAN IDs configured above it, and use those IDs to tag
1707	self generated packets.
1708	
1709		For reasons of simplicity, and to support the use of adapters
1710	that can do VLAN hardware acceleration offloading, the bonding
1711	interface declares itself as fully hardware offloading capable, it gets
1712	the add_vid/kill_vid notifications to gather the necessary
1713	information, and it propagates those actions to the slaves.  In case
1714	of mixed adapter types, hardware accelerated tagged packets that
1715	should go through an adapter that is not offloading capable are
1716	"un-accelerated" by the bonding driver so the VLAN tag sits in the
1717	regular location.
1718	
1719		VLAN interfaces *must* be added on top of a bonding interface
1720	only after enslaving at least one slave.  The bonding interface has a
1721	hardware address of 00:00:00:00:00:00 until the first slave is added.
1722	If the VLAN interface is created prior to the first enslavement, it
1723	would pick up the all-zeroes hardware address.  Once the first slave
1724	is attached to the bond, the bond device itself will pick up the
1725	slave's hardware address, which is then available for the VLAN device.
1726	
1727		Also, be aware that a similar problem can occur if all slaves
1728	are released from a bond that still has one or more VLAN interfaces on
1729	top of it.  When a new slave is added, the bonding interface will
1730	obtain its hardware address from the first slave, which might not
1731	match the hardware address of the VLAN interfaces (which was
1732	ultimately copied from an earlier slave).
1733	
1734		There are two methods to insure that the VLAN device operates
1735	with the correct hardware address if all slaves are removed from a
1736	bond interface:
1737	
1738		1. Remove all VLAN interfaces then recreate them
1739	
1740		2. Set the bonding interface's hardware address so that it
1741	matches the hardware address of the VLAN interfaces.
1742	
1743		Note that changing a VLAN interface's HW address would set the
1744	underlying device -- i.e. the bonding interface -- to promiscuous
1745	mode, which might not be what you want.
1746	
1747	
1748	7. Link Monitoring
1749	==================
1750	
1751		The bonding driver at present supports two schemes for
1752	monitoring a slave device's link state: the ARP monitor and the MII
1753	monitor.
1754	
1755		At the present time, due to implementation restrictions in the
1756	bonding driver itself, it is not possible to enable both ARP and MII
1757	monitoring simultaneously.
1758	
1759	7.1 ARP Monitor Operation
1760	-------------------------
1761	
1762		The ARP monitor operates as its name suggests: it sends ARP
1763	queries to one or more designated peer systems on the network, and
1764	uses the response as an indication that the link is operating.  This
1765	gives some assurance that traffic is actually flowing to and from one
1766	or more peers on the local network.
1767	
1768		The ARP monitor relies on the device driver itself to verify
1769	that traffic is flowing.  In particular, the driver must keep up to
1770	date the last receive time, dev->last_rx, and transmit start time,
1771	dev->trans_start.  If these are not updated by the driver, then the
1772	ARP monitor will immediately fail any slaves using that driver, and
1773	those slaves will stay down.  If networking monitoring (tcpdump, etc)
1774	shows the ARP requests and replies on the network, then it may be that
1775	your device driver is not updating last_rx and trans_start.
1776	
1777	7.2 Configuring Multiple ARP Targets
1778	------------------------------------
1779	
1780		While ARP monitoring can be done with just one target, it can
1781	be useful in a High Availability setup to have several targets to
1782	monitor.  In the case of just one target, the target itself may go
1783	down or have a problem making it unresponsive to ARP requests.  Having
1784	an additional target (or several) increases the reliability of the ARP
1785	monitoring.
1786	
1787		Multiple ARP targets must be separated by commas as follows:
1788	
1789	# example options for ARP monitoring with three targets
1790	alias bond0 bonding
1791	options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1792	
1793		For just a single target the options would resemble:
1794	
1795	# example options for ARP monitoring with one target
1796	alias bond0 bonding
1797	options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1798	
1799	
1800	7.3 MII Monitor Operation
1801	-------------------------
1802	
1803		The MII monitor monitors only the carrier state of the local
1804	network interface.  It accomplishes this in one of three ways: by
1805	depending upon the device driver to maintain its carrier state, by
1806	querying the device's MII registers, or by making an ethtool query to
1807	the device.
1808	
1809		If the use_carrier module parameter is 1 (the default value),
1810	then the MII monitor will rely on the driver for carrier state
1811	information (via the netif_carrier subsystem).  As explained in the
1812	use_carrier parameter information, above, if the MII monitor fails to
1813	detect carrier loss on the device (e.g., when the cable is physically
1814	disconnected), it may be that the driver does not support
1815	netif_carrier.
1816	
1817		If use_carrier is 0, then the MII monitor will first query the
1818	device's (via ioctl) MII registers and check the link state.  If that
1819	request fails (not just that it returns carrier down), then the MII
1820	monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1821	the same information.  If both methods fail (i.e., the driver either
1822	does not support or had some error in processing both the MII register
1823	and ethtool requests), then the MII monitor will assume the link is
1824	up.
1825	
1826	8. Potential Sources of Trouble
1827	===============================
1828	
1829	8.1 Adventures in Routing
1830	-------------------------
1831	
1832		When bonding is configured, it is important that the slave
1833	devices not have routes that supersede routes of the master (or,
1834	generally, not have routes at all).  For example, suppose the bonding
1835	device bond0 has two slaves, eth0 and eth1, and the routing table is
1836	as follows:
1837	
1838	Kernel IP routing table
1839	Destination     Gateway         Genmask         Flags   MSS Window  irtt Iface
1840	10.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 eth0
1841	10.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 eth1
1842	10.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 bond0
1843	127.0.0.0       0.0.0.0         255.0.0.0       U        40 0          0 lo
1844	
1845		This routing configuration will likely still update the
1846	receive/transmit times in the driver (needed by the ARP monitor), but
1847	may bypass the bonding driver (because outgoing traffic to, in this
1848	case, another host on network 10 would use eth0 or eth1 before bond0).
1849	
1850		The ARP monitor (and ARP itself) may become confused by this
1851	configuration, because ARP requests (generated by the ARP monitor)
1852	will be sent on one interface (bond0), but the corresponding reply
1853	will arrive on a different interface (eth0).  This reply looks to ARP
1854	as an unsolicited ARP reply (because ARP matches replies on an
1855	interface basis), and is discarded.  The MII monitor is not affected
1856	by the state of the routing table.
1857	
1858		The solution here is simply to insure that slaves do not have
1859	routes of their own, and if for some reason they must, those routes do
1860	not supersede routes of their master.  This should generally be the
1861	case, but unusual configurations or errant manual or automatic static
1862	route additions may cause trouble.
1863	
1864	8.2 Ethernet Device Renaming
1865	----------------------------
1866	
1867		On systems with network configuration scripts that do not
1868	associate physical devices directly with network interface names (so
1869	that the same physical device always has the same "ethX" name), it may
1870	be necessary to add some special logic to config files in
1871	/etc/modprobe.d/.
1872	
1873		For example, given a modules.conf containing the following:
1874	
1875	alias bond0 bonding
1876	options bond0 mode=some-mode miimon=50
1877	alias eth0 tg3
1878	alias eth1 tg3
1879	alias eth2 e1000
1880	alias eth3 e1000
1881	
1882		If neither eth0 and eth1 are slaves to bond0, then when the
1883	bond0 interface comes up, the devices may end up reordered.  This
1884	happens because bonding is loaded first, then its slave device's
1885	drivers are loaded next.  Since no other drivers have been loaded,
1886	when the e1000 driver loads, it will receive eth0 and eth1 for its
1887	devices, but the bonding configuration tries to enslave eth2 and eth3
1888	(which may later be assigned to the tg3 devices).
1889	
1890		Adding the following:
1891	
1892	add above bonding e1000 tg3
1893	
1894		causes modprobe to load e1000 then tg3, in that order, when
1895	bonding is loaded.  This command is fully documented in the
1896	modules.conf manual page.
1897	
1898		On systems utilizing modprobe an equivalent problem can occur.
1899	In this case, the following can be added to config files in
1900	/etc/modprobe.d/ as:
1901	
1902	softdep bonding pre: tg3 e1000
1903	
1904		This will load tg3 and e1000 modules before loading the bonding one.
1905	Full documentation on this can be found in the modprobe.d and modprobe
1906	manual pages.
1907	
1908	8.3. Painfully Slow Or No Failed Link Detection By Miimon
1909	---------------------------------------------------------
1910	
1911		By default, bonding enables the use_carrier option, which
1912	instructs bonding to trust the driver to maintain carrier state.
1913	
1914		As discussed in the options section, above, some drivers do
1915	not support the netif_carrier_on/_off link state tracking system.
1916	With use_carrier enabled, bonding will always see these links as up,
1917	regardless of their actual state.
1918	
1919		Additionally, other drivers do support netif_carrier, but do
1920	not maintain it in real time, e.g., only polling the link state at
1921	some fixed interval.  In this case, miimon will detect failures, but
1922	only after some long period of time has expired.  If it appears that
1923	miimon is very slow in detecting link failures, try specifying
1924	use_carrier=0 to see if that improves the failure detection time.  If
1925	it does, then it may be that the driver checks the carrier state at a
1926	fixed interval, but does not cache the MII register values (so the
1927	use_carrier=0 method of querying the registers directly works).  If
1928	use_carrier=0 does not improve the failover, then the driver may cache
1929	the registers, or the problem may be elsewhere.
1930	
1931		Also, remember that miimon only checks for the device's
1932	carrier state.  It has no way to determine the state of devices on or
1933	beyond other ports of a switch, or if a switch is refusing to pass
1934	traffic while still maintaining carrier on.
1935	
1936	9. SNMP agents
1937	===============
1938	
1939		If running SNMP agents, the bonding driver should be loaded
1940	before any network drivers participating in a bond.  This requirement
1941	is due to the interface index (ipAdEntIfIndex) being associated to
1942	the first interface found with a given IP address.  That is, there is
1943	only one ipAdEntIfIndex for each IP address.  For example, if eth0 and
1944	eth1 are slaves of bond0 and the driver for eth0 is loaded before the
1945	bonding driver, the interface for the IP address will be associated
1946	with the eth0 interface.  This configuration is shown below, the IP
1947	address 192.168.1.1 has an interface index of 2 which indexes to eth0
1948	in the ifDescr table (ifDescr.2).
1949	
1950	     interfaces.ifTable.ifEntry.ifDescr.1 = lo
1951	     interfaces.ifTable.ifEntry.ifDescr.2 = eth0
1952	     interfaces.ifTable.ifEntry.ifDescr.3 = eth1
1953	     interfaces.ifTable.ifEntry.ifDescr.4 = eth2
1954	     interfaces.ifTable.ifEntry.ifDescr.5 = eth3
1955	     interfaces.ifTable.ifEntry.ifDescr.6 = bond0
1956	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
1957	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1958	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
1959	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1960	
1961		This problem is avoided by loading the bonding driver before
1962	any network drivers participating in a bond.  Below is an example of
1963	loading the bonding driver first, the IP address 192.168.1.1 is
1964	correctly associated with ifDescr.2.
1965	
1966	     interfaces.ifTable.ifEntry.ifDescr.1 = lo
1967	     interfaces.ifTable.ifEntry.ifDescr.2 = bond0
1968	     interfaces.ifTable.ifEntry.ifDescr.3 = eth0
1969	     interfaces.ifTable.ifEntry.ifDescr.4 = eth1
1970	     interfaces.ifTable.ifEntry.ifDescr.5 = eth2
1971	     interfaces.ifTable.ifEntry.ifDescr.6 = eth3
1972	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
1973	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1974	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
1975	     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1976	
1977		While some distributions may not report the interface name in
1978	ifDescr, the association between the IP address and IfIndex remains
1979	and SNMP functions such as Interface_Scan_Next will report that
1980	association.
1981	
1982	10. Promiscuous mode
1983	====================
1984	
1985		When running network monitoring tools, e.g., tcpdump, it is
1986	common to enable promiscuous mode on the device, so that all traffic
1987	is seen (instead of seeing only traffic destined for the local host).
1988	The bonding driver handles promiscuous mode changes to the bonding
1989	master device (e.g., bond0), and propagates the setting to the slave
1990	devices.
1991	
1992		For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
1993	the promiscuous mode setting is propagated to all slaves.
1994	
1995		For the active-backup, balance-tlb and balance-alb modes, the
1996	promiscuous mode setting is propagated only to the active slave.
1997	
1998		For balance-tlb mode, the active slave is the slave currently
1999	receiving inbound traffic.
2000	
2001		For balance-alb mode, the active slave is the slave used as a
2002	"primary."  This slave is used for mode-specific control traffic, for
2003	sending to peers that are unassigned or if the load is unbalanced.
2004	
2005		For the active-backup, balance-tlb and balance-alb modes, when
2006	the active slave changes (e.g., due to a link failure), the
2007	promiscuous setting will be propagated to the new active slave.
2008	
2009	11. Configuring Bonding for High Availability
2010	=============================================
2011	
2012		High Availability refers to configurations that provide
2013	maximum network availability by having redundant or backup devices,
2014	links or switches between the host and the rest of the world.  The
2015	goal is to provide the maximum availability of network connectivity
2016	(i.e., the network always works), even though other configurations
2017	could provide higher throughput.
2018	
2019	11.1 High Availability in a Single Switch Topology
2020	--------------------------------------------------
2021	
2022		If two hosts (or a host and a single switch) are directly
2023	connected via multiple physical links, then there is no availability
2024	penalty to optimizing for maximum bandwidth.  In this case, there is
2025	only one switch (or peer), so if it fails, there is no alternative
2026	access to fail over to.  Additionally, the bonding load balance modes
2027	support link monitoring of their members, so if individual links fail,
2028	the load will be rebalanced across the remaining devices.
2029	
2030		See Section 12, "Configuring Bonding for Maximum Throughput"
2031	for information on configuring bonding with one peer device.
2032	
2033	11.2 High Availability in a Multiple Switch Topology
2034	----------------------------------------------------
2035	
2036		With multiple switches, the configuration of bonding and the
2037	network changes dramatically.  In multiple switch topologies, there is
2038	a trade off between network availability and usable bandwidth.
2039	
2040		Below is a sample network, configured to maximize the
2041	availability of the network:
2042	
2043	                |                                     |
2044	                |port3                           port3|
2045	          +-----+----+                          +-----+----+
2046	          |          |port2       ISL      port2|          |
2047	          | switch A +--------------------------+ switch B |
2048	          |          |                          |          |
2049	          +-----+----+                          +-----++---+
2050	                |port1                           port1|
2051	                |             +-------+               |
2052	                +-------------+ host1 +---------------+
2053	                         eth0 +-------+ eth1
2054	
2055		In this configuration, there is a link between the two
2056	switches (ISL, or inter switch link), and multiple ports connecting to
2057	the outside world ("port3" on each switch).  There is no technical
2058	reason that this could not be extended to a third switch.
2059	
2060	11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
2061	-------------------------------------------------------------
2062	
2063		In a topology such as the example above, the active-backup and
2064	broadcast modes are the only useful bonding modes when optimizing for
2065	availability; the other modes require all links to terminate on the
2066	same peer for them to behave rationally.
2067	
2068	active-backup: This is generally the preferred mode, particularly if
2069		the switches have an ISL and play together well.  If the
2070		network configuration is such that one switch is specifically
2071		a backup switch (e.g., has lower capacity, higher cost, etc),
2072		then the primary option can be used to insure that the
2073		preferred link is always used when it is available.
2074	
2075	broadcast: This mode is really a special purpose mode, and is suitable
2076		only for very specific needs.  For example, if the two
2077		switches are not connected (no ISL), and the networks beyond
2078		them are totally independent.  In this case, if it is
2079		necessary for some specific one-way traffic to reach both
2080		independent networks, then the broadcast mode may be suitable.
2081	
2082	11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
2083	----------------------------------------------------------------
2084	
2085		The choice of link monitoring ultimately depends upon your
2086	switch.  If the switch can reliably fail ports in response to other
2087	failures, then either the MII or ARP monitors should work.  For
2088	example, in the above example, if the "port3" link fails at the remote
2089	end, the MII monitor has no direct means to detect this.  The ARP
2090	monitor could be configured with a target at the remote end of port3,
2091	thus detecting that failure without switch support.
2092	
2093		In general, however, in a multiple switch topology, the ARP
2094	monitor can provide a higher level of reliability in detecting end to
2095	end connectivity failures (which may be caused by the failure of any
2096	individual component to pass traffic for any reason).  Additionally,
2097	the ARP monitor should be configured with multiple targets (at least
2098	one for each switch in the network).  This will insure that,
2099	regardless of which switch is active, the ARP monitor has a suitable
2100	target to query.
2101	
2102		Note, also, that of late many switches now support a functionality
2103	generally referred to as "trunk failover."  This is a feature of the
2104	switch that causes the link state of a particular switch port to be set
2105	down (or up) when the state of another switch port goes down (or up).
2106	Its purpose is to propagate link failures from logically "exterior" ports
2107	to the logically "interior" ports that bonding is able to monitor via
2108	miimon.  Availability and configuration for trunk failover varies by
2109	switch, but this can be a viable alternative to the ARP monitor when using
2110	suitable switches.
2111	
2112	12. Configuring Bonding for Maximum Throughput
2113	==============================================
2114	
2115	12.1 Maximizing Throughput in a Single Switch Topology
2116	------------------------------------------------------
2117	
2118		In a single switch configuration, the best method to maximize
2119	throughput depends upon the application and network environment.  The
2120	various load balancing modes each have strengths and weaknesses in
2121	different environments, as detailed below.
2122	
2123		For this discussion, we will break down the topologies into
2124	two categories.  Depending upon the destination of most traffic, we
2125	categorize them into either "gatewayed" or "local" configurations.
2126	
2127		In a gatewayed configuration, the "switch" is acting primarily
2128	as a router, and the majority of traffic passes through this router to
2129	other networks.  An example would be the following:
2130	
2131	
2132	     +----------+                     +----------+
2133	     |          |eth0            port1|          | to other networks
2134	     | Host A   +---------------------+ router   +------------------->
2135	     |          +---------------------+          | Hosts B and C are out
2136	     |          |eth1            port2|          | here somewhere
2137	     +----------+                     +----------+
2138	
2139		The router may be a dedicated router device, or another host
2140	acting as a gateway.  For our discussion, the important point is that
2141	the majority of traffic from Host A will pass through the router to
2142	some other network before reaching its final destination.
2143	
2144		In a gatewayed network configuration, although Host A may
2145	communicate with many other systems, all of its traffic will be sent
2146	and received via one other peer on the local network, the router.
2147	
2148		Note that the case of two systems connected directly via
2149	multiple physical links is, for purposes of configuring bonding, the
2150	same as a gatewayed configuration.  In that case, it happens that all
2151	traffic is destined for the "gateway" itself, not some other network
2152	beyond the gateway.
2153	
2154		In a local configuration, the "switch" is acting primarily as
2155	a switch, and the majority of traffic passes through this switch to
2156	reach other stations on the same network.  An example would be the
2157	following:
2158	
2159	    +----------+            +----------+       +--------+
2160	    |          |eth0   port1|          +-------+ Host B |
2161	    |  Host A  +------------+  switch  |port3  +--------+
2162	    |          +------------+          |                  +--------+
2163	    |          |eth1   port2|          +------------------+ Host C |
2164	    +----------+            +----------+port4             +--------+
2165	
2166	
2167		Again, the switch may be a dedicated switch device, or another
2168	host acting as a gateway.  For our discussion, the important point is
2169	that the majority of traffic from Host A is destined for other hosts
2170	on the same local network (Hosts B and C in the above example).
2171	
2172		In summary, in a gatewayed configuration, traffic to and from
2173	the bonded device will be to the same MAC level peer on the network
2174	(the gateway itself, i.e., the router), regardless of its final
2175	destination.  In a local configuration, traffic flows directly to and
2176	from the final destinations, thus, each destination (Host B, Host C)
2177	will be addressed directly by their individual MAC addresses.
2178	
2179		This distinction between a gatewayed and a local network
2180	configuration is important because many of the load balancing modes
2181	available use the MAC addresses of the local network source and
2182	destination to make load balancing decisions.  The behavior of each
2183	mode is described below.
2184	
2185	
2186	12.1.1 MT Bonding Mode Selection for Single Switch Topology
2187	-----------------------------------------------------------
2188	
2189		This configuration is the easiest to set up and to understand,
2190	although you will have to decide which bonding mode best suits your
2191	needs.  The trade offs for each mode are detailed below:
2192	
2193	balance-rr: This mode is the only mode that will permit a single
2194		TCP/IP connection to stripe traffic across multiple
2195		interfaces. It is therefore the only mode that will allow a
2196		single TCP/IP stream to utilize more than one interface's
2197		worth of throughput.  This comes at a cost, however: the
2198		striping generally results in peer systems receiving packets out
2199		of order, causing TCP/IP's congestion control system to kick
2200		in, often by retransmitting segments.
2201	
2202		It is possible to adjust TCP/IP's congestion limits by
2203		altering the net.ipv4.tcp_reordering sysctl parameter.  The
2204		usual default value is 3, and the maximum useful value is 127.
2205		For a four interface balance-rr bond, expect that a single
2206		TCP/IP stream will utilize no more than approximately 2.3
2207		interface's worth of throughput, even after adjusting
2208		tcp_reordering.
2209	
2210		Note that the fraction of packets that will be delivered out of
2211		order is highly variable, and is unlikely to be zero.  The level
2212		of reordering depends upon a variety of factors, including the
2213		networking interfaces, the switch, and the topology of the
2214		configuration.  Speaking in general terms, higher speed network
2215		cards produce more reordering (due to factors such as packet
2216		coalescing), and a "many to many" topology will reorder at a
2217		higher rate than a "many slow to one fast" configuration.
2218	
2219		Many switches do not support any modes that stripe traffic
2220		(instead choosing a port based upon IP or MAC level addresses);
2221		for those devices, traffic for a particular connection flowing
2222		through the switch to a balance-rr bond will not utilize greater
2223		than one interface's worth of bandwidth.
2224	
2225		If you are utilizing protocols other than TCP/IP, UDP for
2226		example, and your application can tolerate out of order
2227		delivery, then this mode can allow for single stream datagram
2228		performance that scales near linearly as interfaces are added
2229		to the bond.
2230	
2231		This mode requires the switch to have the appropriate ports
2232		configured for "etherchannel" or "trunking."
2233	
2234	active-backup: There is not much advantage in this network topology to
2235		the active-backup mode, as the inactive backup devices are all
2236		connected to the same peer as the primary.  In this case, a
2237		load balancing mode (with link monitoring) will provide the
2238		same level of network availability, but with increased
2239		available bandwidth.  On the plus side, active-backup mode
2240		does not require any configuration of the switch, so it may
2241		have value if the hardware available does not support any of
2242		the load balance modes.
2243	
2244	balance-xor: This mode will limit traffic such that packets destined
2245		for specific peers will always be sent over the same
2246		interface.  Since the destination is determined by the MAC
2247		addresses involved, this mode works best in a "local" network
2248		configuration (as described above), with destinations all on
2249		the same local network.  This mode is likely to be suboptimal
2250		if all your traffic is passed through a single router (i.e., a
2251		"gatewayed" network configuration, as described above).
2252	
2253		As with balance-rr, the switch ports need to be configured for
2254		"etherchannel" or "trunking."
2255	
2256	broadcast: Like active-backup, there is not much advantage to this
2257		mode in this type of network topology.
2258	
2259	802.3ad: This mode can be a good choice for this type of network
2260		topology.  The 802.3ad mode is an IEEE standard, so all peers
2261		that implement 802.3ad should interoperate well.  The 802.3ad
2262		protocol includes automatic configuration of the aggregates,
2263		so minimal manual configuration of the switch is needed
2264		(typically only to designate that some set of devices is
2265		available for 802.3ad).  The 802.3ad standard also mandates
2266		that frames be delivered in order (within certain limits), so
2267		in general single connections will not see misordering of
2268		packets.  The 802.3ad mode does have some drawbacks: the
2269		standard mandates that all devices in the aggregate operate at
2270		the same speed and duplex.  Also, as with all bonding load
2271		balance modes other than balance-rr, no single connection will
2272		be able to utilize more than a single interface's worth of
2273		bandwidth.  
2274	
2275		Additionally, the linux bonding 802.3ad implementation
2276		distributes traffic by peer (using an XOR of MAC addresses),
2277		so in a "gatewayed" configuration, all outgoing traffic will
2278		generally use the same device.  Incoming traffic may also end
2279		up on a single device, but that is dependent upon the
2280		balancing policy of the peer's 8023.ad implementation.  In a
2281		"local" configuration, traffic will be distributed across the
2282		devices in the bond.
2283	
2284		Finally, the 802.3ad mode mandates the use of the MII monitor,
2285		therefore, the ARP monitor is not available in this mode.
2286	
2287	balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
2288		Since the balancing is done according to MAC address, in a
2289		"gatewayed" configuration (as described above), this mode will
2290		send all traffic across a single device.  However, in a
2291		"local" network configuration, this mode balances multiple
2292		local network peers across devices in a vaguely intelligent
2293		manner (not a simple XOR as in balance-xor or 802.3ad mode),
2294		so that mathematically unlucky MAC addresses (i.e., ones that
2295		XOR to the same value) will not all "bunch up" on a single
2296		interface.
2297	
2298		Unlike 802.3ad, interfaces may be of differing speeds, and no
2299		special switch configuration is required.  On the down side,
2300		in this mode all incoming traffic arrives over a single
2301		interface, this mode requires certain ethtool support in the
2302		network device driver of the slave interfaces, and the ARP
2303		monitor is not available.
2304	
2305	balance-alb: This mode is everything that balance-tlb is, and more.
2306		It has all of the features (and restrictions) of balance-tlb,
2307		and will also balance incoming traffic from local network
2308		peers (as described in the Bonding Module Options section,
2309		above).
2310	
2311		The only additional down side to this mode is that the network
2312		device driver must support changing the hardware address while
2313		the device is open.
2314	
2315	12.1.2 MT Link Monitoring for Single Switch Topology
2316	----------------------------------------------------
2317	
2318		The choice of link monitoring may largely depend upon which
2319	mode you choose to use.  The more advanced load balancing modes do not
2320	support the use of the ARP monitor, and are thus restricted to using
2321	the MII monitor (which does not provide as high a level of end to end
2322	assurance as the ARP monitor).
2323	
2324	12.2 Maximum Throughput in a Multiple Switch Topology
2325	-----------------------------------------------------
2326	
2327		Multiple switches may be utilized to optimize for throughput
2328	when they are configured in parallel as part of an isolated network
2329	between two or more systems, for example:
2330	
2331	                       +-----------+
2332	                       |  Host A   | 
2333	                       +-+---+---+-+
2334	                         |   |   |
2335	                +--------+   |   +---------+
2336	                |            |             |
2337	         +------+---+  +-----+----+  +-----+----+
2338	         | Switch A |  | Switch B |  | Switch C |
2339	         +------+---+  +-----+----+  +-----+----+
2340	                |            |             |
2341	                +--------+   |   +---------+
2342	                         |   |   |
2343	                       +-+---+---+-+
2344	                       |  Host B   | 
2345	                       +-----------+
2346	
2347		In this configuration, the switches are isolated from one
2348	another.  One reason to employ a topology such as this is for an
2349	isolated network with many hosts (a cluster configured for high
2350	performance, for example), using multiple smaller switches can be more
2351	cost effective than a single larger switch, e.g., on a network with 24
2352	hosts, three 24 port switches can be significantly less expensive than
2353	a single 72 port switch.
2354	
2355		If access beyond the network is required, an individual host
2356	can be equipped with an additional network device connected to an
2357	external network; this host then additionally acts as a gateway.
2358	
2359	12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2360	-------------------------------------------------------------
2361	
2362		In actual practice, the bonding mode typically employed in
2363	configurations of this type is balance-rr.  Historically, in this
2364	network configuration, the usual caveats about out of order packet
2365	delivery are mitigated by the use of network adapters that do not do
2366	any kind of packet coalescing (via the use of NAPI, or because the
2367	device itself does not generate interrupts until some number of
2368	packets has arrived).  When employed in this fashion, the balance-rr
2369	mode allows individual connections between two hosts to effectively
2370	utilize greater than one interface's bandwidth.
2371	
2372	12.2.2 MT Link Monitoring for Multiple Switch Topology
2373	------------------------------------------------------
2374	
2375		Again, in actual practice, the MII monitor is most often used
2376	in this configuration, as performance is given preference over
2377	availability.  The ARP monitor will function in this topology, but its
2378	advantages over the MII monitor are mitigated by the volume of probes
2379	needed as the number of systems involved grows (remember that each
2380	host in the network is configured with bonding).
2381	
2382	13. Switch Behavior Issues
2383	==========================
2384	
2385	13.1 Link Establishment and Failover Delays
2386	-------------------------------------------
2387	
2388		Some switches exhibit undesirable behavior with regard to the
2389	timing of link up and down reporting by the switch.
2390	
2391		First, when a link comes up, some switches may indicate that
2392	the link is up (carrier available), but not pass traffic over the
2393	interface for some period of time.  This delay is typically due to
2394	some type of autonegotiation or routing protocol, but may also occur
2395	during switch initialization (e.g., during recovery after a switch
2396	failure).  If you find this to be a problem, specify an appropriate
2397	value to the updelay bonding module option to delay the use of the
2398	relevant interface(s).
2399	
2400		Second, some switches may "bounce" the link state one or more
2401	times while a link is changing state.  This occurs most commonly while
2402	the switch is initializing.  Again, an appropriate updelay value may
2403	help.
2404	
2405		Note that when a bonding interface has no active links, the
2406	driver will immediately reuse the first link that goes up, even if the
2407	updelay parameter has been specified (the updelay is ignored in this
2408	case).  If there are slave interfaces waiting for the updelay timeout
2409	to expire, the interface that first went into that state will be
2410	immediately reused.  This reduces down time of the network if the
2411	value of updelay has been overestimated, and since this occurs only in
2412	cases with no connectivity, there is no additional penalty for
2413	ignoring the updelay.
2414	
2415		In addition to the concerns about switch timings, if your
2416	switches take a long time to go into backup mode, it may be desirable
2417	to not activate a backup interface immediately after a link goes down.
2418	Failover may be delayed via the downdelay bonding module option.
2419	
2420	13.2 Duplicated Incoming Packets
2421	--------------------------------
2422	
2423		NOTE: Starting with version 3.0.2, the bonding driver has logic to
2424	suppress duplicate packets, which should largely eliminate this problem.
2425	The following description is kept for reference.
2426	
2427		It is not uncommon to observe a short burst of duplicated
2428	traffic when the bonding device is first used, or after it has been
2429	idle for some period of time.  This is most easily observed by issuing
2430	a "ping" to some other host on the network, and noticing that the
2431	output from ping flags duplicates (typically one per slave).
2432	
2433		For example, on a bond in active-backup mode with five slaves
2434	all connected to one switch, the output may appear as follows:
2435	
2436	# ping -n 10.0.4.2
2437	PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
2438	64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
2439	64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2440	64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2441	64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2442	64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
2443	64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
2444	64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
2445	64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2446	
2447		This is not due to an error in the bonding driver, rather, it
2448	is a side effect of how many switches update their MAC forwarding
2449	tables.  Initially, the switch does not associate the MAC address in
2450	the packet with a particular switch port, and so it may send the
2451	traffic to all ports until its MAC forwarding table is updated.  Since
2452	the interfaces attached to the bond may occupy multiple ports on a
2453	single switch, when the switch (temporarily) floods the traffic to all
2454	ports, the bond device receives multiple copies of the same packet
2455	(one per slave device).
2456	
2457		The duplicated packet behavior is switch dependent, some
2458	switches exhibit this, and some do not.  On switches that display this
2459	behavior, it can be induced by clearing the MAC forwarding table (on
2460	most Cisco switches, the privileged command "clear mac address-table
2461	dynamic" will accomplish this).
2462	
2463	14. Hardware Specific Considerations
2464	====================================
2465	
2466		This section contains additional information for configuring
2467	bonding on specific hardware platforms, or for interfacing bonding
2468	with particular switches or other devices.
2469	
2470	14.1 IBM BladeCenter
2471	--------------------
2472	
2473		This applies to the JS20 and similar systems.
2474	
2475		On the JS20 blades, the bonding driver supports only
2476	balance-rr, active-backup, balance-tlb and balance-alb modes.  This is
2477	largely due to the network topology inside the BladeCenter, detailed
2478	below.
2479	
2480	JS20 network adapter information
2481	--------------------------------
2482	
2483		All JS20s come with two Broadcom Gigabit Ethernet ports
2484	integrated on the planar (that's "motherboard" in IBM-speak).  In the
2485	BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2486	I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2487	An add-on Broadcom daughter card can be installed on a JS20 to provide
2488	two more Gigabit Ethernet ports.  These ports, eth2 and eth3, are
2489	wired to I/O Modules 3 and 4, respectively.
2490	
2491		Each I/O Module may contain either a switch or a passthrough
2492	module (which allows ports to be directly connected to an external
2493	switch).  Some bonding modes require a specific BladeCenter internal
2494	network topology in order to function; these are detailed below.
2495	
2496		Additional BladeCenter-specific networking information can be
2497	found in two IBM Redbooks (www.ibm.com/redbooks):
2498	
2499	"IBM eServer BladeCenter Networking Options"
2500	"IBM eServer BladeCenter Layer 2-7 Network Switching"
2501	
2502	BladeCenter networking configuration
2503	------------------------------------
2504	
2505		Because a BladeCenter can be configured in a very large number
2506	of ways, this discussion will be confined to describing basic
2507	configurations.
2508	
2509		Normally, Ethernet Switch Modules (ESMs) are used in I/O
2510	modules 1 and 2.  In this configuration, the eth0 and eth1 ports of a
2511	JS20 will be connected to different internal switches (in the
2512	respective I/O modules).
2513	
2514		A passthrough module (OPM or CPM, optical or copper,
2515	passthrough module) connects the I/O module directly to an external
2516	switch.  By using PMs in I/O module #1 and #2, the eth0 and eth1
2517	interfaces of a JS20 can be redirected to the outside world and
2518	connected to a common external switch.
2519	
2520		Depending upon the mix of ESMs and PMs, the network will
2521	appear to bonding as either a single switch topology (all PMs) or as a
2522	multiple switch topology (one or more ESMs, zero or more PMs).  It is
2523	also possible to connect ESMs together, resulting in a configuration
2524	much like the example in "High Availability in a Multiple Switch
2525	Topology," above.
2526	
2527	Requirements for specific modes
2528	-------------------------------
2529	
2530		The balance-rr mode requires the use of passthrough modules
2531	for devices in the bond, all connected to an common external switch.
2532	That switch must be configured for "etherchannel" or "trunking" on the
2533	appropriate ports, as is usual for balance-rr.
2534	
2535		The balance-alb and balance-tlb modes will function with
2536	either switch modules or passthrough modules (or a mix).  The only
2537	specific requirement for these modes is that all network interfaces
2538	must be able to reach all destinations for traffic sent over the
2539	bonding device (i.e., the network must converge at some point outside
2540	the BladeCenter).
2541	
2542		The active-backup mode has no additional requirements.
2543	
2544	Link monitoring issues
2545	----------------------
2546	
2547		When an Ethernet Switch Module is in place, only the ARP
2548	monitor will reliably detect link loss to an external switch.  This is
2549	nothing unusual, but examination of the BladeCenter cabinet would
2550	suggest that the "external" network ports are the ethernet ports for
2551	the system, when it fact there is a switch between these "external"
2552	ports and the devices on the JS20 system itself.  The MII monitor is
2553	only able to detect link failures between the ESM and the JS20 system.
2554	
2555		When a passthrough module is in place, the MII monitor does
2556	detect failures to the "external" port, which is then directly
2557	connected to the JS20 system.
2558	
2559	Other concerns
2560	--------------
2561	
2562		The Serial Over LAN (SoL) link is established over the primary
2563	ethernet (eth0) only, therefore, any loss of link to eth0 will result
2564	in losing your SoL connection.  It will not fail over with other
2565	network traffic, as the SoL system is beyond the control of the
2566	bonding driver.
2567	
2568		It may be desirable to disable spanning tree on the switch
2569	(either the internal Ethernet Switch Module, or an external switch) to
2570	avoid fail-over delay issues when using bonding.
2571	
2572		
2573	15. Frequently Asked Questions
2574	==============================
2575	
2576	1.  Is it SMP safe?
2577	
2578		Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2579	The new driver was designed to be SMP safe from the start.
2580	
2581	2.  What type of cards will work with it?
2582	
2583		Any Ethernet type cards (you can even mix cards - a Intel
2584	EtherExpress PRO/100 and a 3com 3c905b, for example).  For most modes,
2585	devices need not be of the same speed.
2586	
2587		Starting with version 3.2.1, bonding also supports Infiniband
2588	slaves in active-backup mode.
2589	
2590	3.  How many bonding devices can I have?
2591	
2592		There is no limit.
2593	
2594	4.  How many slaves can a bonding device have?
2595	
2596		This is limited only by the number of network interfaces Linux
2597	supports and/or the number of network cards you can place in your
2598	system.
2599	
2600	5.  What happens when a slave link dies?
2601	
2602		If link monitoring is enabled, then the failing device will be
2603	disabled.  The active-backup mode will fail over to a backup link, and
2604	other modes will ignore the failed link.  The link will continue to be
2605	monitored, and should it recover, it will rejoin the bond (in whatever
2606	manner is appropriate for the mode). See the sections on High
2607	Availability and the documentation for each mode for additional
2608	information.
2609		
2610		Link monitoring can be enabled via either the miimon or
2611	arp_interval parameters (described in the module parameters section,
2612	above).  In general, miimon monitors the carrier state as sensed by
2613	the underlying network device, and the arp monitor (arp_interval)
2614	monitors connectivity to another host on the local network.
2615	
2616		If no link monitoring is configured, the bonding driver will
2617	be unable to detect link failures, and will assume that all links are
2618	always available.  This will likely result in lost packets, and a
2619	resulting degradation of performance.  The precise performance loss
2620	depends upon the bonding mode and network configuration.
2621	
2622	6.  Can bonding be used for High Availability?
2623	
2624		Yes.  See the section on High Availability for details.
2625	
2626	7.  Which switches/systems does it work with?
2627	
2628		The full answer to this depends upon the desired mode.
2629	
2630		In the basic balance modes (balance-rr and balance-xor), it
2631	works with any system that supports etherchannel (also called
2632	trunking).  Most managed switches currently available have such
2633	support, and many unmanaged switches as well.
2634	
2635		The advanced balance modes (balance-tlb and balance-alb) do
2636	not have special switch requirements, but do need device drivers that
2637	support specific features (described in the appropriate section under
2638	module parameters, above).
2639	
2640		In 802.3ad mode, it works with systems that support IEEE
2641	802.3ad Dynamic Link Aggregation.  Most managed and many unmanaged
2642	switches currently available support 802.3ad.
2643	
2644	        The active-backup mode should work with any Layer-II switch.
2645	
2646	8.  Where does a bonding device get its MAC address from?
2647	
2648		When using slave devices that have fixed MAC addresses, or when
2649	the fail_over_mac option is enabled, the bonding device's MAC address is
2650	the MAC address of the active slave.
2651	
2652		For other configurations, if not explicitly configured (with
2653	ifconfig or ip link), the MAC address of the bonding device is taken from
2654	its first slave device.  This MAC address is then passed to all following
2655	slaves and remains persistent (even if the first slave is removed) until
2656	the bonding device is brought down or reconfigured.
2657	
2658		If you wish to change the MAC address, you can set it with
2659	ifconfig or ip link:
2660	
2661	# ifconfig bond0 hw ether 00:11:22:33:44:55
2662	
2663	# ip link set bond0 address 66:77:88:99:aa:bb
2664	
2665		The MAC address can be also changed by bringing down/up the
2666	device and then changing its slaves (or their order):
2667	
2668	# ifconfig bond0 down ; modprobe -r bonding
2669	# ifconfig bond0 .... up
2670	# ifenslave bond0 eth...
2671	
2672		This method will automatically take the address from the next
2673	slave that is added.
2674	
2675		To restore your slaves' MAC addresses, you need to detach them
2676	from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2677	then restore the MAC addresses that the slaves had before they were
2678	enslaved.
2679	
2680	16. Resources and Links
2681	=======================
2682	
2683		The latest version of the bonding driver can be found in the latest
2684	version of the linux kernel, found on http://kernel.org
2685	
2686		The latest version of this document can be found in the latest kernel
2687	source (named Documentation/networking/bonding.txt).
2688	
2689		Discussions regarding the usage of the bonding driver take place on the
2690	bonding-devel mailing list, hosted at sourceforge.net. If you have questions or
2691	problems, post them to the list.  The list address is:
2692	
2693	bonding-devel@lists.sourceforge.net
2694	
2695		The administrative interface (to subscribe or unsubscribe) can
2696	be found at:
2697	
2698	https://lists.sourceforge.net/lists/listinfo/bonding-devel
2699	
2700		Discussions regarding the development of the bonding driver take place
2701	on the main Linux network mailing list, hosted at vger.kernel.org. The list
2702	address is:
2703	
2704	netdev@vger.kernel.org
2705	
2706		The administrative interface (to subscribe or unsubscribe) can
2707	be found at:
2708	
2709	http://vger.kernel.org/vger-lists.html#netdev
2710	
2711	Donald Becker's Ethernet Drivers and diag programs may be found at :
2712	 - http://web.archive.org/web/*/http://www.scyld.com/network/ 
2713	
2714	You will also find a lot of information regarding Ethernet, NWay, MII,
2715	etc. at www.scyld.com.
2716	
2717	-- END --
Hide Line Numbers
About Kernel Documentation Linux Kernel Contact Linux Resources Linux Blog

Information is copyright its respective author. All material is available from the Linux Kernel Source distributed under a GPL License. This page is provided as a free service by mjmwired.net.