blob: e3abfbd32f71ebd03b7da401d232e0334a897260 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002 Linux Ethernet Bonding Driver HOWTO
3
Ben Hutchingsad246c992011-04-26 15:25:52 +00004 Latest update: 27 April 2011
Linus Torvalds1da177e2005-04-16 15:20:36 -07005
6Initial release : Thomas Davis <tadavis at lbl.gov>
7Corrections, 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
14Reorganized and updated Feb 2005 by Jay Vosburgh
Auke Kok6224e012006-06-08 11:15:35 -070015Added Sysfs information: 2006/04/24
16 - Mitch Williams <mitch.a.williams at intel.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017
Jay Vosburgh00354cf2005-07-21 12:18:02 -070018Introduction
19============
Linus Torvalds1da177e2005-04-16 15:20:36 -070020
Jay Vosburgh00354cf2005-07-21 12:18:02 -070021 The Linux bonding driver provides a method for aggregating
22multiple network interfaces into a single logical "bonded" interface.
23The behavior of the bonded interfaces depends upon the mode; generally
24speaking, modes provide either hot standby or load balancing services.
25Additionally, link integrity monitoring may be performed.
26
27 The bonding driver originally came from Donald Becker's
28beowulf patches for kernel 2.0. It has changed quite a bit since, and
29the original tools from extreme-linux and beowulf sites will not work
30with this version of the driver.
31
32 For new versions of the driver, updated userspace tools, and
33who to ask for help, please follow the links at the end of this file.
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
35Table of Contents
36=================
37
381. Bonding Driver Installation
39
402. Bonding Driver Options
41
423. Configuring Bonding Devices
Auke Kok6224e012006-06-08 11:15:35 -0700433.1 Configuration with Sysconfig Support
443.1.1 Using DHCP with Sysconfig
453.1.2 Configuring Multiple Bonds with Sysconfig
463.2 Configuration with Initscripts Support
473.2.1 Using DHCP with Initscripts
483.2.2 Configuring Multiple Bonds with Initscripts
493.3 Configuring Bonding Manually with Ifenslave
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700503.3.1 Configuring Multiple Bonds Manually
Auke Kok6224e012006-06-08 11:15:35 -0700513.4 Configuring Bonding Manually via Sysfs
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000523.5 Configuration with Interfaces Support
533.6 Overriding Configuration for Special Cases
Mahesh Bandeward22a5fc2015-05-09 00:01:57 -0700543.7 Configuring LACP for 802.3ad mode in a more secure way
Linus Torvalds1da177e2005-04-16 15:20:36 -070055
Auke Kok6224e012006-06-08 11:15:35 -0700564. Querying Bonding Configuration
574.1 Bonding Configuration
584.2 Network Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
Auke Kok6224e012006-06-08 11:15:35 -0700605. Switch Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
Auke Kok6224e012006-06-08 11:15:35 -0700626. 802.1q VLAN Support
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Auke Kok6224e012006-06-08 11:15:35 -0700647. Link Monitoring
657.1 ARP Monitor Operation
667.2 Configuring Multiple ARP Targets
677.3 MII Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -070068
Auke Kok6224e012006-06-08 11:15:35 -0700698. Potential Trouble Sources
708.1 Adventures in Routing
718.2 Ethernet Device Renaming
728.3 Painfully Slow Or No Failed Link Detection By Miimon
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
Auke Kok6224e012006-06-08 11:15:35 -0700749. SNMP agents
Linus Torvalds1da177e2005-04-16 15:20:36 -070075
Auke Kok6224e012006-06-08 11:15:35 -07007610. Promiscuous mode
Linus Torvalds1da177e2005-04-16 15:20:36 -070077
Auke Kok6224e012006-06-08 11:15:35 -07007811. Configuring Bonding for High Availability
7911.1 High Availability in a Single Switch Topology
8011.2 High Availability in a Multiple Switch Topology
8111.2.1 HA Bonding Mode Selection for Multiple Switch Topology
8211.2.2 HA Link Monitoring for Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -070083
Auke Kok6224e012006-06-08 11:15:35 -07008412. Configuring Bonding for Maximum Throughput
8512.1 Maximum Throughput in a Single Switch Topology
8612.1.1 MT Bonding Mode Selection for Single Switch Topology
8712.1.2 MT Link Monitoring for Single Switch Topology
8812.2 Maximum Throughput in a Multiple Switch Topology
8912.2.1 MT Bonding Mode Selection for Multiple Switch Topology
9012.2.2 MT Link Monitoring for Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
Auke Kok6224e012006-06-08 11:15:35 -07009213. Switch Behavior Issues
9313.1 Link Establishment and Failover Delays
9413.2 Duplicated Incoming Packets
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Auke Kok6224e012006-06-08 11:15:35 -07009614. Hardware Specific Considerations
9714.1 IBM BladeCenter
Jay Vosburgh00354cf2005-07-21 12:18:02 -070098
Auke Kok6224e012006-06-08 11:15:35 -07009915. Frequently Asked Questions
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700100
Auke Kok6224e012006-06-08 11:15:35 -070010116. Resources and Links
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102
103
1041. Bonding Driver Installation
105==============================
106
107 Most popular distro kernels ship with the bonding driver
Cong Wangb1098bb2013-05-27 15:49:16 +0000108already available as a module. If your distro does not, or you
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109have need to compile bonding from source (e.g., configuring and
110installing a mainline kernel from kernel.org), you'll need to perform
111the following steps:
112
1131.1 Configure and build the kernel with bonding
114-----------------------------------------------
115
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700116 The current version of the bonding driver is available in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117drivers/net/bonding subdirectory of the most recent kernel source
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700118(which is available on http://kernel.org). Most users "rolling their
119own" will want to use the most recent kernel from kernel.org.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
121 Configure kernel with "make menuconfig" (or "make xconfig" or
122"make config"), then select "Bonding driver support" in the "Network
123device support" section. It is recommended that you configure the
124driver as module since it is currently the only way to pass parameters
125to the driver or configure more than one bonding device.
126
Cong Wangb1098bb2013-05-27 15:49:16 +0000127 Build and install the new kernel and modules.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
Cong Wangb1098bb2013-05-27 15:49:16 +00001291.2 Bonding Control Utility
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130-------------------------------------
131
Cong Wangb1098bb2013-05-27 15:49:16 +0000132 It is recommended to configure bonding via iproute2 (netlink)
133or sysfs, the old ifenslave control utility is obsolete.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134
1352. Bonding Driver Options
136=========================
137
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800138 Options for the bonding driver are supplied as parameters to the
139bonding module at load time, or are specified via sysfs.
140
141 Module options may be given as command line arguments to the
142insmod or modprobe command, but are usually specified in either the
Tonghao Zhang804c1462018-05-11 02:53:12 -0700143/etc/modprobe.d/*.conf configuration files, or in a distro-specific
Lucas De Marchi970e2482012-03-30 13:37:16 -0700144configuration file (some of which are detailed in the next section).
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800145
146 Details on bonding support for sysfs is provided in the
147"Configuring Bonding Manually via Sysfs" section, below.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
149 The available bonding driver parameters are listed below. If a
150parameter is not specified the default value is used. When initially
151configuring a bond, it is recommended "tail -f /var/log/messages" be
152run in a separate window to watch for bonding driver error messages.
153
154 It is critical that either the miimon or arp_interval and
155arp_ip_target parameters be specified, otherwise serious network
156degradation will occur during link failures. Very few devices do not
157support at least miimon, so there is really no reason not to use it.
158
159 Options with textual values will accept either the text name
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700160or, for backwards compatibility, the option value. E.g.,
161"mode=802.3ad" and "mode=4" set the same mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162
163 The parameters are as follows:
164
Nicolas de Pesloüan1ba9ac72011-12-26 13:35:24 +0000165active_slave
166
167 Specifies the new active slave for modes that support it
168 (active-backup, balance-alb and balance-tlb). Possible values
169 are the name of any currently enslaved interface, or an empty
170 string. If a name is given, the slave and its link must be up in order
171 to be selected as the new active slave. If an empty string is
172 specified, the current active slave is cleared, and a new active
173 slave is selected automatically.
174
175 Note that this is only available through the sysfs interface. No module
176 parameter by this name exists.
177
178 The normal value of this option is the name of the currently
179 active slave, or the empty string if there is no active slave or
180 the current mode does not use an active slave.
181
Mahesh Bandewar6791e462015-05-09 00:01:55 -0700182ad_actor_sys_prio
183
184 In an AD system, this specifies the system priority. The allowed range
185 is 1 - 65535. If the value is not specified, it takes 65535 as the
186 default value.
187
188 This parameter has effect only in 802.3ad mode and is available through
189 SysFs interface.
190
Mahesh Bandewar74514952015-05-09 00:01:56 -0700191ad_actor_system
192
193 In an AD system, this specifies the mac-address for the actor in
194 protocol packet exchanges (LACPDUs). The value cannot be NULL or
195 multicast. It is preferred to have the local-admin bit set for this
196 mac but driver does not enforce it. If the value is not given then
197 system defaults to using the masters' mac address as actors' system
198 address.
199
200 This parameter has effect only in 802.3ad mode and is available through
201 SysFs interface.
202
Jay Vosburghfd989c82008-11-04 17:51:16 -0800203ad_select
204
205 Specifies the 802.3ad aggregation selection logic to use. The
206 possible values and their effects are:
207
208 stable or 0
209
210 The active aggregator is chosen by largest aggregate
211 bandwidth.
212
213 Reselection of the active aggregator occurs only when all
214 slaves of the active aggregator are down or the active
215 aggregator has no slaves.
216
217 This is the default value.
218
219 bandwidth or 1
220
221 The active aggregator is chosen by largest aggregate
222 bandwidth. Reselection occurs if:
223
224 - A slave is added to or removed from the bond
225
226 - Any slave's link state changes
227
228 - Any slave's 802.3ad association state changes
229
Matt LaPlante19f59462009-04-27 15:06:31 +0200230 - The bond's administrative state changes to up
Jay Vosburghfd989c82008-11-04 17:51:16 -0800231
232 count or 2
233
234 The active aggregator is chosen by the largest number of
235 ports (slaves). Reselection occurs as described under the
236 "bandwidth" setting, above.
237
238 The bandwidth and count selection policies permit failover of
239 802.3ad aggregations when partial failure of the active aggregator
240 occurs. This keeps the aggregator with the highest availability
241 (either in bandwidth or in number of ports) active at all times.
242
243 This option was added in bonding version 3.4.0.
244
Mahesh Bandeward22a5fc2015-05-09 00:01:57 -0700245ad_user_port_key
246
247 In an AD system, the port-key has three parts as shown below -
248
249 Bits Use
250 00 Duplex
251 01-05 Speed
252 06-15 User-defined
253
254 This defines the upper 10 bits of the port key. The values can be
255 from 0 - 1023. If not given, the system defaults to 0.
256
257 This parameter has effect only in 802.3ad mode and is available through
258 SysFs interface.
259
Nicolas de Pesloüan025890b2011-08-06 07:06:39 +0000260all_slaves_active
261
262 Specifies that duplicate frames (received on inactive ports) should be
263 dropped (0) or delivered (1).
264
265 Normally, bonding will drop duplicate frames (received on inactive
266 ports), which is desirable for most users. But there are some times
267 it is nice to allow duplicate frames to be delivered.
268
269 The default value is 0 (drop duplicate frames received on inactive
270 ports).
271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272arp_interval
273
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700274 Specifies the ARP link monitoring frequency in milliseconds.
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700275
276 The ARP monitor works by periodically checking the slave
277 devices to determine whether they have sent or received
278 traffic recently (the precise criteria depends upon the
279 bonding mode, and the state of the slave). Regular traffic is
280 generated via ARP probes issued for the addresses specified by
281 the arp_ip_target option.
282
283 This behavior can be modified by the arp_validate option,
284 below.
285
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700286 If ARP monitoring is used in an etherchannel compatible mode
287 (modes 0 and 2), the switch should be configured in a mode
288 that evenly distributes packets across all links. If the
289 switch is configured to distribute the packets in an XOR
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 fashion, all replies from the ARP targets will be received on
291 the same link which could cause the other team members to
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700292 fail. ARP monitoring should not be used in conjunction with
293 miimon. A value of 0 disables ARP monitoring. The default
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 value is 0.
295
296arp_ip_target
297
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700298 Specifies the IP addresses to use as ARP monitoring peers when
299 arp_interval is > 0. These are the targets of the ARP request
300 sent to determine the health of the link to the targets.
301 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
302 addresses must be separated by a comma. At least one IP
303 address must be given for ARP monitoring to function. The
304 maximum number of targets that can be specified is 16. The
305 default value is no IP addresses.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700307arp_validate
308
309 Specifies whether or not ARP probes and replies should be
Veaceslav Falico52f65ef2014-02-18 07:48:41 +0100310 validated in any mode that supports arp monitoring, or whether
311 non-ARP traffic should be filtered (disregarded) for link
312 monitoring purposes.
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700313
314 Possible values are:
315
316 none or 0
317
Veaceslav Falico52f65ef2014-02-18 07:48:41 +0100318 No validation or filtering is performed.
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700319
320 active or 1
321
322 Validation is performed only for the active slave.
323
324 backup or 2
325
326 Validation is performed only for backup slaves.
327
328 all or 3
329
330 Validation is performed for all slaves.
331
Veaceslav Falico52f65ef2014-02-18 07:48:41 +0100332 filter or 4
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700333
Veaceslav Falico52f65ef2014-02-18 07:48:41 +0100334 Filtering is applied to all slaves. No validation is
335 performed.
Veaceslav Falicod7d35c62013-06-24 11:49:33 +0200336
Veaceslav Falico52f65ef2014-02-18 07:48:41 +0100337 filter_active or 5
338
339 Filtering is applied to all slaves, validation is performed
340 only for the active slave.
341
342 filter_backup or 6
343
344 Filtering is applied to all slaves, validation is performed
345 only for backup slaves.
346
347 Validation:
348
349 Enabling validation causes the ARP monitor to examine the incoming
350 ARP requests and replies, and only consider a slave to be up if it
351 is receiving the appropriate ARP traffic.
352
353 For an active slave, the validation checks ARP replies to confirm
354 that they were generated by an arp_ip_target. Since backup slaves
355 do not typically receive these replies, the validation performed
356 for backup slaves is on the broadcast ARP request sent out via the
357 active slave. It is possible that some switch or network
358 configurations may result in situations wherein the backup slaves
359 do not receive the ARP requests; in such a situation, validation
360 of backup slaves must be disabled.
361
362 The validation of ARP requests on backup slaves is mainly helping
363 bonding to decide which slaves are more likely to work in case of
364 the active slave failure, it doesn't really guarantee that the
365 backup slave will work if it's selected as the next active slave.
366
367 Validation is useful in network configurations in which multiple
368 bonding hosts are concurrently issuing ARPs to one or more targets
369 beyond a common switch. Should the link between the switch and
370 target fail (but not the switch itself), the probe traffic
371 generated by the multiple bonding instances will fool the standard
372 ARP monitor into considering the links as still up. Use of
373 validation can resolve this, as the ARP monitor will only consider
374 ARP requests and replies associated with its own instance of
375 bonding.
376
377 Filtering:
378
379 Enabling filtering causes the ARP monitor to only use incoming ARP
380 packets for link availability purposes. Arriving packets that are
381 not ARPs are delivered normally, but do not count when determining
382 if a slave is available.
383
384 Filtering operates by only considering the reception of ARP
385 packets (any ARP packet, regardless of source or destination) when
386 determining if a slave has received traffic for link availability
387 purposes.
388
389 Filtering is useful in network configurations in which significant
390 levels of third party broadcast traffic would fool the standard
391 ARP monitor into considering the links as still up. Use of
392 filtering can resolve this, as only ARP traffic is considered for
393 link availability purposes.
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700394
395 This option was added in bonding version 3.1.0.
396
Veaceslav Falico8599b522013-06-24 11:49:34 +0200397arp_all_targets
398
399 Specifies the quantity of arp_ip_targets that must be reachable
400 in order for the ARP monitor to consider a slave as being up.
401 This option affects only active-backup mode for slaves with
402 arp_validation enabled.
403
404 Possible values are:
405
406 any or 0
407
408 consider the slave up only when any of the arp_ip_targets
409 is reachable
410
411 all or 1
412
413 consider the slave up only when all of the arp_ip_targets
414 are reachable
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416downdelay
417
418 Specifies the time, in milliseconds, to wait before disabling
419 a slave after a link failure has been detected. This option
420 is only valid for the miimon link monitor. The downdelay
421 value should be a multiple of the miimon value; if not, it
422 will be rounded down to the nearest multiple. The default
423 value is 0.
424
Jay Vosburghdd957c52007-10-09 19:57:24 -0700425fail_over_mac
426
427 Specifies whether active-backup mode should set all slaves to
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700428 the same MAC address at enslavement (the traditional
429 behavior), or, when enabled, perform special handling of the
430 bond's MAC address in accordance with the selected policy.
Jay Vosburghdd957c52007-10-09 19:57:24 -0700431
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700432 Possible values are:
Jay Vosburghdd957c52007-10-09 19:57:24 -0700433
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700434 none or 0
Jay Vosburghdd957c52007-10-09 19:57:24 -0700435
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700436 This setting disables fail_over_mac, and causes
437 bonding to set all slaves of an active-backup bond to
438 the same MAC address at enslavement time. This is the
439 default.
Jay Vosburghdd957c52007-10-09 19:57:24 -0700440
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700441 active or 1
Jay Vosburghdd957c52007-10-09 19:57:24 -0700442
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700443 The "active" fail_over_mac policy indicates that the
444 MAC address of the bond should always be the MAC
445 address of the currently active slave. The MAC
446 address of the slaves is not changed; instead, the MAC
447 address of the bond changes during a failover.
448
449 This policy is useful for devices that cannot ever
450 alter their MAC address, or for devices that refuse
451 incoming broadcasts with their own source MAC (which
452 interferes with the ARP monitor).
453
454 The down side of this policy is that every device on
455 the network must be updated via gratuitous ARP,
456 vs. just updating a switch or set of switches (which
457 often takes place for any traffic, not just ARP
458 traffic, if the switch snoops incoming traffic to
459 update its tables) for the traditional method. If the
460 gratuitous ARP is lost, communication may be
461 disrupted.
462
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300463 When this policy is used in conjunction with the mii
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700464 monitor, devices which assert link up prior to being
465 able to actually transmit and receive are particularly
Matt LaPlante19f59462009-04-27 15:06:31 +0200466 susceptible to loss of the gratuitous ARP, and an
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700467 appropriate updelay setting may be required.
468
469 follow or 2
470
471 The "follow" fail_over_mac policy causes the MAC
472 address of the bond to be selected normally (normally
473 the MAC address of the first slave added to the bond).
474 However, the second and subsequent slaves are not set
475 to this MAC address while they are in a backup role; a
476 slave is programmed with the bond's MAC address at
477 failover time (and the formerly active slave receives
478 the newly active slave's MAC address).
479
480 This policy is useful for multiport devices that
481 either become confused or incur a performance penalty
482 when multiple ports are programmed with the same MAC
483 address.
484
485
486 The default policy is none, unless the first slave cannot
487 change its MAC address, in which case the active policy is
488 selected by default.
489
490 This option may be modified via sysfs only when no slaves are
491 present in the bond.
492
493 This option was added in bonding version 3.2.0. The "follow"
494 policy was added in bonding version 3.3.0.
Jay Vosburghdd957c52007-10-09 19:57:24 -0700495
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496lacp_rate
497
498 Option specifying the rate in which we'll ask our link partner
499 to transmit LACPDU packets in 802.3ad mode. Possible values
500 are:
501
502 slow or 0
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700503 Request partner to transmit LACPDUs every 30 seconds
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
505 fast or 1
506 Request partner to transmit LACPDUs every 1 second
507
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700508 The default is slow.
509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510max_bonds
511
512 Specifies the number of bonding devices to create for this
513 instance of the bonding driver. E.g., if max_bonds is 3, and
514 the bonding driver is not already loaded, then bond0, bond1
Jay Vosburghb8a97872008-06-13 18:12:04 -0700515 and bond2 will be created. The default value is 1. Specifying
516 a value of 0 will load bonding, but will not create any devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
518miimon
519
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700520 Specifies the MII link monitoring frequency in milliseconds.
521 This determines how often the link state of each slave is
522 inspected for link failures. A value of zero disables MII
523 link monitoring. A value of 100 is a good starting point.
524 The use_carrier option, below, affects how the link state is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 determined. See the High Availability section for additional
526 information. The default value is 0.
527
Nicolas de Pesloüan025890b2011-08-06 07:06:39 +0000528min_links
529
530 Specifies the minimum number of links that must be active before
531 asserting carrier. It is similar to the Cisco EtherChannel min-links
532 feature. This allows setting the minimum number of member ports that
533 must be up (link-up state) before marking the bond device as up
534 (carrier on). This is useful for situations where higher level services
535 such as clustering want to ensure a minimum number of low bandwidth
536 links are active before switchover. This option only affect 802.3ad
537 mode.
538
539 The default value is 0. This will cause carrier to be asserted (for
540 802.3ad mode) whenever there is an active aggregator, regardless of the
541 number of available links in that aggregator. Note that, because an
542 aggregator cannot be active without at least one available link,
543 setting this option to 0 or to 1 has the exact same effect.
544
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545mode
546
547 Specifies one of the bonding policies. The default is
548 balance-rr (round robin). Possible values are:
549
550 balance-rr or 0
551
552 Round-robin policy: Transmit packets in sequential
553 order from the first available slave through the
554 last. This mode provides load balancing and fault
555 tolerance.
556
557 active-backup or 1
558
559 Active-backup policy: Only one slave in the bond is
560 active. A different slave becomes active if, and only
561 if, the active slave fails. The bond's MAC address is
562 externally visible on only one port (network adapter)
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700563 to avoid confusing the switch.
564
565 In bonding version 2.6.2 or later, when a failover
566 occurs in active-backup mode, bonding will issue one
567 or more gratuitous ARPs on the newly active slave.
Auke Kok6224e012006-06-08 11:15:35 -0700568 One gratuitous ARP is issued for the bonding master
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700569 interface and each VLAN interfaces configured above
570 it, provided that the interface has at least one IP
571 address configured. Gratuitous ARPs issued for VLAN
572 interfaces are tagged with the appropriate VLAN id.
573
574 This mode provides fault tolerance. The primary
575 option, documented below, affects the behavior of this
576 mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577
578 balance-xor or 2
579
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700580 XOR policy: Transmit based on the selected transmit
581 hash policy. The default policy is a simple [(source
Jianhua Xie92abf752014-07-17 14:16:26 +0800582 MAC address XOR'd with destination MAC address XOR
583 packet type ID) modulo slave count]. Alternate transmit
584 policies may be selected via the xmit_hash_policy option,
585 described below.
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700586
587 This mode provides load balancing and fault tolerance.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
589 broadcast or 3
590
591 Broadcast policy: transmits everything on all slave
592 interfaces. This mode provides fault tolerance.
593
594 802.3ad or 4
595
596 IEEE 802.3ad Dynamic link aggregation. Creates
597 aggregation groups that share the same speed and
598 duplex settings. Utilizes all slaves in the active
599 aggregator according to the 802.3ad specification.
600
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700601 Slave selection for outgoing traffic is done according
602 to the transmit hash policy, which may be changed from
603 the default simple XOR policy via the xmit_hash_policy
604 option, documented below. Note that not all transmit
605 policies may be 802.3ad compliant, particularly in
606 regards to the packet mis-ordering requirements of
607 section 43.2.4 of the 802.3ad standard. Differing
608 peer implementations will have varying tolerances for
609 noncompliance.
610
611 Prerequisites:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613 1. Ethtool support in the base drivers for retrieving
614 the speed and duplex of each slave.
615
616 2. A switch that supports IEEE 802.3ad Dynamic link
617 aggregation.
618
619 Most switches will require some type of configuration
620 to enable 802.3ad mode.
621
622 balance-tlb or 5
623
624 Adaptive transmit load balancing: channel bonding that
Mahesh Bandeware9f0fb82014-04-22 16:30:22 -0700625 does not require any special switch support.
626
627 In tlb_dynamic_lb=1 mode; the outgoing traffic is
628 distributed according to the current load (computed
629 relative to the speed) on each slave.
630
631 In tlb_dynamic_lb=0 mode; the load balancing based on
632 current load is disabled and the load is distributed
633 only using the hash distribution.
634
635 Incoming traffic is received by the current slave.
636 If the receiving slave fails, another slave takes over
637 the MAC address of the failed receiving slave.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
639 Prerequisite:
640
641 Ethtool support in the base drivers for retrieving the
642 speed of each slave.
643
644 balance-alb or 6
645
646 Adaptive load balancing: includes balance-tlb plus
647 receive load balancing (rlb) for IPV4 traffic, and
648 does not require any special switch support. The
649 receive load balancing is achieved by ARP negotiation.
650 The bonding driver intercepts the ARP Replies sent by
651 the local system on their way out and overwrites the
652 source hardware address with the unique hardware
653 address of one of the slaves in the bond such that
654 different peers use different hardware addresses for
655 the server.
656
657 Receive traffic from connections created by the server
658 is also balanced. When the local system sends an ARP
659 Request the bonding driver copies and saves the peer's
660 IP information from the ARP packet. When the ARP
661 Reply arrives from the peer, its hardware address is
662 retrieved and the bonding driver initiates an ARP
663 reply to this peer assigning it to one of the slaves
664 in the bond. A problematic outcome of using ARP
665 negotiation for balancing is that each time that an
666 ARP request is broadcast it uses the hardware address
667 of the bond. Hence, peers learn the hardware address
668 of the bond and the balancing of receive traffic
669 collapses to the current slave. This is handled by
670 sending updates (ARP Replies) to all the peers with
671 their individually assigned hardware address such that
672 the traffic is redistributed. Receive traffic is also
673 redistributed when a new slave is added to the bond
674 and when an inactive slave is re-activated. The
675 receive load is distributed sequentially (round robin)
676 among the group of highest speed slaves in the bond.
677
678 When a link is reconnected or a new slave joins the
679 bond the receive traffic is redistributed among all
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700680 active slaves in the bond by initiating ARP Replies
Auke Kok6224e012006-06-08 11:15:35 -0700681 with the selected MAC address to each of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 clients. The updelay parameter (detailed below) must
683 be set to a value equal or greater than the switch's
684 forwarding delay so that the ARP Replies sent to the
685 peers will not be blocked by the switch.
686
687 Prerequisites:
688
689 1. Ethtool support in the base drivers for retrieving
690 the speed of each slave.
691
692 2. Base driver support for setting the hardware
693 address of a device while it is open. This is
694 required so that there will always be one slave in the
695 team using the bond hardware address (the
696 curr_active_slave) while having a unique hardware
697 address for each slave in the bond. If the
698 curr_active_slave fails its hardware address is
699 swapped with the new curr_active_slave that was
700 chosen.
701
Jay Vosburghb59f9f72008-06-13 18:12:03 -0700702num_grat_arp
Brian Haley305d5522008-11-04 17:51:14 -0800703num_unsol_na
704
Ben Hutchingsad246c992011-04-26 15:25:52 +0000705 Specify the number of peer notifications (gratuitous ARPs and
706 unsolicited IPv6 Neighbor Advertisements) to be issued after a
707 failover event. As soon as the link is up on the new slave
708 (possibly immediately) a peer notification is sent on the
Vincent Bernat0307d582019-07-13 16:35:27 +0200709 bonding device and each VLAN sub-device. This is repeated at
710 the rate specified by peer_notif_delay if the number is
711 greater than 1.
Brian Haley305d5522008-11-04 17:51:14 -0800712
Ben Hutchingsad246c992011-04-26 15:25:52 +0000713 The valid range is 0 - 255; the default value is 1. These options
714 affect only the active-backup mode. These options were added for
715 bonding versions 3.3.0 and 3.4.0 respectively.
716
Jesper Juhl8fb4e132011-08-01 17:59:44 -0700717 From Linux 3.0 and bonding version 3.7.1, these notifications
Ben Hutchingsad246c992011-04-26 15:25:52 +0000718 are generated by the ipv4 and ipv6 code and the numbers of
719 repetitions cannot be set independently.
Brian Haley305d5522008-11-04 17:51:14 -0800720
Nikolay Aleksandrov12465fb2013-11-05 13:51:42 +0100721packets_per_slave
722
723 Specify the number of packets to transmit through a slave before
724 moving to the next one. When set to 0 then a slave is chosen at
725 random.
726
727 The valid range is 0 - 65535; the default value is 1. This option
728 has effect only in balance-rr mode.
729
Vincent Bernat0307d582019-07-13 16:35:27 +0200730peer_notif_delay
731
732 Specify the delay, in milliseconds, between each peer
733 notification (gratuitous ARP and unsolicited IPv6 Neighbor
734 Advertisement) when they are issued after a failover event.
735 This delay should be a multiple of the link monitor interval
736 (arp_interval or miimon, whichever is active). The default
737 value is 0 which means to match the value of the link monitor
738 interval.
739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740primary
741
742 A string (eth0, eth2, etc) specifying which slave is the
743 primary device. The specified device will always be the
744 active slave while it is available. Only when the primary is
745 off-line will alternate devices be used. This is useful when
746 one slave is preferred over another, e.g., when one slave has
747 higher throughput than another.
748
dingtianhonge1d206a2014-01-18 16:28:57 +0800749 The primary option is only valid for active-backup(1),
750 balance-tlb (5) and balance-alb (6) mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
Jiri Pirkoa5499522009-09-25 03:28:09 +0000752primary_reselect
753
754 Specifies the reselection policy for the primary slave. This
755 affects how the primary slave is chosen to become the active slave
756 when failure of the active slave or recovery of the primary slave
757 occurs. This option is designed to prevent flip-flopping between
758 the primary slave and other slaves. Possible values are:
759
760 always or 0 (default)
761
762 The primary slave becomes the active slave whenever it
763 comes back up.
764
765 better or 1
766
767 The primary slave becomes the active slave when it comes
768 back up, if the speed and duplex of the primary slave is
769 better than the speed and duplex of the current active
770 slave.
771
772 failure or 2
773
774 The primary slave becomes the active slave only if the
775 current active slave fails and the primary slave is up.
776
777 The primary_reselect setting is ignored in two cases:
778
779 If no slaves are active, the first slave to recover is
780 made the active slave.
781
782 When initially enslaved, the primary slave is always made
783 the active slave.
784
785 Changing the primary_reselect policy via sysfs will cause an
786 immediate selection of the best active slave according to the new
787 policy. This may or may not result in a change of the active
788 slave, depending upon the circumstances.
789
790 This option was added for bonding version 3.6.0.
791
Mahesh Bandeware9f0fb82014-04-22 16:30:22 -0700792tlb_dynamic_lb
793
794 Specifies if dynamic shuffling of flows is enabled in tlb
795 mode. The value has no effect on any other modes.
796
797 The default behavior of tlb mode is to shuffle active flows across
798 slaves based on the load in that interval. This gives nice lb
799 characteristics but can cause packet reordering. If re-ordering is
800 a concern use this variable to disable flow shuffling and rely on
801 load balancing provided solely by the hash distribution.
802 xmit-hash-policy can be used to select the appropriate hashing for
803 the setup.
804
805 The sysfs entry can be used to change the setting per bond device
806 and the initial value is derived from the module parameter. The
807 sysfs entry is allowed to be changed only if the bond device is
808 down.
809
810 The default value is "1" that enables flow shuffling while value "0"
811 disables it. This option was added in bonding driver 3.7.1
812
813
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814updelay
815
816 Specifies the time, in milliseconds, to wait before enabling a
817 slave after a link recovery has been detected. This option is
818 only valid for the miimon link monitor. The updelay value
819 should be a multiple of the miimon value; if not, it will be
820 rounded down to the nearest multiple. The default value is 0.
821
822use_carrier
823
824 Specifies whether or not miimon should use MII or ETHTOOL
825 ioctls vs. netif_carrier_ok() to determine the link
826 status. The MII or ETHTOOL ioctls are less efficient and
827 utilize a deprecated calling sequence within the kernel. The
828 netif_carrier_ok() relies on the device driver to maintain its
829 state with netif_carrier_on/off; at this writing, most, but
830 not all, device drivers support this facility.
831
832 If bonding insists that the link is up when it should not be,
833 it may be that your network device driver does not support
834 netif_carrier_on/off. The default state for netif_carrier is
835 "carrier on," so if a driver does not support netif_carrier,
836 it will appear as if the link is always up. In this case,
837 setting use_carrier to 0 will cause bonding to revert to the
838 MII / ETHTOOL ioctl method to determine the link state.
839
840 A value of 1 enables the use of netif_carrier_ok(), a value of
Debabrata Banerjeeb3c898e2018-05-16 14:02:13 -0400841 0 will use the deprecated MII / ETHTOOL ioctls. The default
842 value is 1.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700844xmit_hash_policy
845
846 Selects the transmit hash policy to use for slave selection in
Mahesh Bandewarf05b42e2014-04-22 16:30:20 -0700847 balance-xor, 802.3ad, and tlb modes. Possible values are:
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700848
849 layer2
850
Jianhua Xie92abf752014-07-17 14:16:26 +0800851 Uses XOR of hardware MAC addresses and packet type ID
852 field to generate the hash. The formula is
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700853
Jianhua Xie92abf752014-07-17 14:16:26 +0800854 hash = source MAC XOR destination MAC XOR packet type ID
855 slave number = hash modulo slave count
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700856
857 This algorithm will place all traffic to a particular
858 network peer on the same slave.
859
860 This algorithm is 802.3ad compliant.
861
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800862 layer2+3
863
864 This policy uses a combination of layer2 and layer3
865 protocol information to generate the hash.
866
867 Uses XOR of hardware MAC addresses and IP addresses to
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200868 generate the hash. The formula is
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800869
Jianhua Xie92abf752014-07-17 14:16:26 +0800870 hash = source MAC XOR destination MAC XOR packet type ID
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200871 hash = hash XOR source IP XOR destination IP
872 hash = hash XOR (hash RSHIFT 16)
873 hash = hash XOR (hash RSHIFT 8)
874 And then hash is reduced modulo slave count.
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800875
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200876 If the protocol is IPv6 then the source and destination
877 addresses are first hashed using ipv6_addr_hash.
John Eaglesham6b923cb2012-08-21 20:43:35 +0000878
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800879 This algorithm will place all traffic to a particular
880 network peer on the same slave. For non-IP traffic,
881 the formula is the same as for the layer2 transmit
882 hash policy.
883
884 This policy is intended to provide a more balanced
885 distribution of traffic than layer2 alone, especially
886 in environments where a layer3 gateway device is
887 required to reach most destinations.
888
Matt LaPlanted9195882008-07-25 19:45:33 -0700889 This algorithm is 802.3ad compliant.
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800890
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700891 layer3+4
892
893 This policy uses upper layer protocol information,
894 when available, to generate the hash. This allows for
895 traffic to a particular network peer to span multiple
896 slaves, although a single connection will not span
897 multiple slaves.
898
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200899 The formula for unfragmented TCP and UDP packets is
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700900
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200901 hash = source port, destination port (as in the header)
902 hash = hash XOR source IP XOR destination IP
903 hash = hash XOR (hash RSHIFT 16)
904 hash = hash XOR (hash RSHIFT 8)
905 And then hash is reduced modulo slave count.
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700906
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200907 If the protocol is IPv6 then the source and destination
908 addresses are first hashed using ipv6_addr_hash.
John Eaglesham6b923cb2012-08-21 20:43:35 +0000909
910 For fragmented TCP or UDP packets and all other IPv4 and
911 IPv6 protocol traffic, the source and destination port
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700912 information is omitted. For non-IP traffic, the
913 formula is the same as for the layer2 transmit hash
914 policy.
915
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700916 This algorithm is not fully 802.3ad compliant. A
917 single TCP or UDP conversation containing both
918 fragmented and unfragmented packets will see packets
919 striped across two interfaces. This may result in out
920 of order delivery. Most traffic types will not meet
921 this criteria, as TCP rarely fragments traffic, and
922 most UDP traffic is not involved in extended
923 conversations. Other implementations of 802.3ad may
924 or may not tolerate this noncompliance.
925
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200926 encap2+3
927
928 This policy uses the same formula as layer2+3 but it
929 relies on skb_flow_dissect to obtain the header fields
930 which might result in the use of inner headers if an
931 encapsulation protocol is used. For example this will
932 improve the performance for tunnel users because the
933 packets will be distributed according to the encapsulated
934 flows.
935
936 encap3+4
937
938 This policy uses the same formula as layer3+4 but it
939 relies on skb_flow_dissect to obtain the header fields
940 which might result in the use of inner headers if an
941 encapsulation protocol is used. For example this will
942 improve the performance for tunnel users because the
943 packets will be distributed according to the encapsulated
944 flows.
945
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700946 The default value is layer2. This option was added in bonding
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800947 version 2.6.3. In earlier versions of bonding, this parameter
948 does not exist, and the layer2 policy is the only policy. The
949 layer2+3 value was added for bonding version 3.2.2.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Flavio Leitnerc2952c32010-10-05 14:23:59 +0000951resend_igmp
952
953 Specifies the number of IGMP membership reports to be issued after
954 a failover event. One membership report is issued immediately after
955 the failover, subsequent packets are sent in each 200ms interval.
956
Flavio Leitner94265cf2011-05-25 08:38:58 +0000957 The valid range is 0 - 255; the default value is 1. A value of 0
958 prevents the IGMP membership report from being issued in response
959 to the failover event.
960
961 This option is useful for bonding modes balance-rr (0), active-backup
962 (1), balance-tlb (5) and balance-alb (6), in which a failover can
963 switch the IGMP traffic from one slave to another. Therefore a fresh
964 IGMP report must be issued to cause the switch to forward the incoming
965 IGMP traffic over the newly selected slave.
966
967 This option was added for bonding version 3.7.0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968
dingtianhong84a6a0a2013-12-21 14:40:22 +0800969lp_interval
970
971 Specifies the number of seconds between instances where the bonding
972 driver sends learning packets to each slaves peer switch.
973
974 The valid range is 1 - 0x7fffffff; the default value is 1. This Option
975 has effect only in balance-tlb and balance-alb modes.
976
Linus Torvalds1da177e2005-04-16 15:20:36 -07009773. Configuring Bonding Devices
978==============================
979
Auke Kok6224e012006-06-08 11:15:35 -0700980 You can configure bonding using either your distro's network
Cong Wangb1098bb2013-05-27 15:49:16 +0000981initialization scripts, or manually using either iproute2 or the
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000982sysfs interface. Distros generally use one of three packages for the
983network initialization scripts: initscripts, sysconfig or interfaces.
984Recent versions of these packages have support for bonding, while older
Auke Kok6224e012006-06-08 11:15:35 -0700985versions do not.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986
987 We will first describe the options for configuring bonding for
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000988distros using versions of initscripts, sysconfig and interfaces with full
989or partial support for bonding, then provide information on enabling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990bonding without support from the network initialization scripts (i.e.,
991older versions of initscripts or sysconfig).
992
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000993 If you're unsure whether your distro uses sysconfig,
994initscripts or interfaces, or don't know if it's new enough, have no fear.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995Determining this is fairly straightforward.
996
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000997 First, look for a file called interfaces in /etc/network directory.
998If this file is present in your system, then your system use interfaces. See
999Configuration with Interfaces Support.
1000
1001 Else, issue the command:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002
1003$ rpm -qf /sbin/ifup
1004
1005 It will respond with a line of text starting with either
1006"initscripts" or "sysconfig," followed by some numbers. This is the
1007package that provides your network initialization scripts.
1008
1009 Next, to determine if your installation supports bonding,
1010issue the command:
1011
1012$ grep ifenslave /sbin/ifup
1013
1014 If this returns any matches, then your initscripts or
1015sysconfig has support for bonding.
1016
Auke Kok6224e012006-06-08 11:15:35 -070010173.1 Configuration with Sysconfig Support
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018----------------------------------------
1019
1020 This section applies to distros using a version of sysconfig
1021with bonding support, for example, SuSE Linux Enterprise Server 9.
1022
1023 SuSE SLES 9's networking configuration system does support
1024bonding, however, at this writing, the YaST system configuration
Auke Kok6224e012006-06-08 11:15:35 -07001025front end does not provide any means to work with bonding devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026Bonding devices can be managed by hand, however, as follows.
1027
1028 First, if they have not already been configured, configure the
1029slave devices. On SLES 9, this is most easily done by running the
1030yast2 sysconfig configuration utility. The goal is for to create an
1031ifcfg-id file for each slave device. The simplest way to accomplish
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001032this is to configure the devices for DHCP (this is only to get the
1033file ifcfg-id file created; see below for some issues with DHCP). The
1034name of the configuration file for each device will be of the form:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035
1036ifcfg-id-xx:xx:xx:xx:xx:xx
1037
1038 Where the "xx" portion will be replaced with the digits from
1039the device's permanent MAC address.
1040
1041 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
1042created, it is necessary to edit the configuration files for the slave
1043devices (the MAC addresses correspond to those of the slave devices).
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001044Before editing, the file will contain multiple lines, and will look
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045something like this:
1046
1047BOOTPROTO='dhcp'
1048STARTMODE='on'
1049USERCTL='no'
1050UNIQUE='XNzu.WeZGOGF+4wE'
1051_nm_name='bus-pci-0001:61:01.0'
1052
1053 Change the BOOTPROTO and STARTMODE lines to the following:
1054
1055BOOTPROTO='none'
1056STARTMODE='off'
1057
1058 Do not alter the UNIQUE or _nm_name lines. Remove any other
1059lines (USERCTL, etc).
1060
1061 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
1062it's time to create the configuration file for the bonding device
1063itself. This file is named ifcfg-bondX, where X is the number of the
1064bonding device to create, starting at 0. The first such file is
1065ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
1066network configuration system will correctly start multiple instances
1067of bonding.
1068
1069 The contents of the ifcfg-bondX file is as follows:
1070
1071BOOTPROTO="static"
1072BROADCAST="10.0.2.255"
1073IPADDR="10.0.2.10"
1074NETMASK="255.255.0.0"
1075NETWORK="10.0.2.0"
1076REMOTE_IPADDR=""
1077STARTMODE="onboot"
1078BONDING_MASTER="yes"
1079BONDING_MODULE_OPTS="mode=active-backup miimon=100"
1080BONDING_SLAVE0="eth0"
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001081BONDING_SLAVE1="bus-pci-0000:06:08.1"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082
1083 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
1084values with the appropriate values for your network.
1085
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 The STARTMODE specifies when the device is brought online.
1087The possible values are:
1088
1089 onboot: The device is started at boot time. If you're not
1090 sure, this is probably what you want.
1091
1092 manual: The device is started only when ifup is called
1093 manually. Bonding devices may be configured this
1094 way if you do not wish them to start automatically
1095 at boot for some reason.
1096
1097 hotplug: The device is started by a hotplug event. This is not
1098 a valid choice for a bonding device.
1099
1100 off or ignore: The device configuration is ignored.
1101
1102 The line BONDING_MASTER='yes' indicates that the device is a
1103bonding master device. The only useful value is "yes."
1104
1105 The contents of BONDING_MODULE_OPTS are supplied to the
1106instance of the bonding module for this device. Specify the options
1107for the bonding mode, link monitoring, and so on here. Do not include
1108the max_bonds bonding parameter; this will confuse the configuration
1109system if you have multiple bonding devices.
1110
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001111 Finally, supply one BONDING_SLAVEn="slave device" for each
1112slave. where "n" is an increasing value, one for each slave. The
1113"slave device" is either an interface name, e.g., "eth0", or a device
1114specifier for the network device. The interface name is easier to
1115find, but the ethN names are subject to change at boot time if, e.g.,
1116a device early in the sequence has failed. The device specifiers
1117(bus-pci-0000:06:08.1 in the example above) specify the physical
1118network device, and will not change unless the device's bus location
1119changes (for example, it is moved from one PCI slot to another). The
1120example above uses one of each type for demonstration purposes; most
1121configurations will choose one or the other for all slave devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122
1123 When all configuration files have been modified or created,
1124networking must be restarted for the configuration changes to take
1125effect. This can be accomplished via the following:
1126
1127# /etc/init.d/network restart
1128
1129 Note that the network control script (/sbin/ifdown) will
1130remove the bonding module as part of the network shutdown processing,
1131so it is not necessary to remove the module by hand if, e.g., the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001132module parameters have changed.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
1134 Also, at this writing, YaST/YaST2 will not manage bonding
1135devices (they do not show bonding interfaces on its list of network
1136devices). It is necessary to edit the configuration file by hand to
1137change the bonding configuration.
1138
1139 Additional general options and details of the ifcfg file
1140format can be found in an example ifcfg template file:
1141
1142/etc/sysconfig/network/ifcfg.template
1143
1144 Note that the template does not document the various BONDING_
1145settings described above, but does describe many of the other options.
1146
Auke Kok6224e012006-06-08 11:15:35 -070011473.1.1 Using DHCP with Sysconfig
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001148-------------------------------
1149
1150 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1151will cause it to query DHCP for its IP address information. At this
1152writing, this does not function for bonding devices; the scripts
1153attempt to obtain the device address from DHCP prior to adding any of
1154the slave devices. Without active slaves, the DHCP requests are not
1155sent to the network.
1156
Auke Kok6224e012006-06-08 11:15:35 -070011573.1.2 Configuring Multiple Bonds with Sysconfig
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001158-----------------------------------------------
1159
1160 The sysconfig network initialization system is capable of
1161handling multiple bonding devices. All that is necessary is for each
1162bonding instance to have an appropriately configured ifcfg-bondX file
1163(as described above). Do not specify the "max_bonds" parameter to any
1164instance of bonding, as this will confuse sysconfig. If you require
1165multiple bonding devices with identical parameters, create multiple
1166ifcfg-bondX files.
1167
1168 Because the sysconfig scripts supply the bonding module
1169options in the ifcfg-bondX file, it is not necessary to add them to
Lucas De Marchi970e2482012-03-30 13:37:16 -07001170the system /etc/modules.d/*.conf configuration files.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001171
Auke Kok6224e012006-06-08 11:15:35 -070011723.2 Configuration with Initscripts Support
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173------------------------------------------
1174
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001175 This section applies to distros using a recent version of
1176initscripts with bonding support, for example, Red Hat Enterprise Linux
1177version 3 or later, Fedora, etc. On these systems, the network
1178initialization scripts have knowledge of bonding, and can be configured to
1179control bonding devices. Note that older versions of the initscripts
1180package have lower levels of support for bonding; this will be noted where
1181applicable.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182
1183 These distros will not automatically load the network adapter
1184driver unless the ethX device is configured with an IP address.
1185Because of this constraint, users must manually configure a
1186network-script file for all physical adapters that will be members of
1187a bondX link. Network script files are located in the directory:
1188
1189/etc/sysconfig/network-scripts
1190
1191 The file name must be prefixed with "ifcfg-eth" and suffixed
1192with the adapter's physical adapter number. For example, the script
1193for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1194Place the following text in the file:
1195
1196DEVICE=eth0
1197USERCTL=no
1198ONBOOT=yes
1199MASTER=bond0
1200SLAVE=yes
1201BOOTPROTO=none
1202
1203 The DEVICE= line will be different for every ethX device and
1204must correspond with the name of the file, i.e., ifcfg-eth1 must have
1205a device line of DEVICE=eth1. The setting of the MASTER= line will
1206also depend on the final bonding interface name chosen for your bond.
1207As with other network devices, these typically start at 0, and go up
1208one for each device, i.e., the first bonding instance is bond0, the
1209second is bond1, and so on.
1210
1211 Next, create a bond network script. The file name for this
1212script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1213the number of the bond. For bond0 the file is named "ifcfg-bond0",
1214for bond1 it is named "ifcfg-bond1", and so on. Within that file,
1215place the following text:
1216
1217DEVICE=bond0
1218IPADDR=192.168.1.1
1219NETMASK=255.255.255.0
1220NETWORK=192.168.1.0
1221BROADCAST=192.168.1.255
1222ONBOOT=yes
1223BOOTPROTO=none
1224USERCTL=no
1225
1226 Be sure to change the networking specific lines (IPADDR,
1227NETMASK, NETWORK and BROADCAST) to match your network configuration.
1228
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001229 For later versions of initscripts, such as that found with Fedora
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +000012307 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1231and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001232file, e.g. a line of the format:
1233
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +00001234BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001235
1236 will configure the bond with the specified options. The options
1237specified in BONDING_OPTS are identical to the bonding module parameters
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +00001238except for the arp_ip_target field when using versions of initscripts older
1239than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
1240using older versions each target should be included as a separate option and
1241should be preceded by a '+' to indicate it should be added to the list of
1242queried targets, e.g.,
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001243
1244 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1245
1246 is the proper syntax to specify multiple targets. When specifying
Lucas De Marchi970e2482012-03-30 13:37:16 -07001247options via BONDING_OPTS, it is not necessary to edit /etc/modprobe.d/*.conf.
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001248
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +00001249 For even older versions of initscripts that do not support
Lucas De Marchi970e2482012-03-30 13:37:16 -07001250BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1251your distro) to load the bonding module with your desired options when the
1252bond0 interface is brought up. The following lines in /etc/modprobe.d/*.conf
1253will load the bonding module, and select its options:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254
1255alias bond0 bonding
1256options bond0 mode=balance-alb miimon=100
1257
1258 Replace the sample parameters with the appropriate set of
1259options for your configuration.
1260
1261 Finally run "/etc/rc.d/init.d/network restart" as root. This
1262will restart the networking subsystem and your bond link should be now
1263up and running.
1264
Auke Kok6224e012006-06-08 11:15:35 -070012653.2.1 Using DHCP with Initscripts
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001266---------------------------------
1267
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001268 Recent versions of initscripts (the versions supplied with Fedora
1269Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1270work) have support for assigning IP information to bonding devices via
1271DHCP.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001272
1273 To configure bonding for DHCP, configure it as described
1274above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1275and add a line consisting of "TYPE=Bonding". Note that the TYPE value
1276is case sensitive.
1277
Auke Kok6224e012006-06-08 11:15:35 -070012783.2.2 Configuring Multiple Bonds with Initscripts
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001279-------------------------------------------------
1280
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001281 Initscripts packages that are included with Fedora 7 and Red Hat
1282Enterprise Linux 5 support multiple bonding interfaces by simply
1283specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1284number of the bond. This support requires sysfs support in the kernel,
1285and a bonding driver of version 3.0.0 or later. Other configurations may
1286not support this method for specifying multiple bonding interfaces; for
1287those instances, see the "Configuring Multiple Bonds Manually" section,
1288below.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289
Cong Wangb1098bb2013-05-27 15:49:16 +000012903.3 Configuring Bonding Manually with iproute2
Auke Kok6224e012006-06-08 11:15:35 -07001291-----------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
1293 This section applies to distros whose network initialization
1294scripts (the sysconfig or initscripts package) do not have specific
1295knowledge of bonding. One such distro is SuSE Linux Enterprise Server
1296version 8.
1297
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001298 The general method for these systems is to place the bonding
Lucas De Marchi970e2482012-03-30 13:37:16 -07001299module parameters into a config file in /etc/modprobe.d/ (as
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001300appropriate for the installed distro), then add modprobe and/or
Cong Wangb1098bb2013-05-27 15:49:16 +00001301`ip link` commands to the system's global init script. The name of
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001302the global init script differs; for sysconfig, it is
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303/etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1304
1305 For example, if you wanted to make a simple bond of two e100
1306devices (presumed to be eth0 and eth1), and have it persist across
1307reboots, edit the appropriate file (/etc/init.d/boot.local or
1308/etc/rc.d/rc.local), and add the following:
1309
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001310modprobe bonding mode=balance-alb miimon=100
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311modprobe e100
1312ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
Cong Wangb1098bb2013-05-27 15:49:16 +00001313ip link set eth0 master bond0
1314ip link set eth1 master bond0
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
1316 Replace the example bonding module parameters and bond0
1317network configuration (IP address, netmask, etc) with the appropriate
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001318values for your configuration.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
1320 Unfortunately, this method will not provide support for the
1321ifup and ifdown scripts on the bond devices. To reload the bonding
1322configuration, it is necessary to run the initialization script, e.g.,
1323
1324# /etc/init.d/boot.local
1325
1326 or
1327
1328# /etc/rc.d/rc.local
1329
1330 It may be desirable in such a case to create a separate script
1331which only initializes the bonding configuration, then call that
1332separate script from within boot.local. This allows for bonding to be
1333enabled without re-running the entire global init script.
1334
1335 To shut down the bonding devices, it is necessary to first
1336mark the bonding device itself as being down, then remove the
1337appropriate device driver modules. For our example above, you can do
1338the following:
1339
1340# ifconfig bond0 down
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001341# rmmod bonding
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342# rmmod e100
1343
1344 Again, for convenience, it may be desirable to create a script
1345with these commands.
1346
1347
Jay Vosburgh00354cf2005-07-21 12:18:02 -070013483.3.1 Configuring Multiple Bonds Manually
1349-----------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350
1351 This section contains information on configuring multiple
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001352bonding devices with differing options for those systems whose network
1353initialization scripts lack support for configuring multiple bonds.
1354
1355 If you require multiple bonding devices, but all with the same
1356options, you may wish to use the "max_bonds" module parameter,
1357documented above.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001359 To create multiple bonding devices with differing options, it is
Rick Jonesf8b72d32012-07-20 10:51:37 +00001360preferable to use bonding parameters exported by sysfs, documented in the
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001361section below.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001363 For versions of bonding without sysfs support, the only means to
1364provide multiple instances of bonding with differing options is to load
1365the bonding driver multiple times. Note that current versions of the
1366sysconfig network initialization scripts handle this automatically; if
1367your distro uses these scripts, no special action is needed. See the
1368section Configuring Bonding Devices, above, if you're not sure about your
1369network initialization scripts.
1370
1371 To load multiple instances of the module, it is necessary to
1372specify a different name for each instance (the module loading system
1373requires that every loaded module, even multiple instances of the same
1374module, have a unique name). This is accomplished by supplying multiple
Lucas De Marchi970e2482012-03-30 13:37:16 -07001375sets of bonding options in /etc/modprobe.d/*.conf, for example:
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001376
1377alias bond0 bonding
1378options bond0 -o bond0 mode=balance-rr miimon=100
1379
1380alias bond1 bonding
1381options bond1 -o bond1 mode=balance-alb miimon=50
1382
1383 will load the bonding module two times. The first instance is
1384named "bond0" and creates the bond0 device in balance-rr mode with an
1385miimon of 100. The second instance is named "bond1" and creates the
1386bond1 device in balance-alb mode with an miimon of 50.
1387
1388 In some circumstances (typically with older distributions),
1389the above does not work, and the second bonding instance never sees
1390its options. In that case, the second options line can be substituted
1391as follows:
1392
1393install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1394 mode=balance-alb miimon=50
1395
1396 This may be repeated any number of times, specifying a new and
1397unique name in place of bond1 for each subsequent instance.
1398
1399 It has been observed that some Red Hat supplied kernels are unable
1400to rename modules at load time (the "-o bond1" part). Attempts to pass
1401that option to modprobe will produce an "Operation not permitted" error.
1402This has been reported on some Fedora Core kernels, and has been seen on
1403RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1404to configure multiple bonds with differing parameters (as they are older
1405kernels, and also lack sysfs support).
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001406
Auke Kok6224e012006-06-08 11:15:35 -070014073.4 Configuring Bonding Manually via Sysfs
1408------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001410 Starting with version 3.0.0, Channel Bonding may be configured
Auke Kok6224e012006-06-08 11:15:35 -07001411via the sysfs interface. This interface allows dynamic configuration
1412of all bonds in the system without unloading the module. It also
1413allows for adding and removing bonds at runtime. Ifenslave is no
1414longer required, though it is still supported.
1415
1416 Use of the sysfs interface allows you to use multiple bonds
1417with different configurations without having to reload the module.
1418It also allows you to use multiple, differently configured bonds when
1419bonding is compiled into the kernel.
1420
1421 You must have the sysfs filesystem mounted to configure
1422bonding this way. The examples in this document assume that you
1423are using the standard mount point for sysfs, e.g. /sys. If your
1424sysfs filesystem is mounted elsewhere, you will need to adjust the
1425example paths accordingly.
1426
1427Creating and Destroying Bonds
1428-----------------------------
1429To add a new bond foo:
1430# echo +foo > /sys/class/net/bonding_masters
1431
1432To remove an existing bond bar:
1433# echo -bar > /sys/class/net/bonding_masters
1434
1435To show all existing bonds:
1436# cat /sys/class/net/bonding_masters
1437
1438NOTE: due to 4K size limitation of sysfs files, this list may be
1439truncated if you have more than a few hundred bonds. This is unlikely
1440to occur under normal operating conditions.
1441
1442Adding and Removing Slaves
1443--------------------------
1444 Interfaces may be enslaved to a bond using the file
1445/sys/class/net/<bond>/bonding/slaves. The semantics for this file
1446are the same as for the bonding_masters file.
1447
1448To enslave interface eth0 to bond bond0:
1449# ifconfig bond0 up
1450# echo +eth0 > /sys/class/net/bond0/bonding/slaves
1451
1452To free slave eth0 from bond bond0:
1453# echo -eth0 > /sys/class/net/bond0/bonding/slaves
1454
Auke Kok6224e012006-06-08 11:15:35 -07001455 When an interface is enslaved to a bond, symlinks between the
1456two are created in the sysfs filesystem. In this case, you would get
1457/sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1458/sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1459
1460 This means that you can tell quickly whether or not an
1461interface is enslaved by looking for the master symlink. Thus:
1462# echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1463will free eth0 from whatever bond it is enslaved to, regardless of
1464the name of the bond interface.
1465
1466Changing a Bond's Configuration
1467-------------------------------
1468 Each bond may be configured individually by manipulating the
1469files located in /sys/class/net/<bond name>/bonding
1470
1471 The names of these files correspond directly with the command-
Paolo Ornati670e9f32006-10-03 22:57:56 +02001472line parameters described elsewhere in this file, and, with the
Auke Kok6224e012006-06-08 11:15:35 -07001473exception of arp_ip_target, they accept the same values. To see the
1474current setting, simply cat the appropriate file.
1475
1476 A few examples will be given here; for specific usage
1477guidelines for each parameter, see the appropriate section in this
1478document.
1479
1480To configure bond0 for balance-alb mode:
1481# ifconfig bond0 down
1482# echo 6 > /sys/class/net/bond0/bonding/mode
1483 - or -
1484# echo balance-alb > /sys/class/net/bond0/bonding/mode
1485 NOTE: The bond interface must be down before the mode can be
1486changed.
1487
1488To enable MII monitoring on bond0 with a 1 second interval:
1489# echo 1000 > /sys/class/net/bond0/bonding/miimon
1490 NOTE: If ARP monitoring is enabled, it will disabled when MII
1491monitoring is enabled, and vice-versa.
1492
1493To add ARP targets:
1494# echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1495# echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
Brian Haley5a31bec2009-04-13 00:11:30 -07001496 NOTE: up to 16 target addresses may be specified.
Auke Kok6224e012006-06-08 11:15:35 -07001497
1498To remove an ARP target:
1499# echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1500
Neil Horman7eacd032013-09-13 11:05:33 -04001501To configure the interval between learning packet transmits:
1502# echo 12 > /sys/class/net/bond0/bonding/lp_interval
Masanari Iida9f80a072018-07-13 01:05:17 +09001503 NOTE: the lp_interval is the number of seconds between instances where
Neil Horman7eacd032013-09-13 11:05:33 -04001504the bonding driver sends learning packets to each slaves peer switch. The
1505default interval is 1 second.
1506
Auke Kok6224e012006-06-08 11:15:35 -07001507Example Configuration
1508---------------------
1509 We begin with the same example that is shown in section 3.3,
1510executed with sysfs, and without using ifenslave.
1511
1512 To make a simple bond of two e100 devices (presumed to be eth0
1513and eth1), and have it persist across reboots, edit the appropriate
1514file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1515following:
1516
1517modprobe bonding
1518modprobe e100
1519echo balance-alb > /sys/class/net/bond0/bonding/mode
1520ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1521echo 100 > /sys/class/net/bond0/bonding/miimon
1522echo +eth0 > /sys/class/net/bond0/bonding/slaves
1523echo +eth1 > /sys/class/net/bond0/bonding/slaves
1524
1525 To add a second bond, with two e1000 interfaces in
1526active-backup mode, using ARP monitoring, add the following lines to
1527your init script:
1528
1529modprobe e1000
1530echo +bond1 > /sys/class/net/bonding_masters
1531echo active-backup > /sys/class/net/bond1/bonding/mode
1532ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1533echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1534echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1535echo +eth2 > /sys/class/net/bond1/bonding/slaves
1536echo +eth3 > /sys/class/net/bond1/bonding/slaves
1537
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +000015383.5 Configuration with Interfaces Support
1539-----------------------------------------
1540
1541 This section applies to distros which use /etc/network/interfaces file
1542to describe network interface configuration, most notably Debian and it's
1543derivatives.
1544
1545 The ifup and ifdown commands on Debian don't support bonding out of
1546the box. The ifenslave-2.6 package should be installed to provide bonding
1547support. Once installed, this package will provide bond-* options to be used
1548into /etc/network/interfaces.
1549
1550 Note that ifenslave-2.6 package will load the bonding module and use
1551the ifenslave command when appropriate.
1552
1553Example Configurations
1554----------------------
1555
1556In /etc/network/interfaces, the following stanza will configure bond0, in
1557active-backup mode, with eth0 and eth1 as slaves.
1558
1559auto bond0
1560iface bond0 inet dhcp
1561 bond-slaves eth0 eth1
1562 bond-mode active-backup
1563 bond-miimon 100
1564 bond-primary eth0 eth1
1565
1566If the above configuration doesn't work, you might have a system using
1567upstart for system startup. This is most notably true for recent
1568Ubuntu versions. The following stanza in /etc/network/interfaces will
1569produce the same result on those systems.
1570
1571auto bond0
1572iface bond0 inet dhcp
1573 bond-slaves none
1574 bond-mode active-backup
1575 bond-miimon 100
1576
1577auto eth0
1578iface eth0 inet manual
1579 bond-master bond0
1580 bond-primary eth0 eth1
1581
1582auto eth1
1583iface eth1 inet manual
1584 bond-master bond0
1585 bond-primary eth0 eth1
1586
1587For a full list of bond-* supported options in /etc/network/interfaces and some
1588more advanced examples tailored to you particular distros, see the files in
1589/usr/share/doc/ifenslave-2.6.
1590
15913.6 Overriding Configuration for Special Cases
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00001592----------------------------------------------
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +00001593
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00001594When using the bonding driver, the physical port which transmits a frame is
1595typically selected by the bonding driver, and is not relevant to the user or
1596system administrator. The output port is simply selected using the policies of
1597the selected bonding mode. On occasion however, it is helpful to direct certain
1598classes of traffic to certain physical interfaces on output to implement
1599slightly more complex policies. For example, to reach a web server over a
1600bonded interface in which eth0 connects to a private network, while eth1
1601connects via a public network, it may be desirous to bias the bond to send said
1602traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1603can safely be sent over either interface. Such configurations may be achieved
1604using the traffic control utilities inherent in linux.
Auke Kok6224e012006-06-08 11:15:35 -07001605
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00001606By default the bonding driver is multiqueue aware and 16 queues are created
1607when the driver initializes (see Documentation/networking/multiqueue.txt
1608for details). If more or less queues are desired the module parameter
1609tx_queues can be used to change this value. There is no sysfs parameter
1610available as the allocation is done at module init time.
1611
1612The output of the file /proc/net/bonding/bondX has changed so the output Queue
1613ID is now printed for each slave:
1614
1615Bonding Mode: fault-tolerance (active-backup)
1616Primary Slave: None
1617Currently Active Slave: eth0
1618MII Status: up
1619MII Polling Interval (ms): 0
1620Up Delay (ms): 0
1621Down Delay (ms): 0
1622
1623Slave Interface: eth0
1624MII Status: up
1625Link Failure Count: 0
1626Permanent HW addr: 00:1a:a0:12:8f:cb
1627Slave queue ID: 0
1628
1629Slave Interface: eth1
1630MII Status: up
1631Link Failure Count: 0
1632Permanent HW addr: 00:1a:a0:12:8f:cc
1633Slave queue ID: 2
1634
1635The queue_id for a slave can be set using the command:
1636
1637# echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1638
1639Any interface that needs a queue_id set should set it with multiple calls
1640like the one above until proper priorities are set for all interfaces. On
1641distributions that allow configuration via initscripts, multiple 'queue_id'
1642arguments can be added to BONDING_OPTS to set all needed slave queues.
1643
1644These queue id's can be used in conjunction with the tc utility to configure
1645a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1646slave devices. For instance, say we wanted, in the above configuration to
1647force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1648device. The following commands would accomplish this:
1649
1650# tc qdisc add dev bond0 handle 1 root multiq
1651
1652# tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip dst \
1653 192.168.1.100 action skbedit queue_mapping 2
1654
1655These commands tell the kernel to attach a multiqueue queue discipline to the
1656bond0 interface and filter traffic enqueued to it, such that packets with a dst
1657ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1658This value is then passed into the driver, causing the normal output path
1659selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1660
1661Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver
1662that normal output policy selection should take place. One benefit to simply
1663leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1664driver that is now present. This awareness allows tc filters to be placed on
1665slave devices as well as bond devices and the bonding driver will simply act as
1666a pass-through for selecting output queues on the slave device rather than
1667output port selection.
1668
1669This feature first appeared in bonding driver version 3.7.0 and support for
1670output slave selection was limited to round-robin and active-backup modes.
1671
Mahesh Bandeward22a5fc2015-05-09 00:01:57 -070016723.7 Configuring LACP for 802.3ad mode in a more secure way
1673----------------------------------------------------------
1674
1675When using 802.3ad bonding mode, the Actor (host) and Partner (switch)
1676exchange LACPDUs. These LACPDUs cannot be sniffed, because they are
1677destined to link local mac addresses (which switches/bridges are not
1678supposed to forward). However, most of the values are easily predictable
1679or are simply the machine's MAC address (which is trivially known to all
1680other hosts in the same L2). This implies that other machines in the L2
1681domain can spoof LACPDU packets from other hosts to the switch and potentially
1682cause mayhem by joining (from the point of view of the switch) another
1683machine's aggregate, thus receiving a portion of that hosts incoming
1684traffic and / or spoofing traffic from that machine themselves (potentially
1685even successfully terminating some portion of flows). Though this is not
1686a likely scenario, one could avoid this possibility by simply configuring
1687few bonding parameters:
1688
1689 (a) ad_actor_system : You can set a random mac-address that can be used for
1690 these LACPDU exchanges. The value can not be either NULL or Multicast.
1691 Also it's preferable to set the local-admin bit. Following shell code
1692 generates a random mac-address as described above.
1693
1694 # sys_mac_addr=$(printf '%02x:%02x:%02x:%02x:%02x:%02x' \
1695 $(( (RANDOM & 0xFE) | 0x02 )) \
1696 $(( RANDOM & 0xFF )) \
1697 $(( RANDOM & 0xFF )) \
1698 $(( RANDOM & 0xFF )) \
1699 $(( RANDOM & 0xFF )) \
1700 $(( RANDOM & 0xFF )))
1701 # echo $sys_mac_addr > /sys/class/net/bond0/bonding/ad_actor_system
1702
1703 (b) ad_actor_sys_prio : Randomize the system priority. The default value
1704 is 65535, but system can take the value from 1 - 65535. Following shell
1705 code generates random priority and sets it.
1706
1707 # sys_prio=$(( 1 + RANDOM + RANDOM ))
1708 # echo $sys_prio > /sys/class/net/bond0/bonding/ad_actor_sys_prio
1709
1710 (c) ad_user_port_key : Use the user portion of the port-key. The default
1711 keeps this empty. These are the upper 10 bits of the port-key and value
1712 ranges from 0 - 1023. Following shell code generates these 10 bits and
1713 sets it.
1714
1715 # usr_port_key=$(( RANDOM & 0x3FF ))
1716 # echo $usr_port_key > /sys/class/net/bond0/bonding/ad_user_port_key
1717
1718
Andy Gospodarekbb1d9122010-06-02 08:40:18 +000017194 Querying Bonding Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720=================================
1721
Auke Kok6224e012006-06-08 11:15:35 -070017224.1 Bonding Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723-------------------------
1724
1725 Each bonding device has a read-only file residing in the
1726/proc/net/bonding directory. The file contents include information
1727about the bonding configuration, options and state of each slave.
1728
1729 For example, the contents of /proc/net/bonding/bond0 after the
1730driver is loaded with parameters of mode=0 and miimon=1000 is
1731generally as follows:
1732
1733 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1734 Bonding Mode: load balancing (round-robin)
1735 Currently Active Slave: eth0
1736 MII Status: up
1737 MII Polling Interval (ms): 1000
1738 Up Delay (ms): 0
1739 Down Delay (ms): 0
1740
1741 Slave Interface: eth1
1742 MII Status: up
1743 Link Failure Count: 1
1744
1745 Slave Interface: eth0
1746 MII Status: up
1747 Link Failure Count: 1
1748
1749 The precise format and contents will change depending upon the
1750bonding configuration, state, and version of the bonding driver.
1751
Auke Kok6224e012006-06-08 11:15:35 -070017524.2 Network configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753-------------------------
1754
1755 The network configuration can be inspected using the ifconfig
1756command. Bonding devices will have the MASTER flag set; Bonding slave
1757devices will have the SLAVE flag set. The ifconfig output does not
1758contain information on which slaves are associated with which masters.
1759
1760 In the example below, the bond0 interface is the master
1761(MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1762bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1763TLB and ALB that require a unique MAC address for each slave.
1764
1765# /sbin/ifconfig
1766bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1767 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1768 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1769 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1770 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1771 collisions:0 txqueuelen:0
1772
1773eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1775 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1776 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1777 collisions:0 txqueuelen:100
1778 Interrupt:10 Base address:0x1080
1779
1780eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1782 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1783 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1784 collisions:0 txqueuelen:100
1785 Interrupt:9 Base address:0x1400
1786
Auke Kok6224e012006-06-08 11:15:35 -070017875. Switch Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788=======================
1789
1790 For this section, "switch" refers to whatever system the
1791bonded devices are directly connected to (i.e., where the other end of
1792the cable plugs into). This may be an actual dedicated switch device,
1793or it may be another regular system (e.g., another computer running
1794Linux),
1795
1796 The active-backup, balance-tlb and balance-alb modes do not
1797require any specific configuration of the switch.
1798
1799 The 802.3ad mode requires that the switch have the appropriate
1800ports configured as an 802.3ad aggregation. The precise method used
1801to configure this varies from switch to switch, but, for example, a
1802Cisco 3550 series switch requires that the appropriate ports first be
1803grouped together in a single etherchannel instance, then that
1804etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1805standard EtherChannel).
1806
1807 The balance-rr, balance-xor and broadcast modes generally
1808require that the switch have the appropriate ports grouped together.
1809The nomenclature for such a group differs between switches, it may be
1810called an "etherchannel" (as in the Cisco example, above), a "trunk
1811group" or some other similar variation. For these modes, each switch
1812will also have its own configuration options for the switch's transmit
1813policy to the bond. Typical choices include XOR of either the MAC or
1814IP addresses. The transmit policy of the two peers does not need to
1815match. For these three modes, the bonding mode really selects a
1816transmit policy for an EtherChannel group; all three will interoperate
1817with another EtherChannel group.
1818
1819
Auke Kok6224e012006-06-08 11:15:35 -070018206. 802.1q VLAN Support
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821======================
1822
1823 It is possible to configure VLAN devices over a bond interface
1824using the 8021q driver. However, only packets coming from the 8021q
1825driver and passing through bonding will be tagged by default. Self
1826generated packets, for example, bonding's learning packets or ARP
1827packets generated by either ALB mode or the ARP monitor mechanism, are
1828tagged internally by bonding itself. As a result, bonding must
1829"learn" the VLAN IDs configured above it, and use those IDs to tag
1830self generated packets.
1831
1832 For reasons of simplicity, and to support the use of adapters
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001833that can do VLAN hardware acceleration offloading, the bonding
1834interface declares itself as fully hardware offloading capable, it gets
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835the add_vid/kill_vid notifications to gather the necessary
1836information, and it propagates those actions to the slaves. In case
1837of mixed adapter types, hardware accelerated tagged packets that
1838should go through an adapter that is not offloading capable are
1839"un-accelerated" by the bonding driver so the VLAN tag sits in the
1840regular location.
1841
1842 VLAN interfaces *must* be added on top of a bonding interface
1843only after enslaving at least one slave. The bonding interface has a
1844hardware address of 00:00:00:00:00:00 until the first slave is added.
1845If the VLAN interface is created prior to the first enslavement, it
1846would pick up the all-zeroes hardware address. Once the first slave
1847is attached to the bond, the bond device itself will pick up the
1848slave's hardware address, which is then available for the VLAN device.
1849
1850 Also, be aware that a similar problem can occur if all slaves
1851are released from a bond that still has one or more VLAN interfaces on
1852top of it. When a new slave is added, the bonding interface will
1853obtain its hardware address from the first slave, which might not
1854match the hardware address of the VLAN interfaces (which was
1855ultimately copied from an earlier slave).
1856
1857 There are two methods to insure that the VLAN device operates
1858with the correct hardware address if all slaves are removed from a
1859bond interface:
1860
1861 1. Remove all VLAN interfaces then recreate them
1862
1863 2. Set the bonding interface's hardware address so that it
1864matches the hardware address of the VLAN interfaces.
1865
1866 Note that changing a VLAN interface's HW address would set the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001867underlying device -- i.e. the bonding interface -- to promiscuous
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868mode, which might not be what you want.
1869
1870
Auke Kok6224e012006-06-08 11:15:35 -070018717. Link Monitoring
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872==================
1873
1874 The bonding driver at present supports two schemes for
1875monitoring a slave device's link state: the ARP monitor and the MII
1876monitor.
1877
1878 At the present time, due to implementation restrictions in the
1879bonding driver itself, it is not possible to enable both ARP and MII
1880monitoring simultaneously.
1881
Auke Kok6224e012006-06-08 11:15:35 -070018827.1 ARP Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883-------------------------
1884
1885 The ARP monitor operates as its name suggests: it sends ARP
1886queries to one or more designated peer systems on the network, and
1887uses the response as an indication that the link is operating. This
1888gives some assurance that traffic is actually flowing to and from one
1889or more peers on the local network.
1890
1891 The ARP monitor relies on the device driver itself to verify
1892that traffic is flowing. In particular, the driver must keep up to
Florian Westphal5c2a9642016-05-04 22:51:47 +02001893date the last receive time, dev->last_rx. Drivers that use NETIF_F_LLTX
1894flag must also update netdev_queue->trans_start. If they do not, then the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895ARP monitor will immediately fail any slaves using that driver, and
1896those slaves will stay down. If networking monitoring (tcpdump, etc)
1897shows the ARP requests and replies on the network, then it may be that
1898your device driver is not updating last_rx and trans_start.
1899
Auke Kok6224e012006-06-08 11:15:35 -070019007.2 Configuring Multiple ARP Targets
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901------------------------------------
1902
1903 While ARP monitoring can be done with just one target, it can
1904be useful in a High Availability setup to have several targets to
1905monitor. In the case of just one target, the target itself may go
1906down or have a problem making it unresponsive to ARP requests. Having
1907an additional target (or several) increases the reliability of the ARP
1908monitoring.
1909
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001910 Multiple ARP targets must be separated by commas as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
1912# example options for ARP monitoring with three targets
1913alias bond0 bonding
1914options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1915
1916 For just a single target the options would resemble:
1917
1918# example options for ARP monitoring with one target
1919alias bond0 bonding
1920options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1921
1922
Auke Kok6224e012006-06-08 11:15:35 -070019237.3 MII Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924-------------------------
1925
1926 The MII monitor monitors only the carrier state of the local
1927network interface. It accomplishes this in one of three ways: by
1928depending upon the device driver to maintain its carrier state, by
1929querying the device's MII registers, or by making an ethtool query to
1930the device.
1931
1932 If the use_carrier module parameter is 1 (the default value),
1933then the MII monitor will rely on the driver for carrier state
1934information (via the netif_carrier subsystem). As explained in the
1935use_carrier parameter information, above, if the MII monitor fails to
1936detect carrier loss on the device (e.g., when the cable is physically
1937disconnected), it may be that the driver does not support
1938netif_carrier.
1939
1940 If use_carrier is 0, then the MII monitor will first query the
1941device's (via ioctl) MII registers and check the link state. If that
1942request fails (not just that it returns carrier down), then the MII
1943monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1944the same information. If both methods fail (i.e., the driver either
1945does not support or had some error in processing both the MII register
1946and ethtool requests), then the MII monitor will assume the link is
1947up.
1948
Auke Kok6224e012006-06-08 11:15:35 -070019498. Potential Sources of Trouble
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950===============================
1951
Auke Kok6224e012006-06-08 11:15:35 -070019528.1 Adventures in Routing
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953-------------------------
1954
1955 When bonding is configured, it is important that the slave
Auke Kok6224e012006-06-08 11:15:35 -07001956devices not have routes that supersede routes of the master (or,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957generally, not have routes at all). For example, suppose the bonding
1958device bond0 has two slaves, eth0 and eth1, and the routing table is
1959as follows:
1960
1961Kernel IP routing table
1962Destination Gateway Genmask Flags MSS Window irtt Iface
196310.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
196410.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
196510.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
1966127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
1967
1968 This routing configuration will likely still update the
1969receive/transmit times in the driver (needed by the ARP monitor), but
1970may bypass the bonding driver (because outgoing traffic to, in this
1971case, another host on network 10 would use eth0 or eth1 before bond0).
1972
1973 The ARP monitor (and ARP itself) may become confused by this
1974configuration, because ARP requests (generated by the ARP monitor)
1975will be sent on one interface (bond0), but the corresponding reply
1976will arrive on a different interface (eth0). This reply looks to ARP
1977as an unsolicited ARP reply (because ARP matches replies on an
1978interface basis), and is discarded. The MII monitor is not affected
1979by the state of the routing table.
1980
1981 The solution here is simply to insure that slaves do not have
1982routes of their own, and if for some reason they must, those routes do
Auke Kok6224e012006-06-08 11:15:35 -07001983not supersede routes of their master. This should generally be the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984case, but unusual configurations or errant manual or automatic static
1985route additions may cause trouble.
1986
Auke Kok6224e012006-06-08 11:15:35 -070019878.2 Ethernet Device Renaming
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988----------------------------
1989
1990 On systems with network configuration scripts that do not
1991associate physical devices directly with network interface names (so
1992that the same physical device always has the same "ethX" name), it may
Lucas De Marchi970e2482012-03-30 13:37:16 -07001993be necessary to add some special logic to config files in
1994/etc/modprobe.d/.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995
1996 For example, given a modules.conf containing the following:
1997
1998alias bond0 bonding
1999options bond0 mode=some-mode miimon=50
2000alias eth0 tg3
2001alias eth1 tg3
2002alias eth2 e1000
2003alias eth3 e1000
2004
2005 If neither eth0 and eth1 are slaves to bond0, then when the
2006bond0 interface comes up, the devices may end up reordered. This
2007happens because bonding is loaded first, then its slave device's
2008drivers are loaded next. Since no other drivers have been loaded,
2009when the e1000 driver loads, it will receive eth0 and eth1 for its
2010devices, but the bonding configuration tries to enslave eth2 and eth3
2011(which may later be assigned to the tg3 devices).
2012
2013 Adding the following:
2014
2015add above bonding e1000 tg3
2016
2017 causes modprobe to load e1000 then tg3, in that order, when
2018bonding is loaded. This command is fully documented in the
2019modules.conf manual page.
2020
Lucas De Marchi970e2482012-03-30 13:37:16 -07002021 On systems utilizing modprobe an equivalent problem can occur.
2022In this case, the following can be added to config files in
2023/etc/modprobe.d/ as:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024
Lucas De Marchi78286cd2012-03-30 13:37:20 -07002025softdep bonding pre: tg3 e1000
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
Lucas De Marchi970e2482012-03-30 13:37:16 -07002027 This will load tg3 and e1000 modules before loading the bonding one.
2028Full documentation on this can be found in the modprobe.d and modprobe
2029manual pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030
Auke Kok6224e012006-06-08 11:15:35 -070020318.3. Painfully Slow Or No Failed Link Detection By Miimon
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032---------------------------------------------------------
2033
2034 By default, bonding enables the use_carrier option, which
2035instructs bonding to trust the driver to maintain carrier state.
2036
2037 As discussed in the options section, above, some drivers do
2038not support the netif_carrier_on/_off link state tracking system.
2039With use_carrier enabled, bonding will always see these links as up,
2040regardless of their actual state.
2041
2042 Additionally, other drivers do support netif_carrier, but do
2043not maintain it in real time, e.g., only polling the link state at
2044some fixed interval. In this case, miimon will detect failures, but
2045only after some long period of time has expired. If it appears that
2046miimon is very slow in detecting link failures, try specifying
2047use_carrier=0 to see if that improves the failure detection time. If
2048it does, then it may be that the driver checks the carrier state at a
2049fixed interval, but does not cache the MII register values (so the
2050use_carrier=0 method of querying the registers directly works). If
2051use_carrier=0 does not improve the failover, then the driver may cache
2052the registers, or the problem may be elsewhere.
2053
2054 Also, remember that miimon only checks for the device's
2055carrier state. It has no way to determine the state of devices on or
2056beyond other ports of a switch, or if a switch is refusing to pass
2057traffic while still maintaining carrier on.
2058
Auke Kok6224e012006-06-08 11:15:35 -070020599. SNMP agents
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060===============
2061
2062 If running SNMP agents, the bonding driver should be loaded
2063before any network drivers participating in a bond. This requirement
Tobias Klauserd533f672005-09-10 00:26:46 -07002064is due to the interface index (ipAdEntIfIndex) being associated to
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065the first interface found with a given IP address. That is, there is
2066only one ipAdEntIfIndex for each IP address. For example, if eth0 and
2067eth1 are slaves of bond0 and the driver for eth0 is loaded before the
2068bonding driver, the interface for the IP address will be associated
2069with the eth0 interface. This configuration is shown below, the IP
2070address 192.168.1.1 has an interface index of 2 which indexes to eth0
2071in the ifDescr table (ifDescr.2).
2072
2073 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2074 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
2075 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
2076 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
2077 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
2078 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
2079 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
2080 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2081 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
2082 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2083
2084 This problem is avoided by loading the bonding driver before
2085any network drivers participating in a bond. Below is an example of
2086loading the bonding driver first, the IP address 192.168.1.1 is
2087correctly associated with ifDescr.2.
2088
2089 interfaces.ifTable.ifEntry.ifDescr.1 = lo
2090 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
2091 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
2092 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
2093 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
2094 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
2095 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
2096 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2097 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
2098 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2099
2100 While some distributions may not report the interface name in
2101ifDescr, the association between the IP address and IfIndex remains
2102and SNMP functions such as Interface_Scan_Next will report that
2103association.
2104
Auke Kok6224e012006-06-08 11:15:35 -0700210510. Promiscuous mode
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106====================
2107
2108 When running network monitoring tools, e.g., tcpdump, it is
2109common to enable promiscuous mode on the device, so that all traffic
2110is seen (instead of seeing only traffic destined for the local host).
2111The bonding driver handles promiscuous mode changes to the bonding
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002112master device (e.g., bond0), and propagates the setting to the slave
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113devices.
2114
2115 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002116the promiscuous mode setting is propagated to all slaves.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117
2118 For the active-backup, balance-tlb and balance-alb modes, the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002119promiscuous mode setting is propagated only to the active slave.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120
2121 For balance-tlb mode, the active slave is the slave currently
2122receiving inbound traffic.
2123
2124 For balance-alb mode, the active slave is the slave used as a
2125"primary." This slave is used for mode-specific control traffic, for
2126sending to peers that are unassigned or if the load is unbalanced.
2127
2128 For the active-backup, balance-tlb and balance-alb modes, when
2129the active slave changes (e.g., due to a link failure), the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002130promiscuous setting will be propagated to the new active slave.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131
Auke Kok6224e012006-06-08 11:15:35 -0700213211. Configuring Bonding for High Availability
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002133=============================================
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134
2135 High Availability refers to configurations that provide
2136maximum network availability by having redundant or backup devices,
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002137links or switches between the host and the rest of the world. The
2138goal is to provide the maximum availability of network connectivity
2139(i.e., the network always works), even though other configurations
2140could provide higher throughput.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141
Auke Kok6224e012006-06-08 11:15:35 -0700214211.1 High Availability in a Single Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143--------------------------------------------------
2144
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002145 If two hosts (or a host and a single switch) are directly
2146connected via multiple physical links, then there is no availability
2147penalty to optimizing for maximum bandwidth. In this case, there is
2148only one switch (or peer), so if it fails, there is no alternative
2149access to fail over to. Additionally, the bonding load balance modes
2150support link monitoring of their members, so if individual links fail,
2151the load will be rebalanced across the remaining devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152
Rick Jonesf8b72d32012-07-20 10:51:37 +00002153 See Section 12, "Configuring Bonding for Maximum Throughput"
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002154for information on configuring bonding with one peer device.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155
Auke Kok6224e012006-06-08 11:15:35 -0700215611.2 High Availability in a Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157----------------------------------------------------
2158
2159 With multiple switches, the configuration of bonding and the
2160network changes dramatically. In multiple switch topologies, there is
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002161a trade off between network availability and usable bandwidth.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162
2163 Below is a sample network, configured to maximize the
2164availability of the network:
2165
2166 | |
2167 |port3 port3|
2168 +-----+----+ +-----+----+
2169 | |port2 ISL port2| |
2170 | switch A +--------------------------+ switch B |
2171 | | | |
2172 +-----+----+ +-----++---+
2173 |port1 port1|
2174 | +-------+ |
2175 +-------------+ host1 +---------------+
2176 eth0 +-------+ eth1
2177
2178 In this configuration, there is a link between the two
2179switches (ISL, or inter switch link), and multiple ports connecting to
2180the outside world ("port3" on each switch). There is no technical
2181reason that this could not be extended to a third switch.
2182
Auke Kok6224e012006-06-08 11:15:35 -0700218311.2.1 HA Bonding Mode Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002184-------------------------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002186 In a topology such as the example above, the active-backup and
2187broadcast modes are the only useful bonding modes when optimizing for
2188availability; the other modes require all links to terminate on the
2189same peer for them to behave rationally.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190
2191active-backup: This is generally the preferred mode, particularly if
2192 the switches have an ISL and play together well. If the
2193 network configuration is such that one switch is specifically
2194 a backup switch (e.g., has lower capacity, higher cost, etc),
2195 then the primary option can be used to insure that the
2196 preferred link is always used when it is available.
2197
2198broadcast: This mode is really a special purpose mode, and is suitable
2199 only for very specific needs. For example, if the two
2200 switches are not connected (no ISL), and the networks beyond
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002201 them are totally independent. In this case, if it is
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202 necessary for some specific one-way traffic to reach both
2203 independent networks, then the broadcast mode may be suitable.
2204
Auke Kok6224e012006-06-08 11:15:35 -0700220511.2.2 HA Link Monitoring Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002206----------------------------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207
2208 The choice of link monitoring ultimately depends upon your
2209switch. If the switch can reliably fail ports in response to other
2210failures, then either the MII or ARP monitors should work. For
2211example, in the above example, if the "port3" link fails at the remote
2212end, the MII monitor has no direct means to detect this. The ARP
2213monitor could be configured with a target at the remote end of port3,
2214thus detecting that failure without switch support.
2215
2216 In general, however, in a multiple switch topology, the ARP
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002217monitor can provide a higher level of reliability in detecting end to
2218end connectivity failures (which may be caused by the failure of any
2219individual component to pass traffic for any reason). Additionally,
2220the ARP monitor should be configured with multiple targets (at least
2221one for each switch in the network). This will insure that,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222regardless of which switch is active, the ARP monitor has a suitable
2223target to query.
2224
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002225 Note, also, that of late many switches now support a functionality
2226generally referred to as "trunk failover." This is a feature of the
2227switch that causes the link state of a particular switch port to be set
2228down (or up) when the state of another switch port goes down (or up).
Matt LaPlante19f59462009-04-27 15:06:31 +02002229Its purpose is to propagate link failures from logically "exterior" ports
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002230to the logically "interior" ports that bonding is able to monitor via
2231miimon. Availability and configuration for trunk failover varies by
2232switch, but this can be a viable alternative to the ARP monitor when using
2233suitable switches.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234
Auke Kok6224e012006-06-08 11:15:35 -0700223512. Configuring Bonding for Maximum Throughput
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002236==============================================
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237
Auke Kok6224e012006-06-08 11:15:35 -0700223812.1 Maximizing Throughput in a Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002239------------------------------------------------------
2240
2241 In a single switch configuration, the best method to maximize
2242throughput depends upon the application and network environment. The
2243various load balancing modes each have strengths and weaknesses in
2244different environments, as detailed below.
2245
2246 For this discussion, we will break down the topologies into
2247two categories. Depending upon the destination of most traffic, we
2248categorize them into either "gatewayed" or "local" configurations.
2249
2250 In a gatewayed configuration, the "switch" is acting primarily
2251as a router, and the majority of traffic passes through this router to
2252other networks. An example would be the following:
2253
2254
2255 +----------+ +----------+
2256 | |eth0 port1| | to other networks
2257 | Host A +---------------------+ router +------------------->
2258 | +---------------------+ | Hosts B and C are out
2259 | |eth1 port2| | here somewhere
2260 +----------+ +----------+
2261
2262 The router may be a dedicated router device, or another host
2263acting as a gateway. For our discussion, the important point is that
2264the majority of traffic from Host A will pass through the router to
2265some other network before reaching its final destination.
2266
2267 In a gatewayed network configuration, although Host A may
2268communicate with many other systems, all of its traffic will be sent
2269and received via one other peer on the local network, the router.
2270
2271 Note that the case of two systems connected directly via
2272multiple physical links is, for purposes of configuring bonding, the
2273same as a gatewayed configuration. In that case, it happens that all
2274traffic is destined for the "gateway" itself, not some other network
2275beyond the gateway.
2276
2277 In a local configuration, the "switch" is acting primarily as
2278a switch, and the majority of traffic passes through this switch to
2279reach other stations on the same network. An example would be the
2280following:
2281
2282 +----------+ +----------+ +--------+
2283 | |eth0 port1| +-------+ Host B |
2284 | Host A +------------+ switch |port3 +--------+
2285 | +------------+ | +--------+
2286 | |eth1 port2| +------------------+ Host C |
2287 +----------+ +----------+port4 +--------+
2288
2289
2290 Again, the switch may be a dedicated switch device, or another
2291host acting as a gateway. For our discussion, the important point is
2292that the majority of traffic from Host A is destined for other hosts
2293on the same local network (Hosts B and C in the above example).
2294
2295 In summary, in a gatewayed configuration, traffic to and from
2296the bonded device will be to the same MAC level peer on the network
2297(the gateway itself, i.e., the router), regardless of its final
2298destination. In a local configuration, traffic flows directly to and
2299from the final destinations, thus, each destination (Host B, Host C)
2300will be addressed directly by their individual MAC addresses.
2301
2302 This distinction between a gatewayed and a local network
2303configuration is important because many of the load balancing modes
2304available use the MAC addresses of the local network source and
2305destination to make load balancing decisions. The behavior of each
2306mode is described below.
2307
2308
Auke Kok6224e012006-06-08 11:15:35 -0700230912.1.1 MT Bonding Mode Selection for Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002310-----------------------------------------------------------
2311
2312 This configuration is the easiest to set up and to understand,
2313although you will have to decide which bonding mode best suits your
2314needs. The trade offs for each mode are detailed below:
2315
2316balance-rr: This mode is the only mode that will permit a single
2317 TCP/IP connection to stripe traffic across multiple
2318 interfaces. It is therefore the only mode that will allow a
2319 single TCP/IP stream to utilize more than one interface's
2320 worth of throughput. This comes at a cost, however: the
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002321 striping generally results in peer systems receiving packets out
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002322 of order, causing TCP/IP's congestion control system to kick
2323 in, often by retransmitting segments.
2324
2325 It is possible to adjust TCP/IP's congestion limits by
2326 altering the net.ipv4.tcp_reordering sysctl parameter. The
Eric Dumazetdca145f2014-10-27 21:45:24 -07002327 usual default value is 3. But keep in mind TCP stack is able
2328 to automatically increase this when it detects reorders.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002329
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002330 Note that the fraction of packets that will be delivered out of
2331 order is highly variable, and is unlikely to be zero. The level
2332 of reordering depends upon a variety of factors, including the
2333 networking interfaces, the switch, and the topology of the
2334 configuration. Speaking in general terms, higher speed network
2335 cards produce more reordering (due to factors such as packet
2336 coalescing), and a "many to many" topology will reorder at a
2337 higher rate than a "many slow to one fast" configuration.
2338
2339 Many switches do not support any modes that stripe traffic
2340 (instead choosing a port based upon IP or MAC level addresses);
2341 for those devices, traffic for a particular connection flowing
2342 through the switch to a balance-rr bond will not utilize greater
2343 than one interface's worth of bandwidth.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002344
2345 If you are utilizing protocols other than TCP/IP, UDP for
2346 example, and your application can tolerate out of order
2347 delivery, then this mode can allow for single stream datagram
2348 performance that scales near linearly as interfaces are added
2349 to the bond.
2350
2351 This mode requires the switch to have the appropriate ports
2352 configured for "etherchannel" or "trunking."
2353
2354active-backup: There is not much advantage in this network topology to
2355 the active-backup mode, as the inactive backup devices are all
2356 connected to the same peer as the primary. In this case, a
2357 load balancing mode (with link monitoring) will provide the
2358 same level of network availability, but with increased
2359 available bandwidth. On the plus side, active-backup mode
2360 does not require any configuration of the switch, so it may
2361 have value if the hardware available does not support any of
2362 the load balance modes.
2363
2364balance-xor: This mode will limit traffic such that packets destined
2365 for specific peers will always be sent over the same
2366 interface. Since the destination is determined by the MAC
2367 addresses involved, this mode works best in a "local" network
2368 configuration (as described above), with destinations all on
2369 the same local network. This mode is likely to be suboptimal
2370 if all your traffic is passed through a single router (i.e., a
2371 "gatewayed" network configuration, as described above).
2372
2373 As with balance-rr, the switch ports need to be configured for
2374 "etherchannel" or "trunking."
2375
2376broadcast: Like active-backup, there is not much advantage to this
2377 mode in this type of network topology.
2378
2379802.3ad: This mode can be a good choice for this type of network
2380 topology. The 802.3ad mode is an IEEE standard, so all peers
2381 that implement 802.3ad should interoperate well. The 802.3ad
2382 protocol includes automatic configuration of the aggregates,
2383 so minimal manual configuration of the switch is needed
2384 (typically only to designate that some set of devices is
2385 available for 802.3ad). The 802.3ad standard also mandates
2386 that frames be delivered in order (within certain limits), so
2387 in general single connections will not see misordering of
2388 packets. The 802.3ad mode does have some drawbacks: the
2389 standard mandates that all devices in the aggregate operate at
2390 the same speed and duplex. Also, as with all bonding load
2391 balance modes other than balance-rr, no single connection will
2392 be able to utilize more than a single interface's worth of
2393 bandwidth.
2394
2395 Additionally, the linux bonding 802.3ad implementation
Jianhua Xie92abf752014-07-17 14:16:26 +08002396 distributes traffic by peer (using an XOR of MAC addresses
2397 and packet type ID), so in a "gatewayed" configuration, all
2398 outgoing traffic will generally use the same device. Incoming
2399 traffic may also end up on a single device, but that is
Axel Beckert00a534e2017-10-05 22:00:33 +02002400 dependent upon the balancing policy of the peer's 802.3ad
Jianhua Xie92abf752014-07-17 14:16:26 +08002401 implementation. In a "local" configuration, traffic will be
2402 distributed across the devices in the bond.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002403
2404 Finally, the 802.3ad mode mandates the use of the MII monitor,
2405 therefore, the ARP monitor is not available in this mode.
2406
2407balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
2408 Since the balancing is done according to MAC address, in a
2409 "gatewayed" configuration (as described above), this mode will
2410 send all traffic across a single device. However, in a
2411 "local" network configuration, this mode balances multiple
2412 local network peers across devices in a vaguely intelligent
2413 manner (not a simple XOR as in balance-xor or 802.3ad mode),
2414 so that mathematically unlucky MAC addresses (i.e., ones that
2415 XOR to the same value) will not all "bunch up" on a single
2416 interface.
2417
2418 Unlike 802.3ad, interfaces may be of differing speeds, and no
2419 special switch configuration is required. On the down side,
2420 in this mode all incoming traffic arrives over a single
2421 interface, this mode requires certain ethtool support in the
2422 network device driver of the slave interfaces, and the ARP
2423 monitor is not available.
2424
2425balance-alb: This mode is everything that balance-tlb is, and more.
2426 It has all of the features (and restrictions) of balance-tlb,
2427 and will also balance incoming traffic from local network
2428 peers (as described in the Bonding Module Options section,
2429 above).
2430
2431 The only additional down side to this mode is that the network
2432 device driver must support changing the hardware address while
2433 the device is open.
2434
Auke Kok6224e012006-06-08 11:15:35 -0700243512.1.2 MT Link Monitoring for Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002436----------------------------------------------------
2437
2438 The choice of link monitoring may largely depend upon which
2439mode you choose to use. The more advanced load balancing modes do not
2440support the use of the ARP monitor, and are thus restricted to using
2441the MII monitor (which does not provide as high a level of end to end
2442assurance as the ARP monitor).
2443
Auke Kok6224e012006-06-08 11:15:35 -0700244412.2 Maximum Throughput in a Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002445-----------------------------------------------------
2446
2447 Multiple switches may be utilized to optimize for throughput
2448when they are configured in parallel as part of an isolated network
2449between two or more systems, for example:
2450
2451 +-----------+
2452 | Host A |
2453 +-+---+---+-+
2454 | | |
2455 +--------+ | +---------+
2456 | | |
2457 +------+---+ +-----+----+ +-----+----+
2458 | Switch A | | Switch B | | Switch C |
2459 +------+---+ +-----+----+ +-----+----+
2460 | | |
2461 +--------+ | +---------+
2462 | | |
2463 +-+---+---+-+
2464 | Host B |
2465 +-----------+
2466
2467 In this configuration, the switches are isolated from one
2468another. One reason to employ a topology such as this is for an
2469isolated network with many hosts (a cluster configured for high
2470performance, for example), using multiple smaller switches can be more
2471cost effective than a single larger switch, e.g., on a network with 24
2472hosts, three 24 port switches can be significantly less expensive than
2473a single 72 port switch.
2474
2475 If access beyond the network is required, an individual host
2476can be equipped with an additional network device connected to an
2477external network; this host then additionally acts as a gateway.
2478
Auke Kok6224e012006-06-08 11:15:35 -0700247912.2.1 MT Bonding Mode Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002480-------------------------------------------------------------
2481
2482 In actual practice, the bonding mode typically employed in
2483configurations of this type is balance-rr. Historically, in this
2484network configuration, the usual caveats about out of order packet
2485delivery are mitigated by the use of network adapters that do not do
2486any kind of packet coalescing (via the use of NAPI, or because the
2487device itself does not generate interrupts until some number of
2488packets has arrived). When employed in this fashion, the balance-rr
2489mode allows individual connections between two hosts to effectively
2490utilize greater than one interface's bandwidth.
2491
Auke Kok6224e012006-06-08 11:15:35 -0700249212.2.2 MT Link Monitoring for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002493------------------------------------------------------
2494
2495 Again, in actual practice, the MII monitor is most often used
2496in this configuration, as performance is given preference over
2497availability. The ARP monitor will function in this topology, but its
2498advantages over the MII monitor are mitigated by the volume of probes
2499needed as the number of systems involved grows (remember that each
2500host in the network is configured with bonding).
2501
Auke Kok6224e012006-06-08 11:15:35 -0700250213. Switch Behavior Issues
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002503==========================
2504
Auke Kok6224e012006-06-08 11:15:35 -0700250513.1 Link Establishment and Failover Delays
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002506-------------------------------------------
2507
2508 Some switches exhibit undesirable behavior with regard to the
2509timing of link up and down reporting by the switch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002510
2511 First, when a link comes up, some switches may indicate that
2512the link is up (carrier available), but not pass traffic over the
2513interface for some period of time. This delay is typically due to
2514some type of autonegotiation or routing protocol, but may also occur
2515during switch initialization (e.g., during recovery after a switch
2516failure). If you find this to be a problem, specify an appropriate
2517value to the updelay bonding module option to delay the use of the
2518relevant interface(s).
2519
2520 Second, some switches may "bounce" the link state one or more
2521times while a link is changing state. This occurs most commonly while
2522the switch is initializing. Again, an appropriate updelay value may
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002523help.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002524
2525 Note that when a bonding interface has no active links, the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002526driver will immediately reuse the first link that goes up, even if the
2527updelay parameter has been specified (the updelay is ignored in this
2528case). If there are slave interfaces waiting for the updelay timeout
2529to expire, the interface that first went into that state will be
2530immediately reused. This reduces down time of the network if the
2531value of updelay has been overestimated, and since this occurs only in
2532cases with no connectivity, there is no additional penalty for
2533ignoring the updelay.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534
2535 In addition to the concerns about switch timings, if your
2536switches take a long time to go into backup mode, it may be desirable
2537to not activate a backup interface immediately after a link goes down.
2538Failover may be delayed via the downdelay bonding module option.
2539
Auke Kok6224e012006-06-08 11:15:35 -0700254013.2 Duplicated Incoming Packets
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002541--------------------------------
2542
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002543 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2544suppress duplicate packets, which should largely eliminate this problem.
2545The following description is kept for reference.
2546
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002547 It is not uncommon to observe a short burst of duplicated
2548traffic when the bonding device is first used, or after it has been
2549idle for some period of time. This is most easily observed by issuing
2550a "ping" to some other host on the network, and noticing that the
2551output from ping flags duplicates (typically one per slave).
2552
2553 For example, on a bond in active-backup mode with five slaves
2554all connected to one switch, the output may appear as follows:
2555
2556# ping -n 10.0.4.2
2557PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
255864 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
255964 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
256064 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
256164 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
256264 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
256364 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
256464 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
256564 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2566
2567 This is not due to an error in the bonding driver, rather, it
2568is a side effect of how many switches update their MAC forwarding
2569tables. Initially, the switch does not associate the MAC address in
2570the packet with a particular switch port, and so it may send the
2571traffic to all ports until its MAC forwarding table is updated. Since
2572the interfaces attached to the bond may occupy multiple ports on a
2573single switch, when the switch (temporarily) floods the traffic to all
2574ports, the bond device receives multiple copies of the same packet
2575(one per slave device).
2576
2577 The duplicated packet behavior is switch dependent, some
2578switches exhibit this, and some do not. On switches that display this
2579behavior, it can be induced by clearing the MAC forwarding table (on
2580most Cisco switches, the privileged command "clear mac address-table
2581dynamic" will accomplish this).
2582
Auke Kok6224e012006-06-08 11:15:35 -0700258314. Hardware Specific Considerations
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584====================================
2585
2586 This section contains additional information for configuring
2587bonding on specific hardware platforms, or for interfacing bonding
2588with particular switches or other devices.
2589
Auke Kok6224e012006-06-08 11:15:35 -0700259014.1 IBM BladeCenter
Linus Torvalds1da177e2005-04-16 15:20:36 -07002591--------------------
2592
2593 This applies to the JS20 and similar systems.
2594
2595 On the JS20 blades, the bonding driver supports only
2596balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2597largely due to the network topology inside the BladeCenter, detailed
2598below.
2599
2600JS20 network adapter information
2601--------------------------------
2602
2603 All JS20s come with two Broadcom Gigabit Ethernet ports
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002604integrated on the planar (that's "motherboard" in IBM-speak). In the
2605BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2606I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2607An add-on Broadcom daughter card can be installed on a JS20 to provide
2608two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2609wired to I/O Modules 3 and 4, respectively.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610
2611 Each I/O Module may contain either a switch or a passthrough
2612module (which allows ports to be directly connected to an external
2613switch). Some bonding modes require a specific BladeCenter internal
2614network topology in order to function; these are detailed below.
2615
2616 Additional BladeCenter-specific networking information can be
2617found in two IBM Redbooks (www.ibm.com/redbooks):
2618
2619"IBM eServer BladeCenter Networking Options"
2620"IBM eServer BladeCenter Layer 2-7 Network Switching"
2621
2622BladeCenter networking configuration
2623------------------------------------
2624
2625 Because a BladeCenter can be configured in a very large number
2626of ways, this discussion will be confined to describing basic
2627configurations.
2628
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002629 Normally, Ethernet Switch Modules (ESMs) are used in I/O
Linus Torvalds1da177e2005-04-16 15:20:36 -07002630modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2631JS20 will be connected to different internal switches (in the
2632respective I/O modules).
2633
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002634 A passthrough module (OPM or CPM, optical or copper,
2635passthrough module) connects the I/O module directly to an external
2636switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2637interfaces of a JS20 can be redirected to the outside world and
2638connected to a common external switch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002639
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002640 Depending upon the mix of ESMs and PMs, the network will
2641appear to bonding as either a single switch topology (all PMs) or as a
2642multiple switch topology (one or more ESMs, zero or more PMs). It is
2643also possible to connect ESMs together, resulting in a configuration
2644much like the example in "High Availability in a Multiple Switch
2645Topology," above.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002646
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002647Requirements for specific modes
2648-------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07002649
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002650 The balance-rr mode requires the use of passthrough modules
2651for devices in the bond, all connected to an common external switch.
2652That switch must be configured for "etherchannel" or "trunking" on the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653appropriate ports, as is usual for balance-rr.
2654
2655 The balance-alb and balance-tlb modes will function with
2656either switch modules or passthrough modules (or a mix). The only
2657specific requirement for these modes is that all network interfaces
2658must be able to reach all destinations for traffic sent over the
2659bonding device (i.e., the network must converge at some point outside
2660the BladeCenter).
2661
2662 The active-backup mode has no additional requirements.
2663
2664Link monitoring issues
2665----------------------
2666
2667 When an Ethernet Switch Module is in place, only the ARP
2668monitor will reliably detect link loss to an external switch. This is
2669nothing unusual, but examination of the BladeCenter cabinet would
2670suggest that the "external" network ports are the ethernet ports for
2671the system, when it fact there is a switch between these "external"
2672ports and the devices on the JS20 system itself. The MII monitor is
2673only able to detect link failures between the ESM and the JS20 system.
2674
2675 When a passthrough module is in place, the MII monitor does
2676detect failures to the "external" port, which is then directly
2677connected to the JS20 system.
2678
2679Other concerns
2680--------------
2681
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002682 The Serial Over LAN (SoL) link is established over the primary
Linus Torvalds1da177e2005-04-16 15:20:36 -07002683ethernet (eth0) only, therefore, any loss of link to eth0 will result
2684in losing your SoL connection. It will not fail over with other
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002685network traffic, as the SoL system is beyond the control of the
2686bonding driver.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002687
2688 It may be desirable to disable spanning tree on the switch
2689(either the internal Ethernet Switch Module, or an external switch) to
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002690avoid fail-over delay issues when using bonding.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002691
2692
Auke Kok6224e012006-06-08 11:15:35 -0700269315. Frequently Asked Questions
Linus Torvalds1da177e2005-04-16 15:20:36 -07002694==============================
2695
26961. Is it SMP safe?
2697
2698 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2699The new driver was designed to be SMP safe from the start.
2700
27012. What type of cards will work with it?
2702
2703 Any Ethernet type cards (you can even mix cards - a Intel
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002704EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2705devices need not be of the same speed.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002707 Starting with version 3.2.1, bonding also supports Infiniband
2708slaves in active-backup mode.
2709
Linus Torvalds1da177e2005-04-16 15:20:36 -070027103. How many bonding devices can I have?
2711
2712 There is no limit.
2713
27144. How many slaves can a bonding device have?
2715
2716 This is limited only by the number of network interfaces Linux
2717supports and/or the number of network cards you can place in your
2718system.
2719
27205. What happens when a slave link dies?
2721
2722 If link monitoring is enabled, then the failing device will be
2723disabled. The active-backup mode will fail over to a backup link, and
2724other modes will ignore the failed link. The link will continue to be
2725monitored, and should it recover, it will rejoin the bond (in whatever
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002726manner is appropriate for the mode). See the sections on High
2727Availability and the documentation for each mode for additional
2728information.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002729
2730 Link monitoring can be enabled via either the miimon or
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002731arp_interval parameters (described in the module parameters section,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732above). In general, miimon monitors the carrier state as sensed by
2733the underlying network device, and the arp monitor (arp_interval)
2734monitors connectivity to another host on the local network.
2735
2736 If no link monitoring is configured, the bonding driver will
2737be unable to detect link failures, and will assume that all links are
2738always available. This will likely result in lost packets, and a
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002739resulting degradation of performance. The precise performance loss
Linus Torvalds1da177e2005-04-16 15:20:36 -07002740depends upon the bonding mode and network configuration.
2741
27426. Can bonding be used for High Availability?
2743
2744 Yes. See the section on High Availability for details.
2745
27467. Which switches/systems does it work with?
2747
2748 The full answer to this depends upon the desired mode.
2749
2750 In the basic balance modes (balance-rr and balance-xor), it
2751works with any system that supports etherchannel (also called
2752trunking). Most managed switches currently available have such
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002753support, and many unmanaged switches as well.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002754
2755 The advanced balance modes (balance-tlb and balance-alb) do
2756not have special switch requirements, but do need device drivers that
2757support specific features (described in the appropriate section under
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002758module parameters, above).
Linus Torvalds1da177e2005-04-16 15:20:36 -07002759
Auke Kok6224e012006-06-08 11:15:35 -07002760 In 802.3ad mode, it works with systems that support IEEE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002761802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2762switches currently available support 802.3ad.
2763
2764 The active-backup mode should work with any Layer-II switch.
2765
27668. Where does a bonding device get its MAC address from?
2767
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002768 When using slave devices that have fixed MAC addresses, or when
2769the fail_over_mac option is enabled, the bonding device's MAC address is
2770the MAC address of the active slave.
2771
2772 For other configurations, if not explicitly configured (with
2773ifconfig or ip link), the MAC address of the bonding device is taken from
2774its first slave device. This MAC address is then passed to all following
2775slaves and remains persistent (even if the first slave is removed) until
2776the bonding device is brought down or reconfigured.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002777
2778 If you wish to change the MAC address, you can set it with
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002779ifconfig or ip link:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002780
2781# ifconfig bond0 hw ether 00:11:22:33:44:55
2782
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002783# ip link set bond0 address 66:77:88:99:aa:bb
2784
Linus Torvalds1da177e2005-04-16 15:20:36 -07002785 The MAC address can be also changed by bringing down/up the
2786device and then changing its slaves (or their order):
2787
2788# ifconfig bond0 down ; modprobe -r bonding
2789# ifconfig bond0 .... up
2790# ifenslave bond0 eth...
2791
2792 This method will automatically take the address from the next
2793slave that is added.
2794
2795 To restore your slaves' MAC addresses, you need to detach them
2796from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2797then restore the MAC addresses that the slaves had before they were
2798enslaved.
2799
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700280016. Resources and Links
Linus Torvalds1da177e2005-04-16 15:20:36 -07002801=======================
2802
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002803 The latest version of the bonding driver can be found in the latest
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804version of the linux kernel, found on http://kernel.org
2805
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002806 The latest version of this document can be found in the latest kernel
2807source (named Documentation/networking/bonding.txt).
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002808
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002809 Discussions regarding the usage of the bonding driver take place on the
2810bonding-devel mailing list, hosted at sourceforge.net. If you have questions or
2811problems, post them to the list. The list address is:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002812
2813bonding-devel@lists.sourceforge.net
2814
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002815 The administrative interface (to subscribe or unsubscribe) can
2816be found at:
2817
Linus Torvalds1da177e2005-04-16 15:20:36 -07002818https://lists.sourceforge.net/lists/listinfo/bonding-devel
2819
Rick Jonesf8b72d32012-07-20 10:51:37 +00002820 Discussions regarding the development of the bonding driver take place
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002821on the main Linux network mailing list, hosted at vger.kernel.org. The list
2822address is:
2823
2824netdev@vger.kernel.org
2825
2826 The administrative interface (to subscribe or unsubscribe) can
2827be found at:
2828
2829http://vger.kernel.org/vger-lists.html#netdev
2830
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831Donald Becker's Ethernet Drivers and diag programs may be found at :
Justin P. Mattock0ea6e612010-07-23 20:51:24 -07002832 - http://web.archive.org/web/*/http://www.scyld.com/network/
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833
2834You will also find a lot of information regarding Ethernet, NWay, MII,
2835etc. at www.scyld.com.
2836
2837-- END --