blob: 5cdb22971d19753607739a66108bd85fdf4a492c [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 Hutchingsad246c92011-04-26 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
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
Auke Kok6224e012006-06-08 11:15:35 -0700554. Querying Bonding Configuration
564.1 Bonding Configuration
574.2 Network Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
Auke Kok6224e012006-06-08 11:15:35 -0700595. Switch Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Auke Kok6224e012006-06-08 11:15:35 -0700616. 802.1q VLAN Support
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
Auke Kok6224e012006-06-08 11:15:35 -0700637. Link Monitoring
647.1 ARP Monitor Operation
657.2 Configuring Multiple ARP Targets
667.3 MII Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
Auke Kok6224e012006-06-08 11:15:35 -0700688. Potential Trouble Sources
698.1 Adventures in Routing
708.2 Ethernet Device Renaming
718.3 Painfully Slow Or No Failed Link Detection By Miimon
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
Auke Kok6224e012006-06-08 11:15:35 -0700739. SNMP agents
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Auke Kok6224e012006-06-08 11:15:35 -07007510. Promiscuous mode
Linus Torvalds1da177e2005-04-16 15:20:36 -070076
Auke Kok6224e012006-06-08 11:15:35 -07007711. Configuring Bonding for High Availability
7811.1 High Availability in a Single Switch Topology
7911.2 High Availability in a Multiple Switch Topology
8011.2.1 HA Bonding Mode Selection for Multiple Switch Topology
8111.2.2 HA Link Monitoring for Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
Auke Kok6224e012006-06-08 11:15:35 -07008312. Configuring Bonding for Maximum Throughput
8412.1 Maximum Throughput in a Single Switch Topology
8512.1.1 MT Bonding Mode Selection for Single Switch Topology
8612.1.2 MT Link Monitoring for Single Switch Topology
8712.2 Maximum Throughput in a Multiple Switch Topology
8812.2.1 MT Bonding Mode Selection for Multiple Switch Topology
8912.2.2 MT Link Monitoring for Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
Auke Kok6224e012006-06-08 11:15:35 -07009113. Switch Behavior Issues
9213.1 Link Establishment and Failover Delays
9313.2 Duplicated Incoming Packets
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
Auke Kok6224e012006-06-08 11:15:35 -07009514. Hardware Specific Considerations
9614.1 IBM BladeCenter
Jay Vosburgh00354cf2005-07-21 12:18:02 -070097
Auke Kok6224e012006-06-08 11:15:35 -07009815. Frequently Asked Questions
Jay Vosburgh00354cf2005-07-21 12:18:02 -070099
Auke Kok6224e012006-06-08 11:15:35 -070010016. Resources and Links
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
102
1031. Bonding Driver Installation
104==============================
105
106 Most popular distro kernels ship with the bonding driver
Cong Wangb1098bb2013-05-27 15:49:16 +0000107already available as a module. If your distro does not, or you
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108have need to compile bonding from source (e.g., configuring and
109installing a mainline kernel from kernel.org), you'll need to perform
110the following steps:
111
1121.1 Configure and build the kernel with bonding
113-----------------------------------------------
114
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700115 The current version of the bonding driver is available in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116drivers/net/bonding subdirectory of the most recent kernel source
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700117(which is available on http://kernel.org). Most users "rolling their
118own" will want to use the most recent kernel from kernel.org.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
120 Configure kernel with "make menuconfig" (or "make xconfig" or
121"make config"), then select "Bonding driver support" in the "Network
122device support" section. It is recommended that you configure the
123driver as module since it is currently the only way to pass parameters
124to the driver or configure more than one bonding device.
125
Cong Wangb1098bb2013-05-27 15:49:16 +0000126 Build and install the new kernel and modules.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
Cong Wangb1098bb2013-05-27 15:49:16 +00001281.2 Bonding Control Utility
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129-------------------------------------
130
Cong Wangb1098bb2013-05-27 15:49:16 +0000131 It is recommended to configure bonding via iproute2 (netlink)
132or sysfs, the old ifenslave control utility is obsolete.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133
1342. Bonding Driver Options
135=========================
136
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800137 Options for the bonding driver are supplied as parameters to the
138bonding module at load time, or are specified via sysfs.
139
140 Module options may be given as command line arguments to the
141insmod or modprobe command, but are usually specified in either the
Lucas De Marchi970e2482012-03-30 13:37:16 -0700142/etc/modrobe.d/*.conf configuration files, or in a distro-specific
143configuration file (some of which are detailed in the next section).
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800144
145 Details on bonding support for sysfs is provided in the
146"Configuring Bonding Manually via Sysfs" section, below.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147
148 The available bonding driver parameters are listed below. If a
149parameter is not specified the default value is used. When initially
150configuring a bond, it is recommended "tail -f /var/log/messages" be
151run in a separate window to watch for bonding driver error messages.
152
153 It is critical that either the miimon or arp_interval and
154arp_ip_target parameters be specified, otherwise serious network
155degradation will occur during link failures. Very few devices do not
156support at least miimon, so there is really no reason not to use it.
157
158 Options with textual values will accept either the text name
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700159or, for backwards compatibility, the option value. E.g.,
160"mode=802.3ad" and "mode=4" set the same mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
162 The parameters are as follows:
163
Nicolas de Pesloüan1ba9ac72011-12-26 13:35:24 +0000164active_slave
165
166 Specifies the new active slave for modes that support it
167 (active-backup, balance-alb and balance-tlb). Possible values
168 are the name of any currently enslaved interface, or an empty
169 string. If a name is given, the slave and its link must be up in order
170 to be selected as the new active slave. If an empty string is
171 specified, the current active slave is cleared, and a new active
172 slave is selected automatically.
173
174 Note that this is only available through the sysfs interface. No module
175 parameter by this name exists.
176
177 The normal value of this option is the name of the currently
178 active slave, or the empty string if there is no active slave or
179 the current mode does not use an active slave.
180
Jay Vosburghfd989c82008-11-04 17:51:16 -0800181ad_select
182
183 Specifies the 802.3ad aggregation selection logic to use. The
184 possible values and their effects are:
185
186 stable or 0
187
188 The active aggregator is chosen by largest aggregate
189 bandwidth.
190
191 Reselection of the active aggregator occurs only when all
192 slaves of the active aggregator are down or the active
193 aggregator has no slaves.
194
195 This is the default value.
196
197 bandwidth or 1
198
199 The active aggregator is chosen by largest aggregate
200 bandwidth. Reselection occurs if:
201
202 - A slave is added to or removed from the bond
203
204 - Any slave's link state changes
205
206 - Any slave's 802.3ad association state changes
207
Matt LaPlante19f59462009-04-27 15:06:31 +0200208 - The bond's administrative state changes to up
Jay Vosburghfd989c82008-11-04 17:51:16 -0800209
210 count or 2
211
212 The active aggregator is chosen by the largest number of
213 ports (slaves). Reselection occurs as described under the
214 "bandwidth" setting, above.
215
216 The bandwidth and count selection policies permit failover of
217 802.3ad aggregations when partial failure of the active aggregator
218 occurs. This keeps the aggregator with the highest availability
219 (either in bandwidth or in number of ports) active at all times.
220
221 This option was added in bonding version 3.4.0.
222
Nicolas de Pesloüan025890b2011-08-06 07:06:39 +0000223all_slaves_active
224
225 Specifies that duplicate frames (received on inactive ports) should be
226 dropped (0) or delivered (1).
227
228 Normally, bonding will drop duplicate frames (received on inactive
229 ports), which is desirable for most users. But there are some times
230 it is nice to allow duplicate frames to be delivered.
231
232 The default value is 0 (drop duplicate frames received on inactive
233 ports).
234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235arp_interval
236
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700237 Specifies the ARP link monitoring frequency in milliseconds.
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700238
239 The ARP monitor works by periodically checking the slave
240 devices to determine whether they have sent or received
241 traffic recently (the precise criteria depends upon the
242 bonding mode, and the state of the slave). Regular traffic is
243 generated via ARP probes issued for the addresses specified by
244 the arp_ip_target option.
245
246 This behavior can be modified by the arp_validate option,
247 below.
248
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700249 If ARP monitoring is used in an etherchannel compatible mode
250 (modes 0 and 2), the switch should be configured in a mode
251 that evenly distributes packets across all links. If the
252 switch is configured to distribute the packets in an XOR
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 fashion, all replies from the ARP targets will be received on
254 the same link which could cause the other team members to
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700255 fail. ARP monitoring should not be used in conjunction with
256 miimon. A value of 0 disables ARP monitoring. The default
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 value is 0.
258
259arp_ip_target
260
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700261 Specifies the IP addresses to use as ARP monitoring peers when
262 arp_interval is > 0. These are the targets of the ARP request
263 sent to determine the health of the link to the targets.
264 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
265 addresses must be separated by a comma. At least one IP
266 address must be given for ARP monitoring to function. The
267 maximum number of targets that can be specified is 16. The
268 default value is no IP addresses.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700270arp_validate
271
272 Specifies whether or not ARP probes and replies should be
273 validated in the active-backup mode. This causes the ARP
274 monitor to examine the incoming ARP requests and replies, and
275 only consider a slave to be up if it is receiving the
276 appropriate ARP traffic.
277
278 Possible values are:
279
280 none or 0
281
282 No validation is performed. This is the default.
283
284 active or 1
285
286 Validation is performed only for the active slave.
287
288 backup or 2
289
290 Validation is performed only for backup slaves.
291
292 all or 3
293
294 Validation is performed for all slaves.
295
296 For the active slave, the validation checks ARP replies to
297 confirm that they were generated by an arp_ip_target. Since
298 backup slaves do not typically receive these replies, the
299 validation performed for backup slaves is on the ARP request
300 sent out via the active slave. It is possible that some
301 switch or network configurations may result in situations
302 wherein the backup slaves do not receive the ARP requests; in
303 such a situation, validation of backup slaves must be
304 disabled.
305
Veaceslav Falicod7d35c62013-06-24 11:49:33 +0200306 The validation of ARP requests on backup slaves is mainly
307 helping bonding to decide which slaves are more likely to
308 work in case of the active slave failure, it doesn't really
309 guarantee that the backup slave will work if it's selected
310 as the next active slave.
311
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700312 This option is useful in network configurations in which
313 multiple bonding hosts are concurrently issuing ARPs to one or
314 more targets beyond a common switch. Should the link between
315 the switch and target fail (but not the switch itself), the
316 probe traffic generated by the multiple bonding instances will
317 fool the standard ARP monitor into considering the links as
318 still up. Use of the arp_validate option can resolve this, as
319 the ARP monitor will only consider ARP requests and replies
320 associated with its own instance of bonding.
321
322 This option was added in bonding version 3.1.0.
323
Veaceslav Falico8599b522013-06-24 11:49:34 +0200324arp_all_targets
325
326 Specifies the quantity of arp_ip_targets that must be reachable
327 in order for the ARP monitor to consider a slave as being up.
328 This option affects only active-backup mode for slaves with
329 arp_validation enabled.
330
331 Possible values are:
332
333 any or 0
334
335 consider the slave up only when any of the arp_ip_targets
336 is reachable
337
338 all or 1
339
340 consider the slave up only when all of the arp_ip_targets
341 are reachable
342
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343downdelay
344
345 Specifies the time, in milliseconds, to wait before disabling
346 a slave after a link failure has been detected. This option
347 is only valid for the miimon link monitor. The downdelay
348 value should be a multiple of the miimon value; if not, it
349 will be rounded down to the nearest multiple. The default
350 value is 0.
351
Jay Vosburghdd957c52007-10-09 19:57:24 -0700352fail_over_mac
353
354 Specifies whether active-backup mode should set all slaves to
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700355 the same MAC address at enslavement (the traditional
356 behavior), or, when enabled, perform special handling of the
357 bond's MAC address in accordance with the selected policy.
Jay Vosburghdd957c52007-10-09 19:57:24 -0700358
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700359 Possible values are:
Jay Vosburghdd957c52007-10-09 19:57:24 -0700360
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700361 none or 0
Jay Vosburghdd957c52007-10-09 19:57:24 -0700362
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700363 This setting disables fail_over_mac, and causes
364 bonding to set all slaves of an active-backup bond to
365 the same MAC address at enslavement time. This is the
366 default.
Jay Vosburghdd957c52007-10-09 19:57:24 -0700367
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700368 active or 1
Jay Vosburghdd957c52007-10-09 19:57:24 -0700369
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700370 The "active" fail_over_mac policy indicates that the
371 MAC address of the bond should always be the MAC
372 address of the currently active slave. The MAC
373 address of the slaves is not changed; instead, the MAC
374 address of the bond changes during a failover.
375
376 This policy is useful for devices that cannot ever
377 alter their MAC address, or for devices that refuse
378 incoming broadcasts with their own source MAC (which
379 interferes with the ARP monitor).
380
381 The down side of this policy is that every device on
382 the network must be updated via gratuitous ARP,
383 vs. just updating a switch or set of switches (which
384 often takes place for any traffic, not just ARP
385 traffic, if the switch snoops incoming traffic to
386 update its tables) for the traditional method. If the
387 gratuitous ARP is lost, communication may be
388 disrupted.
389
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300390 When this policy is used in conjunction with the mii
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700391 monitor, devices which assert link up prior to being
392 able to actually transmit and receive are particularly
Matt LaPlante19f59462009-04-27 15:06:31 +0200393 susceptible to loss of the gratuitous ARP, and an
Jay Vosburgh3915c1e82008-05-17 21:10:14 -0700394 appropriate updelay setting may be required.
395
396 follow or 2
397
398 The "follow" fail_over_mac policy causes the MAC
399 address of the bond to be selected normally (normally
400 the MAC address of the first slave added to the bond).
401 However, the second and subsequent slaves are not set
402 to this MAC address while they are in a backup role; a
403 slave is programmed with the bond's MAC address at
404 failover time (and the formerly active slave receives
405 the newly active slave's MAC address).
406
407 This policy is useful for multiport devices that
408 either become confused or incur a performance penalty
409 when multiple ports are programmed with the same MAC
410 address.
411
412
413 The default policy is none, unless the first slave cannot
414 change its MAC address, in which case the active policy is
415 selected by default.
416
417 This option may be modified via sysfs only when no slaves are
418 present in the bond.
419
420 This option was added in bonding version 3.2.0. The "follow"
421 policy was added in bonding version 3.3.0.
Jay Vosburghdd957c52007-10-09 19:57:24 -0700422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423lacp_rate
424
425 Option specifying the rate in which we'll ask our link partner
426 to transmit LACPDU packets in 802.3ad mode. Possible values
427 are:
428
429 slow or 0
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700430 Request partner to transmit LACPDUs every 30 seconds
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
432 fast or 1
433 Request partner to transmit LACPDUs every 1 second
434
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700435 The default is slow.
436
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437max_bonds
438
439 Specifies the number of bonding devices to create for this
440 instance of the bonding driver. E.g., if max_bonds is 3, and
441 the bonding driver is not already loaded, then bond0, bond1
Jay Vosburghb8a97872008-06-13 18:12:04 -0700442 and bond2 will be created. The default value is 1. Specifying
443 a value of 0 will load bonding, but will not create any devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
445miimon
446
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700447 Specifies the MII link monitoring frequency in milliseconds.
448 This determines how often the link state of each slave is
449 inspected for link failures. A value of zero disables MII
450 link monitoring. A value of 100 is a good starting point.
451 The use_carrier option, below, affects how the link state is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 determined. See the High Availability section for additional
453 information. The default value is 0.
454
Nicolas de Pesloüan025890b2011-08-06 07:06:39 +0000455min_links
456
457 Specifies the minimum number of links that must be active before
458 asserting carrier. It is similar to the Cisco EtherChannel min-links
459 feature. This allows setting the minimum number of member ports that
460 must be up (link-up state) before marking the bond device as up
461 (carrier on). This is useful for situations where higher level services
462 such as clustering want to ensure a minimum number of low bandwidth
463 links are active before switchover. This option only affect 802.3ad
464 mode.
465
466 The default value is 0. This will cause carrier to be asserted (for
467 802.3ad mode) whenever there is an active aggregator, regardless of the
468 number of available links in that aggregator. Note that, because an
469 aggregator cannot be active without at least one available link,
470 setting this option to 0 or to 1 has the exact same effect.
471
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472mode
473
474 Specifies one of the bonding policies. The default is
475 balance-rr (round robin). Possible values are:
476
477 balance-rr or 0
478
479 Round-robin policy: Transmit packets in sequential
480 order from the first available slave through the
481 last. This mode provides load balancing and fault
482 tolerance.
483
484 active-backup or 1
485
486 Active-backup policy: Only one slave in the bond is
487 active. A different slave becomes active if, and only
488 if, the active slave fails. The bond's MAC address is
489 externally visible on only one port (network adapter)
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700490 to avoid confusing the switch.
491
492 In bonding version 2.6.2 or later, when a failover
493 occurs in active-backup mode, bonding will issue one
494 or more gratuitous ARPs on the newly active slave.
Auke Kok6224e012006-06-08 11:15:35 -0700495 One gratuitous ARP is issued for the bonding master
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700496 interface and each VLAN interfaces configured above
497 it, provided that the interface has at least one IP
498 address configured. Gratuitous ARPs issued for VLAN
499 interfaces are tagged with the appropriate VLAN id.
500
501 This mode provides fault tolerance. The primary
502 option, documented below, affects the behavior of this
503 mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
505 balance-xor or 2
506
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700507 XOR policy: Transmit based on the selected transmit
508 hash policy. The default policy is a simple [(source
509 MAC address XOR'd with destination MAC address) modulo
510 slave count]. Alternate transmit policies may be
511 selected via the xmit_hash_policy option, described
512 below.
513
514 This mode provides load balancing and fault tolerance.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
516 broadcast or 3
517
518 Broadcast policy: transmits everything on all slave
519 interfaces. This mode provides fault tolerance.
520
521 802.3ad or 4
522
523 IEEE 802.3ad Dynamic link aggregation. Creates
524 aggregation groups that share the same speed and
525 duplex settings. Utilizes all slaves in the active
526 aggregator according to the 802.3ad specification.
527
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700528 Slave selection for outgoing traffic is done according
529 to the transmit hash policy, which may be changed from
530 the default simple XOR policy via the xmit_hash_policy
531 option, documented below. Note that not all transmit
532 policies may be 802.3ad compliant, particularly in
533 regards to the packet mis-ordering requirements of
534 section 43.2.4 of the 802.3ad standard. Differing
535 peer implementations will have varying tolerances for
536 noncompliance.
537
538 Prerequisites:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
540 1. Ethtool support in the base drivers for retrieving
541 the speed and duplex of each slave.
542
543 2. A switch that supports IEEE 802.3ad Dynamic link
544 aggregation.
545
546 Most switches will require some type of configuration
547 to enable 802.3ad mode.
548
549 balance-tlb or 5
550
551 Adaptive transmit load balancing: channel bonding that
552 does not require any special switch support. The
553 outgoing traffic is distributed according to the
554 current load (computed relative to the speed) on each
555 slave. Incoming traffic is received by the current
556 slave. If the receiving slave fails, another slave
557 takes over the MAC address of the failed receiving
558 slave.
559
560 Prerequisite:
561
562 Ethtool support in the base drivers for retrieving the
563 speed of each slave.
564
565 balance-alb or 6
566
567 Adaptive load balancing: includes balance-tlb plus
568 receive load balancing (rlb) for IPV4 traffic, and
569 does not require any special switch support. The
570 receive load balancing is achieved by ARP negotiation.
571 The bonding driver intercepts the ARP Replies sent by
572 the local system on their way out and overwrites the
573 source hardware address with the unique hardware
574 address of one of the slaves in the bond such that
575 different peers use different hardware addresses for
576 the server.
577
578 Receive traffic from connections created by the server
579 is also balanced. When the local system sends an ARP
580 Request the bonding driver copies and saves the peer's
581 IP information from the ARP packet. When the ARP
582 Reply arrives from the peer, its hardware address is
583 retrieved and the bonding driver initiates an ARP
584 reply to this peer assigning it to one of the slaves
585 in the bond. A problematic outcome of using ARP
586 negotiation for balancing is that each time that an
587 ARP request is broadcast it uses the hardware address
588 of the bond. Hence, peers learn the hardware address
589 of the bond and the balancing of receive traffic
590 collapses to the current slave. This is handled by
591 sending updates (ARP Replies) to all the peers with
592 their individually assigned hardware address such that
593 the traffic is redistributed. Receive traffic is also
594 redistributed when a new slave is added to the bond
595 and when an inactive slave is re-activated. The
596 receive load is distributed sequentially (round robin)
597 among the group of highest speed slaves in the bond.
598
599 When a link is reconnected or a new slave joins the
600 bond the receive traffic is redistributed among all
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700601 active slaves in the bond by initiating ARP Replies
Auke Kok6224e012006-06-08 11:15:35 -0700602 with the selected MAC address to each of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 clients. The updelay parameter (detailed below) must
604 be set to a value equal or greater than the switch's
605 forwarding delay so that the ARP Replies sent to the
606 peers will not be blocked by the switch.
607
608 Prerequisites:
609
610 1. Ethtool support in the base drivers for retrieving
611 the speed of each slave.
612
613 2. Base driver support for setting the hardware
614 address of a device while it is open. This is
615 required so that there will always be one slave in the
616 team using the bond hardware address (the
617 curr_active_slave) while having a unique hardware
618 address for each slave in the bond. If the
619 curr_active_slave fails its hardware address is
620 swapped with the new curr_active_slave that was
621 chosen.
622
Jay Vosburghb59f9f72008-06-13 18:12:03 -0700623num_grat_arp
Brian Haley305d5522008-11-04 17:51:14 -0800624num_unsol_na
625
Ben Hutchingsad246c92011-04-26 15:25:52 +0000626 Specify the number of peer notifications (gratuitous ARPs and
627 unsolicited IPv6 Neighbor Advertisements) to be issued after a
628 failover event. As soon as the link is up on the new slave
629 (possibly immediately) a peer notification is sent on the
630 bonding device and each VLAN sub-device. This is repeated at
631 each link monitor interval (arp_interval or miimon, whichever
632 is active) if the number is greater than 1.
Brian Haley305d5522008-11-04 17:51:14 -0800633
Ben Hutchingsad246c92011-04-26 15:25:52 +0000634 The valid range is 0 - 255; the default value is 1. These options
635 affect only the active-backup mode. These options were added for
636 bonding versions 3.3.0 and 3.4.0 respectively.
637
Jesper Juhl8fb4e132011-08-01 17:59:44 -0700638 From Linux 3.0 and bonding version 3.7.1, these notifications
Ben Hutchingsad246c92011-04-26 15:25:52 +0000639 are generated by the ipv4 and ipv6 code and the numbers of
640 repetitions cannot be set independently.
Brian Haley305d5522008-11-04 17:51:14 -0800641
Nikolay Aleksandrov12465fb2013-11-05 13:51:42 +0100642packets_per_slave
643
644 Specify the number of packets to transmit through a slave before
645 moving to the next one. When set to 0 then a slave is chosen at
646 random.
647
648 The valid range is 0 - 65535; the default value is 1. This option
649 has effect only in balance-rr mode.
650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651primary
652
653 A string (eth0, eth2, etc) specifying which slave is the
654 primary device. The specified device will always be the
655 active slave while it is available. Only when the primary is
656 off-line will alternate devices be used. This is useful when
657 one slave is preferred over another, e.g., when one slave has
658 higher throughput than another.
659
dingtianhonge1d206a2014-01-18 16:28:57 +0800660 The primary option is only valid for active-backup(1),
661 balance-tlb (5) and balance-alb (6) mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662
Jiri Pirkoa5499522009-09-25 03:28:09 +0000663primary_reselect
664
665 Specifies the reselection policy for the primary slave. This
666 affects how the primary slave is chosen to become the active slave
667 when failure of the active slave or recovery of the primary slave
668 occurs. This option is designed to prevent flip-flopping between
669 the primary slave and other slaves. Possible values are:
670
671 always or 0 (default)
672
673 The primary slave becomes the active slave whenever it
674 comes back up.
675
676 better or 1
677
678 The primary slave becomes the active slave when it comes
679 back up, if the speed and duplex of the primary slave is
680 better than the speed and duplex of the current active
681 slave.
682
683 failure or 2
684
685 The primary slave becomes the active slave only if the
686 current active slave fails and the primary slave is up.
687
688 The primary_reselect setting is ignored in two cases:
689
690 If no slaves are active, the first slave to recover is
691 made the active slave.
692
693 When initially enslaved, the primary slave is always made
694 the active slave.
695
696 Changing the primary_reselect policy via sysfs will cause an
697 immediate selection of the best active slave according to the new
698 policy. This may or may not result in a change of the active
699 slave, depending upon the circumstances.
700
701 This option was added for bonding version 3.6.0.
702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703updelay
704
705 Specifies the time, in milliseconds, to wait before enabling a
706 slave after a link recovery has been detected. This option is
707 only valid for the miimon link monitor. The updelay value
708 should be a multiple of the miimon value; if not, it will be
709 rounded down to the nearest multiple. The default value is 0.
710
711use_carrier
712
713 Specifies whether or not miimon should use MII or ETHTOOL
714 ioctls vs. netif_carrier_ok() to determine the link
715 status. The MII or ETHTOOL ioctls are less efficient and
716 utilize a deprecated calling sequence within the kernel. The
717 netif_carrier_ok() relies on the device driver to maintain its
718 state with netif_carrier_on/off; at this writing, most, but
719 not all, device drivers support this facility.
720
721 If bonding insists that the link is up when it should not be,
722 it may be that your network device driver does not support
723 netif_carrier_on/off. The default state for netif_carrier is
724 "carrier on," so if a driver does not support netif_carrier,
725 it will appear as if the link is always up. In this case,
726 setting use_carrier to 0 will cause bonding to revert to the
727 MII / ETHTOOL ioctl method to determine the link state.
728
729 A value of 1 enables the use of netif_carrier_ok(), a value of
730 0 will use the deprecated MII / ETHTOOL ioctls. The default
731 value is 1.
732
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700733xmit_hash_policy
734
735 Selects the transmit hash policy to use for slave selection in
736 balance-xor and 802.3ad modes. Possible values are:
737
738 layer2
739
740 Uses XOR of hardware MAC addresses to generate the
741 hash. The formula is
742
743 (source MAC XOR destination MAC) modulo slave count
744
745 This algorithm will place all traffic to a particular
746 network peer on the same slave.
747
748 This algorithm is 802.3ad compliant.
749
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800750 layer2+3
751
752 This policy uses a combination of layer2 and layer3
753 protocol information to generate the hash.
754
755 Uses XOR of hardware MAC addresses and IP addresses to
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200756 generate the hash. The formula is
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800757
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200758 hash = source MAC XOR destination MAC
759 hash = hash XOR source IP XOR destination IP
760 hash = hash XOR (hash RSHIFT 16)
761 hash = hash XOR (hash RSHIFT 8)
762 And then hash is reduced modulo slave count.
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800763
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200764 If the protocol is IPv6 then the source and destination
765 addresses are first hashed using ipv6_addr_hash.
John Eaglesham6b923cb2012-08-21 20:43:35 +0000766
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800767 This algorithm will place all traffic to a particular
768 network peer on the same slave. For non-IP traffic,
769 the formula is the same as for the layer2 transmit
770 hash policy.
771
772 This policy is intended to provide a more balanced
773 distribution of traffic than layer2 alone, especially
774 in environments where a layer3 gateway device is
775 required to reach most destinations.
776
Matt LaPlanted9195882008-07-25 19:45:33 -0700777 This algorithm is 802.3ad compliant.
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800778
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700779 layer3+4
780
781 This policy uses upper layer protocol information,
782 when available, to generate the hash. This allows for
783 traffic to a particular network peer to span multiple
784 slaves, although a single connection will not span
785 multiple slaves.
786
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200787 The formula for unfragmented TCP and UDP packets is
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700788
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200789 hash = source port, destination port (as in the header)
790 hash = hash XOR source IP XOR destination IP
791 hash = hash XOR (hash RSHIFT 16)
792 hash = hash XOR (hash RSHIFT 8)
793 And then hash is reduced modulo slave count.
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700794
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200795 If the protocol is IPv6 then the source and destination
796 addresses are first hashed using ipv6_addr_hash.
John Eaglesham6b923cb2012-08-21 20:43:35 +0000797
798 For fragmented TCP or UDP packets and all other IPv4 and
799 IPv6 protocol traffic, the source and destination port
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700800 information is omitted. For non-IP traffic, the
801 formula is the same as for the layer2 transmit hash
802 policy.
803
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700804 This algorithm is not fully 802.3ad compliant. A
805 single TCP or UDP conversation containing both
806 fragmented and unfragmented packets will see packets
807 striped across two interfaces. This may result in out
808 of order delivery. Most traffic types will not meet
809 this criteria, as TCP rarely fragments traffic, and
810 most UDP traffic is not involved in extended
811 conversations. Other implementations of 802.3ad may
812 or may not tolerate this noncompliance.
813
Nikolay Aleksandrov7a6afab2013-10-02 13:39:26 +0200814 encap2+3
815
816 This policy uses the same formula as layer2+3 but it
817 relies on skb_flow_dissect to obtain the header fields
818 which might result in the use of inner headers if an
819 encapsulation protocol is used. For example this will
820 improve the performance for tunnel users because the
821 packets will be distributed according to the encapsulated
822 flows.
823
824 encap3+4
825
826 This policy uses the same formula as layer3+4 but it
827 relies on skb_flow_dissect to obtain the header fields
828 which might result in the use of inner headers if an
829 encapsulation protocol is used. For example this will
830 improve the performance for tunnel users because the
831 packets will be distributed according to the encapsulated
832 flows.
833
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700834 The default value is layer2. This option was added in bonding
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800835 version 2.6.3. In earlier versions of bonding, this parameter
836 does not exist, and the layer2 policy is the only policy. The
837 layer2+3 value was added for bonding version 3.2.2.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
Flavio Leitnerc2952c32010-10-05 14:23:59 +0000839resend_igmp
840
841 Specifies the number of IGMP membership reports to be issued after
842 a failover event. One membership report is issued immediately after
843 the failover, subsequent packets are sent in each 200ms interval.
844
Flavio Leitner94265cf2011-05-25 08:38:58 +0000845 The valid range is 0 - 255; the default value is 1. A value of 0
846 prevents the IGMP membership report from being issued in response
847 to the failover event.
848
849 This option is useful for bonding modes balance-rr (0), active-backup
850 (1), balance-tlb (5) and balance-alb (6), in which a failover can
851 switch the IGMP traffic from one slave to another. Therefore a fresh
852 IGMP report must be issued to cause the switch to forward the incoming
853 IGMP traffic over the newly selected slave.
854
855 This option was added for bonding version 3.7.0.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
dingtianhong84a6a0a2013-12-21 14:40:22 +0800857lp_interval
858
859 Specifies the number of seconds between instances where the bonding
860 driver sends learning packets to each slaves peer switch.
861
862 The valid range is 1 - 0x7fffffff; the default value is 1. This Option
863 has effect only in balance-tlb and balance-alb modes.
864
Linus Torvalds1da177e2005-04-16 15:20:36 -07008653. Configuring Bonding Devices
866==============================
867
Auke Kok6224e012006-06-08 11:15:35 -0700868 You can configure bonding using either your distro's network
Cong Wangb1098bb2013-05-27 15:49:16 +0000869initialization scripts, or manually using either iproute2 or the
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000870sysfs interface. Distros generally use one of three packages for the
871network initialization scripts: initscripts, sysconfig or interfaces.
872Recent versions of these packages have support for bonding, while older
Auke Kok6224e012006-06-08 11:15:35 -0700873versions do not.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
875 We will first describe the options for configuring bonding for
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000876distros using versions of initscripts, sysconfig and interfaces with full
877or partial support for bonding, then provide information on enabling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878bonding without support from the network initialization scripts (i.e.,
879older versions of initscripts or sysconfig).
880
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000881 If you're unsure whether your distro uses sysconfig,
882initscripts or interfaces, or don't know if it's new enough, have no fear.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883Determining this is fairly straightforward.
884
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +0000885 First, look for a file called interfaces in /etc/network directory.
886If this file is present in your system, then your system use interfaces. See
887Configuration with Interfaces Support.
888
889 Else, issue the command:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
891$ rpm -qf /sbin/ifup
892
893 It will respond with a line of text starting with either
894"initscripts" or "sysconfig," followed by some numbers. This is the
895package that provides your network initialization scripts.
896
897 Next, to determine if your installation supports bonding,
898issue the command:
899
900$ grep ifenslave /sbin/ifup
901
902 If this returns any matches, then your initscripts or
903sysconfig has support for bonding.
904
Auke Kok6224e012006-06-08 11:15:35 -07009053.1 Configuration with Sysconfig Support
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906----------------------------------------
907
908 This section applies to distros using a version of sysconfig
909with bonding support, for example, SuSE Linux Enterprise Server 9.
910
911 SuSE SLES 9's networking configuration system does support
912bonding, however, at this writing, the YaST system configuration
Auke Kok6224e012006-06-08 11:15:35 -0700913front end does not provide any means to work with bonding devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914Bonding devices can be managed by hand, however, as follows.
915
916 First, if they have not already been configured, configure the
917slave devices. On SLES 9, this is most easily done by running the
918yast2 sysconfig configuration utility. The goal is for to create an
919ifcfg-id file for each slave device. The simplest way to accomplish
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700920this is to configure the devices for DHCP (this is only to get the
921file ifcfg-id file created; see below for some issues with DHCP). The
922name of the configuration file for each device will be of the form:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923
924ifcfg-id-xx:xx:xx:xx:xx:xx
925
926 Where the "xx" portion will be replaced with the digits from
927the device's permanent MAC address.
928
929 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
930created, it is necessary to edit the configuration files for the slave
931devices (the MAC addresses correspond to those of the slave devices).
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700932Before editing, the file will contain multiple lines, and will look
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933something like this:
934
935BOOTPROTO='dhcp'
936STARTMODE='on'
937USERCTL='no'
938UNIQUE='XNzu.WeZGOGF+4wE'
939_nm_name='bus-pci-0001:61:01.0'
940
941 Change the BOOTPROTO and STARTMODE lines to the following:
942
943BOOTPROTO='none'
944STARTMODE='off'
945
946 Do not alter the UNIQUE or _nm_name lines. Remove any other
947lines (USERCTL, etc).
948
949 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
950it's time to create the configuration file for the bonding device
951itself. This file is named ifcfg-bondX, where X is the number of the
952bonding device to create, starting at 0. The first such file is
953ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
954network configuration system will correctly start multiple instances
955of bonding.
956
957 The contents of the ifcfg-bondX file is as follows:
958
959BOOTPROTO="static"
960BROADCAST="10.0.2.255"
961IPADDR="10.0.2.10"
962NETMASK="255.255.0.0"
963NETWORK="10.0.2.0"
964REMOTE_IPADDR=""
965STARTMODE="onboot"
966BONDING_MASTER="yes"
967BONDING_MODULE_OPTS="mode=active-backup miimon=100"
968BONDING_SLAVE0="eth0"
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700969BONDING_SLAVE1="bus-pci-0000:06:08.1"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970
971 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
972values with the appropriate values for your network.
973
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 The STARTMODE specifies when the device is brought online.
975The possible values are:
976
977 onboot: The device is started at boot time. If you're not
978 sure, this is probably what you want.
979
980 manual: The device is started only when ifup is called
981 manually. Bonding devices may be configured this
982 way if you do not wish them to start automatically
983 at boot for some reason.
984
985 hotplug: The device is started by a hotplug event. This is not
986 a valid choice for a bonding device.
987
988 off or ignore: The device configuration is ignored.
989
990 The line BONDING_MASTER='yes' indicates that the device is a
991bonding master device. The only useful value is "yes."
992
993 The contents of BONDING_MODULE_OPTS are supplied to the
994instance of the bonding module for this device. Specify the options
995for the bonding mode, link monitoring, and so on here. Do not include
996the max_bonds bonding parameter; this will confuse the configuration
997system if you have multiple bonding devices.
998
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700999 Finally, supply one BONDING_SLAVEn="slave device" for each
1000slave. where "n" is an increasing value, one for each slave. The
1001"slave device" is either an interface name, e.g., "eth0", or a device
1002specifier for the network device. The interface name is easier to
1003find, but the ethN names are subject to change at boot time if, e.g.,
1004a device early in the sequence has failed. The device specifiers
1005(bus-pci-0000:06:08.1 in the example above) specify the physical
1006network device, and will not change unless the device's bus location
1007changes (for example, it is moved from one PCI slot to another). The
1008example above uses one of each type for demonstration purposes; most
1009configurations will choose one or the other for all slave devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
1011 When all configuration files have been modified or created,
1012networking must be restarted for the configuration changes to take
1013effect. This can be accomplished via the following:
1014
1015# /etc/init.d/network restart
1016
1017 Note that the network control script (/sbin/ifdown) will
1018remove the bonding module as part of the network shutdown processing,
1019so it is not necessary to remove the module by hand if, e.g., the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001020module parameters have changed.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
1022 Also, at this writing, YaST/YaST2 will not manage bonding
1023devices (they do not show bonding interfaces on its list of network
1024devices). It is necessary to edit the configuration file by hand to
1025change the bonding configuration.
1026
1027 Additional general options and details of the ifcfg file
1028format can be found in an example ifcfg template file:
1029
1030/etc/sysconfig/network/ifcfg.template
1031
1032 Note that the template does not document the various BONDING_
1033settings described above, but does describe many of the other options.
1034
Auke Kok6224e012006-06-08 11:15:35 -070010353.1.1 Using DHCP with Sysconfig
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001036-------------------------------
1037
1038 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1039will cause it to query DHCP for its IP address information. At this
1040writing, this does not function for bonding devices; the scripts
1041attempt to obtain the device address from DHCP prior to adding any of
1042the slave devices. Without active slaves, the DHCP requests are not
1043sent to the network.
1044
Auke Kok6224e012006-06-08 11:15:35 -070010453.1.2 Configuring Multiple Bonds with Sysconfig
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001046-----------------------------------------------
1047
1048 The sysconfig network initialization system is capable of
1049handling multiple bonding devices. All that is necessary is for each
1050bonding instance to have an appropriately configured ifcfg-bondX file
1051(as described above). Do not specify the "max_bonds" parameter to any
1052instance of bonding, as this will confuse sysconfig. If you require
1053multiple bonding devices with identical parameters, create multiple
1054ifcfg-bondX files.
1055
1056 Because the sysconfig scripts supply the bonding module
1057options in the ifcfg-bondX file, it is not necessary to add them to
Lucas De Marchi970e2482012-03-30 13:37:16 -07001058the system /etc/modules.d/*.conf configuration files.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001059
Auke Kok6224e012006-06-08 11:15:35 -070010603.2 Configuration with Initscripts Support
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061------------------------------------------
1062
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001063 This section applies to distros using a recent version of
1064initscripts with bonding support, for example, Red Hat Enterprise Linux
1065version 3 or later, Fedora, etc. On these systems, the network
1066initialization scripts have knowledge of bonding, and can be configured to
1067control bonding devices. Note that older versions of the initscripts
1068package have lower levels of support for bonding; this will be noted where
1069applicable.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
1071 These distros will not automatically load the network adapter
1072driver unless the ethX device is configured with an IP address.
1073Because of this constraint, users must manually configure a
1074network-script file for all physical adapters that will be members of
1075a bondX link. Network script files are located in the directory:
1076
1077/etc/sysconfig/network-scripts
1078
1079 The file name must be prefixed with "ifcfg-eth" and suffixed
1080with the adapter's physical adapter number. For example, the script
1081for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1082Place the following text in the file:
1083
1084DEVICE=eth0
1085USERCTL=no
1086ONBOOT=yes
1087MASTER=bond0
1088SLAVE=yes
1089BOOTPROTO=none
1090
1091 The DEVICE= line will be different for every ethX device and
1092must correspond with the name of the file, i.e., ifcfg-eth1 must have
1093a device line of DEVICE=eth1. The setting of the MASTER= line will
1094also depend on the final bonding interface name chosen for your bond.
1095As with other network devices, these typically start at 0, and go up
1096one for each device, i.e., the first bonding instance is bond0, the
1097second is bond1, and so on.
1098
1099 Next, create a bond network script. The file name for this
1100script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1101the number of the bond. For bond0 the file is named "ifcfg-bond0",
1102for bond1 it is named "ifcfg-bond1", and so on. Within that file,
1103place the following text:
1104
1105DEVICE=bond0
1106IPADDR=192.168.1.1
1107NETMASK=255.255.255.0
1108NETWORK=192.168.1.0
1109BROADCAST=192.168.1.255
1110ONBOOT=yes
1111BOOTPROTO=none
1112USERCTL=no
1113
1114 Be sure to change the networking specific lines (IPADDR,
1115NETMASK, NETWORK and BROADCAST) to match your network configuration.
1116
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001117 For later versions of initscripts, such as that found with Fedora
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +000011187 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1119and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001120file, e.g. a line of the format:
1121
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +00001122BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001123
1124 will configure the bond with the specified options. The options
1125specified in BONDING_OPTS are identical to the bonding module parameters
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +00001126except for the arp_ip_target field when using versions of initscripts older
1127than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
1128using older versions each target should be included as a separate option and
1129should be preceded by a '+' to indicate it should be added to the list of
1130queried targets, e.g.,
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001131
1132 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1133
1134 is the proper syntax to specify multiple targets. When specifying
Lucas De Marchi970e2482012-03-30 13:37:16 -07001135options via BONDING_OPTS, it is not necessary to edit /etc/modprobe.d/*.conf.
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001136
Andy Gospodarek3f8b4b12008-10-22 11:19:48 +00001137 For even older versions of initscripts that do not support
Lucas De Marchi970e2482012-03-30 13:37:16 -07001138BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1139your distro) to load the bonding module with your desired options when the
1140bond0 interface is brought up. The following lines in /etc/modprobe.d/*.conf
1141will load the bonding module, and select its options:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142
1143alias bond0 bonding
1144options bond0 mode=balance-alb miimon=100
1145
1146 Replace the sample parameters with the appropriate set of
1147options for your configuration.
1148
1149 Finally run "/etc/rc.d/init.d/network restart" as root. This
1150will restart the networking subsystem and your bond link should be now
1151up and running.
1152
Auke Kok6224e012006-06-08 11:15:35 -070011533.2.1 Using DHCP with Initscripts
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001154---------------------------------
1155
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001156 Recent versions of initscripts (the versions supplied with Fedora
1157Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1158work) have support for assigning IP information to bonding devices via
1159DHCP.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001160
1161 To configure bonding for DHCP, configure it as described
1162above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1163and add a line consisting of "TYPE=Bonding". Note that the TYPE value
1164is case sensitive.
1165
Auke Kok6224e012006-06-08 11:15:35 -070011663.2.2 Configuring Multiple Bonds with Initscripts
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001167-------------------------------------------------
1168
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001169 Initscripts packages that are included with Fedora 7 and Red Hat
1170Enterprise Linux 5 support multiple bonding interfaces by simply
1171specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1172number of the bond. This support requires sysfs support in the kernel,
1173and a bonding driver of version 3.0.0 or later. Other configurations may
1174not support this method for specifying multiple bonding interfaces; for
1175those instances, see the "Configuring Multiple Bonds Manually" section,
1176below.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177
Cong Wangb1098bb2013-05-27 15:49:16 +000011783.3 Configuring Bonding Manually with iproute2
Auke Kok6224e012006-06-08 11:15:35 -07001179-----------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
1181 This section applies to distros whose network initialization
1182scripts (the sysconfig or initscripts package) do not have specific
1183knowledge of bonding. One such distro is SuSE Linux Enterprise Server
1184version 8.
1185
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001186 The general method for these systems is to place the bonding
Lucas De Marchi970e2482012-03-30 13:37:16 -07001187module parameters into a config file in /etc/modprobe.d/ (as
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001188appropriate for the installed distro), then add modprobe and/or
Cong Wangb1098bb2013-05-27 15:49:16 +00001189`ip link` commands to the system's global init script. The name of
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001190the global init script differs; for sysconfig, it is
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191/etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1192
1193 For example, if you wanted to make a simple bond of two e100
1194devices (presumed to be eth0 and eth1), and have it persist across
1195reboots, edit the appropriate file (/etc/init.d/boot.local or
1196/etc/rc.d/rc.local), and add the following:
1197
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001198modprobe bonding mode=balance-alb miimon=100
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199modprobe e100
1200ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
Cong Wangb1098bb2013-05-27 15:49:16 +00001201ip link set eth0 master bond0
1202ip link set eth1 master bond0
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
1204 Replace the example bonding module parameters and bond0
1205network configuration (IP address, netmask, etc) with the appropriate
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001206values for your configuration.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
1208 Unfortunately, this method will not provide support for the
1209ifup and ifdown scripts on the bond devices. To reload the bonding
1210configuration, it is necessary to run the initialization script, e.g.,
1211
1212# /etc/init.d/boot.local
1213
1214 or
1215
1216# /etc/rc.d/rc.local
1217
1218 It may be desirable in such a case to create a separate script
1219which only initializes the bonding configuration, then call that
1220separate script from within boot.local. This allows for bonding to be
1221enabled without re-running the entire global init script.
1222
1223 To shut down the bonding devices, it is necessary to first
1224mark the bonding device itself as being down, then remove the
1225appropriate device driver modules. For our example above, you can do
1226the following:
1227
1228# ifconfig bond0 down
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001229# rmmod bonding
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230# rmmod e100
1231
1232 Again, for convenience, it may be desirable to create a script
1233with these commands.
1234
1235
Jay Vosburgh00354cf2005-07-21 12:18:02 -070012363.3.1 Configuring Multiple Bonds Manually
1237-----------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238
1239 This section contains information on configuring multiple
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001240bonding devices with differing options for those systems whose network
1241initialization scripts lack support for configuring multiple bonds.
1242
1243 If you require multiple bonding devices, but all with the same
1244options, you may wish to use the "max_bonds" module parameter,
1245documented above.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001247 To create multiple bonding devices with differing options, it is
Rick Jonesf8b72d32012-07-20 10:51:37 +00001248preferable to use bonding parameters exported by sysfs, documented in the
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001249section below.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001251 For versions of bonding without sysfs support, the only means to
1252provide multiple instances of bonding with differing options is to load
1253the bonding driver multiple times. Note that current versions of the
1254sysconfig network initialization scripts handle this automatically; if
1255your distro uses these scripts, no special action is needed. See the
1256section Configuring Bonding Devices, above, if you're not sure about your
1257network initialization scripts.
1258
1259 To load multiple instances of the module, it is necessary to
1260specify a different name for each instance (the module loading system
1261requires that every loaded module, even multiple instances of the same
1262module, have a unique name). This is accomplished by supplying multiple
Lucas De Marchi970e2482012-03-30 13:37:16 -07001263sets of bonding options in /etc/modprobe.d/*.conf, for example:
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001264
1265alias bond0 bonding
1266options bond0 -o bond0 mode=balance-rr miimon=100
1267
1268alias bond1 bonding
1269options bond1 -o bond1 mode=balance-alb miimon=50
1270
1271 will load the bonding module two times. The first instance is
1272named "bond0" and creates the bond0 device in balance-rr mode with an
1273miimon of 100. The second instance is named "bond1" and creates the
1274bond1 device in balance-alb mode with an miimon of 50.
1275
1276 In some circumstances (typically with older distributions),
1277the above does not work, and the second bonding instance never sees
1278its options. In that case, the second options line can be substituted
1279as follows:
1280
1281install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1282 mode=balance-alb miimon=50
1283
1284 This may be repeated any number of times, specifying a new and
1285unique name in place of bond1 for each subsequent instance.
1286
1287 It has been observed that some Red Hat supplied kernels are unable
1288to rename modules at load time (the "-o bond1" part). Attempts to pass
1289that option to modprobe will produce an "Operation not permitted" error.
1290This has been reported on some Fedora Core kernels, and has been seen on
1291RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1292to configure multiple bonds with differing parameters (as they are older
1293kernels, and also lack sysfs support).
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001294
Auke Kok6224e012006-06-08 11:15:35 -070012953.4 Configuring Bonding Manually via Sysfs
1296------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001298 Starting with version 3.0.0, Channel Bonding may be configured
Auke Kok6224e012006-06-08 11:15:35 -07001299via the sysfs interface. This interface allows dynamic configuration
1300of all bonds in the system without unloading the module. It also
1301allows for adding and removing bonds at runtime. Ifenslave is no
1302longer required, though it is still supported.
1303
1304 Use of the sysfs interface allows you to use multiple bonds
1305with different configurations without having to reload the module.
1306It also allows you to use multiple, differently configured bonds when
1307bonding is compiled into the kernel.
1308
1309 You must have the sysfs filesystem mounted to configure
1310bonding this way. The examples in this document assume that you
1311are using the standard mount point for sysfs, e.g. /sys. If your
1312sysfs filesystem is mounted elsewhere, you will need to adjust the
1313example paths accordingly.
1314
1315Creating and Destroying Bonds
1316-----------------------------
1317To add a new bond foo:
1318# echo +foo > /sys/class/net/bonding_masters
1319
1320To remove an existing bond bar:
1321# echo -bar > /sys/class/net/bonding_masters
1322
1323To show all existing bonds:
1324# cat /sys/class/net/bonding_masters
1325
1326NOTE: due to 4K size limitation of sysfs files, this list may be
1327truncated if you have more than a few hundred bonds. This is unlikely
1328to occur under normal operating conditions.
1329
1330Adding and Removing Slaves
1331--------------------------
1332 Interfaces may be enslaved to a bond using the file
1333/sys/class/net/<bond>/bonding/slaves. The semantics for this file
1334are the same as for the bonding_masters file.
1335
1336To enslave interface eth0 to bond bond0:
1337# ifconfig bond0 up
1338# echo +eth0 > /sys/class/net/bond0/bonding/slaves
1339
1340To free slave eth0 from bond bond0:
1341# echo -eth0 > /sys/class/net/bond0/bonding/slaves
1342
Auke Kok6224e012006-06-08 11:15:35 -07001343 When an interface is enslaved to a bond, symlinks between the
1344two are created in the sysfs filesystem. In this case, you would get
1345/sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1346/sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1347
1348 This means that you can tell quickly whether or not an
1349interface is enslaved by looking for the master symlink. Thus:
1350# echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1351will free eth0 from whatever bond it is enslaved to, regardless of
1352the name of the bond interface.
1353
1354Changing a Bond's Configuration
1355-------------------------------
1356 Each bond may be configured individually by manipulating the
1357files located in /sys/class/net/<bond name>/bonding
1358
1359 The names of these files correspond directly with the command-
Paolo Ornati670e9f32006-10-03 22:57:56 +02001360line parameters described elsewhere in this file, and, with the
Auke Kok6224e012006-06-08 11:15:35 -07001361exception of arp_ip_target, they accept the same values. To see the
1362current setting, simply cat the appropriate file.
1363
1364 A few examples will be given here; for specific usage
1365guidelines for each parameter, see the appropriate section in this
1366document.
1367
1368To configure bond0 for balance-alb mode:
1369# ifconfig bond0 down
1370# echo 6 > /sys/class/net/bond0/bonding/mode
1371 - or -
1372# echo balance-alb > /sys/class/net/bond0/bonding/mode
1373 NOTE: The bond interface must be down before the mode can be
1374changed.
1375
1376To enable MII monitoring on bond0 with a 1 second interval:
1377# echo 1000 > /sys/class/net/bond0/bonding/miimon
1378 NOTE: If ARP monitoring is enabled, it will disabled when MII
1379monitoring is enabled, and vice-versa.
1380
1381To add ARP targets:
1382# echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1383# echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
Brian Haley5a31bec2009-04-13 00:11:30 -07001384 NOTE: up to 16 target addresses may be specified.
Auke Kok6224e012006-06-08 11:15:35 -07001385
1386To remove an ARP target:
1387# echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1388
Neil Horman7eacd032013-09-13 11:05:33 -04001389To configure the interval between learning packet transmits:
1390# echo 12 > /sys/class/net/bond0/bonding/lp_interval
1391 NOTE: the lp_inteval is the number of seconds between instances where
1392the bonding driver sends learning packets to each slaves peer switch. The
1393default interval is 1 second.
1394
Auke Kok6224e012006-06-08 11:15:35 -07001395Example Configuration
1396---------------------
1397 We begin with the same example that is shown in section 3.3,
1398executed with sysfs, and without using ifenslave.
1399
1400 To make a simple bond of two e100 devices (presumed to be eth0
1401and eth1), and have it persist across reboots, edit the appropriate
1402file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1403following:
1404
1405modprobe bonding
1406modprobe e100
1407echo balance-alb > /sys/class/net/bond0/bonding/mode
1408ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1409echo 100 > /sys/class/net/bond0/bonding/miimon
1410echo +eth0 > /sys/class/net/bond0/bonding/slaves
1411echo +eth1 > /sys/class/net/bond0/bonding/slaves
1412
1413 To add a second bond, with two e1000 interfaces in
1414active-backup mode, using ARP monitoring, add the following lines to
1415your init script:
1416
1417modprobe e1000
1418echo +bond1 > /sys/class/net/bonding_masters
1419echo active-backup > /sys/class/net/bond1/bonding/mode
1420ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1421echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1422echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1423echo +eth2 > /sys/class/net/bond1/bonding/slaves
1424echo +eth3 > /sys/class/net/bond1/bonding/slaves
1425
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +000014263.5 Configuration with Interfaces Support
1427-----------------------------------------
1428
1429 This section applies to distros which use /etc/network/interfaces file
1430to describe network interface configuration, most notably Debian and it's
1431derivatives.
1432
1433 The ifup and ifdown commands on Debian don't support bonding out of
1434the box. The ifenslave-2.6 package should be installed to provide bonding
1435support. Once installed, this package will provide bond-* options to be used
1436into /etc/network/interfaces.
1437
1438 Note that ifenslave-2.6 package will load the bonding module and use
1439the ifenslave command when appropriate.
1440
1441Example Configurations
1442----------------------
1443
1444In /etc/network/interfaces, the following stanza will configure bond0, in
1445active-backup mode, with eth0 and eth1 as slaves.
1446
1447auto bond0
1448iface bond0 inet dhcp
1449 bond-slaves eth0 eth1
1450 bond-mode active-backup
1451 bond-miimon 100
1452 bond-primary eth0 eth1
1453
1454If the above configuration doesn't work, you might have a system using
1455upstart for system startup. This is most notably true for recent
1456Ubuntu versions. The following stanza in /etc/network/interfaces will
1457produce the same result on those systems.
1458
1459auto bond0
1460iface bond0 inet dhcp
1461 bond-slaves none
1462 bond-mode active-backup
1463 bond-miimon 100
1464
1465auto eth0
1466iface eth0 inet manual
1467 bond-master bond0
1468 bond-primary eth0 eth1
1469
1470auto eth1
1471iface eth1 inet manual
1472 bond-master bond0
1473 bond-primary eth0 eth1
1474
1475For a full list of bond-* supported options in /etc/network/interfaces and some
1476more advanced examples tailored to you particular distros, see the files in
1477/usr/share/doc/ifenslave-2.6.
1478
14793.6 Overriding Configuration for Special Cases
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00001480----------------------------------------------
Nicolas de Pesloüande221bd2011-01-24 13:21:37 +00001481
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00001482When using the bonding driver, the physical port which transmits a frame is
1483typically selected by the bonding driver, and is not relevant to the user or
1484system administrator. The output port is simply selected using the policies of
1485the selected bonding mode. On occasion however, it is helpful to direct certain
1486classes of traffic to certain physical interfaces on output to implement
1487slightly more complex policies. For example, to reach a web server over a
1488bonded interface in which eth0 connects to a private network, while eth1
1489connects via a public network, it may be desirous to bias the bond to send said
1490traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1491can safely be sent over either interface. Such configurations may be achieved
1492using the traffic control utilities inherent in linux.
Auke Kok6224e012006-06-08 11:15:35 -07001493
Andy Gospodarekbb1d9122010-06-02 08:40:18 +00001494By default the bonding driver is multiqueue aware and 16 queues are created
1495when the driver initializes (see Documentation/networking/multiqueue.txt
1496for details). If more or less queues are desired the module parameter
1497tx_queues can be used to change this value. There is no sysfs parameter
1498available as the allocation is done at module init time.
1499
1500The output of the file /proc/net/bonding/bondX has changed so the output Queue
1501ID is now printed for each slave:
1502
1503Bonding Mode: fault-tolerance (active-backup)
1504Primary Slave: None
1505Currently Active Slave: eth0
1506MII Status: up
1507MII Polling Interval (ms): 0
1508Up Delay (ms): 0
1509Down Delay (ms): 0
1510
1511Slave Interface: eth0
1512MII Status: up
1513Link Failure Count: 0
1514Permanent HW addr: 00:1a:a0:12:8f:cb
1515Slave queue ID: 0
1516
1517Slave Interface: eth1
1518MII Status: up
1519Link Failure Count: 0
1520Permanent HW addr: 00:1a:a0:12:8f:cc
1521Slave queue ID: 2
1522
1523The queue_id for a slave can be set using the command:
1524
1525# echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1526
1527Any interface that needs a queue_id set should set it with multiple calls
1528like the one above until proper priorities are set for all interfaces. On
1529distributions that allow configuration via initscripts, multiple 'queue_id'
1530arguments can be added to BONDING_OPTS to set all needed slave queues.
1531
1532These queue id's can be used in conjunction with the tc utility to configure
1533a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1534slave devices. For instance, say we wanted, in the above configuration to
1535force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1536device. The following commands would accomplish this:
1537
1538# tc qdisc add dev bond0 handle 1 root multiq
1539
1540# tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip dst \
1541 192.168.1.100 action skbedit queue_mapping 2
1542
1543These commands tell the kernel to attach a multiqueue queue discipline to the
1544bond0 interface and filter traffic enqueued to it, such that packets with a dst
1545ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1546This value is then passed into the driver, causing the normal output path
1547selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1548
1549Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver
1550that normal output policy selection should take place. One benefit to simply
1551leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1552driver that is now present. This awareness allows tc filters to be placed on
1553slave devices as well as bond devices and the bonding driver will simply act as
1554a pass-through for selecting output queues on the slave device rather than
1555output port selection.
1556
1557This feature first appeared in bonding driver version 3.7.0 and support for
1558output slave selection was limited to round-robin and active-backup modes.
1559
15604 Querying Bonding Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561=================================
1562
Auke Kok6224e012006-06-08 11:15:35 -070015634.1 Bonding Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564-------------------------
1565
1566 Each bonding device has a read-only file residing in the
1567/proc/net/bonding directory. The file contents include information
1568about the bonding configuration, options and state of each slave.
1569
1570 For example, the contents of /proc/net/bonding/bond0 after the
1571driver is loaded with parameters of mode=0 and miimon=1000 is
1572generally as follows:
1573
1574 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1575 Bonding Mode: load balancing (round-robin)
1576 Currently Active Slave: eth0
1577 MII Status: up
1578 MII Polling Interval (ms): 1000
1579 Up Delay (ms): 0
1580 Down Delay (ms): 0
1581
1582 Slave Interface: eth1
1583 MII Status: up
1584 Link Failure Count: 1
1585
1586 Slave Interface: eth0
1587 MII Status: up
1588 Link Failure Count: 1
1589
1590 The precise format and contents will change depending upon the
1591bonding configuration, state, and version of the bonding driver.
1592
Auke Kok6224e012006-06-08 11:15:35 -070015934.2 Network configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594-------------------------
1595
1596 The network configuration can be inspected using the ifconfig
1597command. Bonding devices will have the MASTER flag set; Bonding slave
1598devices will have the SLAVE flag set. The ifconfig output does not
1599contain information on which slaves are associated with which masters.
1600
1601 In the example below, the bond0 interface is the master
1602(MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1603bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1604TLB and ALB that require a unique MAC address for each slave.
1605
1606# /sbin/ifconfig
1607bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1608 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1609 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1610 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1611 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1612 collisions:0 txqueuelen:0
1613
1614eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1616 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1617 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1618 collisions:0 txqueuelen:100
1619 Interrupt:10 Base address:0x1080
1620
1621eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1623 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1624 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1625 collisions:0 txqueuelen:100
1626 Interrupt:9 Base address:0x1400
1627
Auke Kok6224e012006-06-08 11:15:35 -070016285. Switch Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629=======================
1630
1631 For this section, "switch" refers to whatever system the
1632bonded devices are directly connected to (i.e., where the other end of
1633the cable plugs into). This may be an actual dedicated switch device,
1634or it may be another regular system (e.g., another computer running
1635Linux),
1636
1637 The active-backup, balance-tlb and balance-alb modes do not
1638require any specific configuration of the switch.
1639
1640 The 802.3ad mode requires that the switch have the appropriate
1641ports configured as an 802.3ad aggregation. The precise method used
1642to configure this varies from switch to switch, but, for example, a
1643Cisco 3550 series switch requires that the appropriate ports first be
1644grouped together in a single etherchannel instance, then that
1645etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1646standard EtherChannel).
1647
1648 The balance-rr, balance-xor and broadcast modes generally
1649require that the switch have the appropriate ports grouped together.
1650The nomenclature for such a group differs between switches, it may be
1651called an "etherchannel" (as in the Cisco example, above), a "trunk
1652group" or some other similar variation. For these modes, each switch
1653will also have its own configuration options for the switch's transmit
1654policy to the bond. Typical choices include XOR of either the MAC or
1655IP addresses. The transmit policy of the two peers does not need to
1656match. For these three modes, the bonding mode really selects a
1657transmit policy for an EtherChannel group; all three will interoperate
1658with another EtherChannel group.
1659
1660
Auke Kok6224e012006-06-08 11:15:35 -070016616. 802.1q VLAN Support
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662======================
1663
1664 It is possible to configure VLAN devices over a bond interface
1665using the 8021q driver. However, only packets coming from the 8021q
1666driver and passing through bonding will be tagged by default. Self
1667generated packets, for example, bonding's learning packets or ARP
1668packets generated by either ALB mode or the ARP monitor mechanism, are
1669tagged internally by bonding itself. As a result, bonding must
1670"learn" the VLAN IDs configured above it, and use those IDs to tag
1671self generated packets.
1672
1673 For reasons of simplicity, and to support the use of adapters
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001674that can do VLAN hardware acceleration offloading, the bonding
1675interface declares itself as fully hardware offloading capable, it gets
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676the add_vid/kill_vid notifications to gather the necessary
1677information, and it propagates those actions to the slaves. In case
1678of mixed adapter types, hardware accelerated tagged packets that
1679should go through an adapter that is not offloading capable are
1680"un-accelerated" by the bonding driver so the VLAN tag sits in the
1681regular location.
1682
1683 VLAN interfaces *must* be added on top of a bonding interface
1684only after enslaving at least one slave. The bonding interface has a
1685hardware address of 00:00:00:00:00:00 until the first slave is added.
1686If the VLAN interface is created prior to the first enslavement, it
1687would pick up the all-zeroes hardware address. Once the first slave
1688is attached to the bond, the bond device itself will pick up the
1689slave's hardware address, which is then available for the VLAN device.
1690
1691 Also, be aware that a similar problem can occur if all slaves
1692are released from a bond that still has one or more VLAN interfaces on
1693top of it. When a new slave is added, the bonding interface will
1694obtain its hardware address from the first slave, which might not
1695match the hardware address of the VLAN interfaces (which was
1696ultimately copied from an earlier slave).
1697
1698 There are two methods to insure that the VLAN device operates
1699with the correct hardware address if all slaves are removed from a
1700bond interface:
1701
1702 1. Remove all VLAN interfaces then recreate them
1703
1704 2. Set the bonding interface's hardware address so that it
1705matches the hardware address of the VLAN interfaces.
1706
1707 Note that changing a VLAN interface's HW address would set the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001708underlying device -- i.e. the bonding interface -- to promiscuous
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709mode, which might not be what you want.
1710
1711
Auke Kok6224e012006-06-08 11:15:35 -070017127. Link Monitoring
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713==================
1714
1715 The bonding driver at present supports two schemes for
1716monitoring a slave device's link state: the ARP monitor and the MII
1717monitor.
1718
1719 At the present time, due to implementation restrictions in the
1720bonding driver itself, it is not possible to enable both ARP and MII
1721monitoring simultaneously.
1722
Auke Kok6224e012006-06-08 11:15:35 -070017237.1 ARP Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724-------------------------
1725
1726 The ARP monitor operates as its name suggests: it sends ARP
1727queries to one or more designated peer systems on the network, and
1728uses the response as an indication that the link is operating. This
1729gives some assurance that traffic is actually flowing to and from one
1730or more peers on the local network.
1731
1732 The ARP monitor relies on the device driver itself to verify
1733that traffic is flowing. In particular, the driver must keep up to
1734date the last receive time, dev->last_rx, and transmit start time,
1735dev->trans_start. If these are not updated by the driver, then the
1736ARP monitor will immediately fail any slaves using that driver, and
1737those slaves will stay down. If networking monitoring (tcpdump, etc)
1738shows the ARP requests and replies on the network, then it may be that
1739your device driver is not updating last_rx and trans_start.
1740
Auke Kok6224e012006-06-08 11:15:35 -070017417.2 Configuring Multiple ARP Targets
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742------------------------------------
1743
1744 While ARP monitoring can be done with just one target, it can
1745be useful in a High Availability setup to have several targets to
1746monitor. In the case of just one target, the target itself may go
1747down or have a problem making it unresponsive to ARP requests. Having
1748an additional target (or several) increases the reliability of the ARP
1749monitoring.
1750
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001751 Multiple ARP targets must be separated by commas as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752
1753# example options for ARP monitoring with three targets
1754alias bond0 bonding
1755options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1756
1757 For just a single target the options would resemble:
1758
1759# example options for ARP monitoring with one target
1760alias bond0 bonding
1761options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1762
1763
Auke Kok6224e012006-06-08 11:15:35 -070017647.3 MII Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765-------------------------
1766
1767 The MII monitor monitors only the carrier state of the local
1768network interface. It accomplishes this in one of three ways: by
1769depending upon the device driver to maintain its carrier state, by
1770querying the device's MII registers, or by making an ethtool query to
1771the device.
1772
1773 If the use_carrier module parameter is 1 (the default value),
1774then the MII monitor will rely on the driver for carrier state
1775information (via the netif_carrier subsystem). As explained in the
1776use_carrier parameter information, above, if the MII monitor fails to
1777detect carrier loss on the device (e.g., when the cable is physically
1778disconnected), it may be that the driver does not support
1779netif_carrier.
1780
1781 If use_carrier is 0, then the MII monitor will first query the
1782device's (via ioctl) MII registers and check the link state. If that
1783request fails (not just that it returns carrier down), then the MII
1784monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1785the same information. If both methods fail (i.e., the driver either
1786does not support or had some error in processing both the MII register
1787and ethtool requests), then the MII monitor will assume the link is
1788up.
1789
Auke Kok6224e012006-06-08 11:15:35 -070017908. Potential Sources of Trouble
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791===============================
1792
Auke Kok6224e012006-06-08 11:15:35 -070017938.1 Adventures in Routing
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794-------------------------
1795
1796 When bonding is configured, it is important that the slave
Auke Kok6224e012006-06-08 11:15:35 -07001797devices not have routes that supersede routes of the master (or,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798generally, not have routes at all). For example, suppose the bonding
1799device bond0 has two slaves, eth0 and eth1, and the routing table is
1800as follows:
1801
1802Kernel IP routing table
1803Destination Gateway Genmask Flags MSS Window irtt Iface
180410.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
180510.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
180610.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
1807127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
1808
1809 This routing configuration will likely still update the
1810receive/transmit times in the driver (needed by the ARP monitor), but
1811may bypass the bonding driver (because outgoing traffic to, in this
1812case, another host on network 10 would use eth0 or eth1 before bond0).
1813
1814 The ARP monitor (and ARP itself) may become confused by this
1815configuration, because ARP requests (generated by the ARP monitor)
1816will be sent on one interface (bond0), but the corresponding reply
1817will arrive on a different interface (eth0). This reply looks to ARP
1818as an unsolicited ARP reply (because ARP matches replies on an
1819interface basis), and is discarded. The MII monitor is not affected
1820by the state of the routing table.
1821
1822 The solution here is simply to insure that slaves do not have
1823routes of their own, and if for some reason they must, those routes do
Auke Kok6224e012006-06-08 11:15:35 -07001824not supersede routes of their master. This should generally be the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825case, but unusual configurations or errant manual or automatic static
1826route additions may cause trouble.
1827
Auke Kok6224e012006-06-08 11:15:35 -070018288.2 Ethernet Device Renaming
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829----------------------------
1830
1831 On systems with network configuration scripts that do not
1832associate physical devices directly with network interface names (so
1833that the same physical device always has the same "ethX" name), it may
Lucas De Marchi970e2482012-03-30 13:37:16 -07001834be necessary to add some special logic to config files in
1835/etc/modprobe.d/.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836
1837 For example, given a modules.conf containing the following:
1838
1839alias bond0 bonding
1840options bond0 mode=some-mode miimon=50
1841alias eth0 tg3
1842alias eth1 tg3
1843alias eth2 e1000
1844alias eth3 e1000
1845
1846 If neither eth0 and eth1 are slaves to bond0, then when the
1847bond0 interface comes up, the devices may end up reordered. This
1848happens because bonding is loaded first, then its slave device's
1849drivers are loaded next. Since no other drivers have been loaded,
1850when the e1000 driver loads, it will receive eth0 and eth1 for its
1851devices, but the bonding configuration tries to enslave eth2 and eth3
1852(which may later be assigned to the tg3 devices).
1853
1854 Adding the following:
1855
1856add above bonding e1000 tg3
1857
1858 causes modprobe to load e1000 then tg3, in that order, when
1859bonding is loaded. This command is fully documented in the
1860modules.conf manual page.
1861
Lucas De Marchi970e2482012-03-30 13:37:16 -07001862 On systems utilizing modprobe an equivalent problem can occur.
1863In this case, the following can be added to config files in
1864/etc/modprobe.d/ as:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865
Lucas De Marchi78286cd2012-03-30 13:37:20 -07001866softdep bonding pre: tg3 e1000
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867
Lucas De Marchi970e2482012-03-30 13:37:16 -07001868 This will load tg3 and e1000 modules before loading the bonding one.
1869Full documentation on this can be found in the modprobe.d and modprobe
1870manual pages.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871
Auke Kok6224e012006-06-08 11:15:35 -070018728.3. Painfully Slow Or No Failed Link Detection By Miimon
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873---------------------------------------------------------
1874
1875 By default, bonding enables the use_carrier option, which
1876instructs bonding to trust the driver to maintain carrier state.
1877
1878 As discussed in the options section, above, some drivers do
1879not support the netif_carrier_on/_off link state tracking system.
1880With use_carrier enabled, bonding will always see these links as up,
1881regardless of their actual state.
1882
1883 Additionally, other drivers do support netif_carrier, but do
1884not maintain it in real time, e.g., only polling the link state at
1885some fixed interval. In this case, miimon will detect failures, but
1886only after some long period of time has expired. If it appears that
1887miimon is very slow in detecting link failures, try specifying
1888use_carrier=0 to see if that improves the failure detection time. If
1889it does, then it may be that the driver checks the carrier state at a
1890fixed interval, but does not cache the MII register values (so the
1891use_carrier=0 method of querying the registers directly works). If
1892use_carrier=0 does not improve the failover, then the driver may cache
1893the registers, or the problem may be elsewhere.
1894
1895 Also, remember that miimon only checks for the device's
1896carrier state. It has no way to determine the state of devices on or
1897beyond other ports of a switch, or if a switch is refusing to pass
1898traffic while still maintaining carrier on.
1899
Auke Kok6224e012006-06-08 11:15:35 -070019009. SNMP agents
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901===============
1902
1903 If running SNMP agents, the bonding driver should be loaded
1904before any network drivers participating in a bond. This requirement
Tobias Klauserd533f672005-09-10 00:26:46 -07001905is due to the interface index (ipAdEntIfIndex) being associated to
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906the first interface found with a given IP address. That is, there is
1907only one ipAdEntIfIndex for each IP address. For example, if eth0 and
1908eth1 are slaves of bond0 and the driver for eth0 is loaded before the
1909bonding driver, the interface for the IP address will be associated
1910with the eth0 interface. This configuration is shown below, the IP
1911address 192.168.1.1 has an interface index of 2 which indexes to eth0
1912in the ifDescr table (ifDescr.2).
1913
1914 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1915 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
1916 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
1917 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
1918 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
1919 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
1920 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
1921 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1922 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
1923 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1924
1925 This problem is avoided by loading the bonding driver before
1926any network drivers participating in a bond. Below is an example of
1927loading the bonding driver first, the IP address 192.168.1.1 is
1928correctly associated with ifDescr.2.
1929
1930 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1931 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
1932 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
1933 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
1934 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
1935 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
1936 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
1937 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1938 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
1939 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1940
1941 While some distributions may not report the interface name in
1942ifDescr, the association between the IP address and IfIndex remains
1943and SNMP functions such as Interface_Scan_Next will report that
1944association.
1945
Auke Kok6224e012006-06-08 11:15:35 -0700194610. Promiscuous mode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947====================
1948
1949 When running network monitoring tools, e.g., tcpdump, it is
1950common to enable promiscuous mode on the device, so that all traffic
1951is seen (instead of seeing only traffic destined for the local host).
1952The bonding driver handles promiscuous mode changes to the bonding
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001953master device (e.g., bond0), and propagates the setting to the slave
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954devices.
1955
1956 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001957the promiscuous mode setting is propagated to all slaves.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958
1959 For the active-backup, balance-tlb and balance-alb modes, the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001960promiscuous mode setting is propagated only to the active slave.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961
1962 For balance-tlb mode, the active slave is the slave currently
1963receiving inbound traffic.
1964
1965 For balance-alb mode, the active slave is the slave used as a
1966"primary." This slave is used for mode-specific control traffic, for
1967sending to peers that are unassigned or if the load is unbalanced.
1968
1969 For the active-backup, balance-tlb and balance-alb modes, when
1970the active slave changes (e.g., due to a link failure), the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001971promiscuous setting will be propagated to the new active slave.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972
Auke Kok6224e012006-06-08 11:15:35 -0700197311. Configuring Bonding for High Availability
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001974=============================================
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975
1976 High Availability refers to configurations that provide
1977maximum network availability by having redundant or backup devices,
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001978links or switches between the host and the rest of the world. The
1979goal is to provide the maximum availability of network connectivity
1980(i.e., the network always works), even though other configurations
1981could provide higher throughput.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982
Auke Kok6224e012006-06-08 11:15:35 -0700198311.1 High Availability in a Single Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984--------------------------------------------------
1985
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001986 If two hosts (or a host and a single switch) are directly
1987connected via multiple physical links, then there is no availability
1988penalty to optimizing for maximum bandwidth. In this case, there is
1989only one switch (or peer), so if it fails, there is no alternative
1990access to fail over to. Additionally, the bonding load balance modes
1991support link monitoring of their members, so if individual links fail,
1992the load will be rebalanced across the remaining devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993
Rick Jonesf8b72d32012-07-20 10:51:37 +00001994 See Section 12, "Configuring Bonding for Maximum Throughput"
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001995for information on configuring bonding with one peer device.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996
Auke Kok6224e012006-06-08 11:15:35 -0700199711.2 High Availability in a Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998----------------------------------------------------
1999
2000 With multiple switches, the configuration of bonding and the
2001network changes dramatically. In multiple switch topologies, there is
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002002a trade off between network availability and usable bandwidth.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003
2004 Below is a sample network, configured to maximize the
2005availability of the network:
2006
2007 | |
2008 |port3 port3|
2009 +-----+----+ +-----+----+
2010 | |port2 ISL port2| |
2011 | switch A +--------------------------+ switch B |
2012 | | | |
2013 +-----+----+ +-----++---+
2014 |port1 port1|
2015 | +-------+ |
2016 +-------------+ host1 +---------------+
2017 eth0 +-------+ eth1
2018
2019 In this configuration, there is a link between the two
2020switches (ISL, or inter switch link), and multiple ports connecting to
2021the outside world ("port3" on each switch). There is no technical
2022reason that this could not be extended to a third switch.
2023
Auke Kok6224e012006-06-08 11:15:35 -0700202411.2.1 HA Bonding Mode Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002025-------------------------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002027 In a topology such as the example above, the active-backup and
2028broadcast modes are the only useful bonding modes when optimizing for
2029availability; the other modes require all links to terminate on the
2030same peer for them to behave rationally.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031
2032active-backup: This is generally the preferred mode, particularly if
2033 the switches have an ISL and play together well. If the
2034 network configuration is such that one switch is specifically
2035 a backup switch (e.g., has lower capacity, higher cost, etc),
2036 then the primary option can be used to insure that the
2037 preferred link is always used when it is available.
2038
2039broadcast: This mode is really a special purpose mode, and is suitable
2040 only for very specific needs. For example, if the two
2041 switches are not connected (no ISL), and the networks beyond
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002042 them are totally independent. In this case, if it is
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 necessary for some specific one-way traffic to reach both
2044 independent networks, then the broadcast mode may be suitable.
2045
Auke Kok6224e012006-06-08 11:15:35 -0700204611.2.2 HA Link Monitoring Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002047----------------------------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048
2049 The choice of link monitoring ultimately depends upon your
2050switch. If the switch can reliably fail ports in response to other
2051failures, then either the MII or ARP monitors should work. For
2052example, in the above example, if the "port3" link fails at the remote
2053end, the MII monitor has no direct means to detect this. The ARP
2054monitor could be configured with a target at the remote end of port3,
2055thus detecting that failure without switch support.
2056
2057 In general, however, in a multiple switch topology, the ARP
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002058monitor can provide a higher level of reliability in detecting end to
2059end connectivity failures (which may be caused by the failure of any
2060individual component to pass traffic for any reason). Additionally,
2061the ARP monitor should be configured with multiple targets (at least
2062one for each switch in the network). This will insure that,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063regardless of which switch is active, the ARP monitor has a suitable
2064target to query.
2065
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002066 Note, also, that of late many switches now support a functionality
2067generally referred to as "trunk failover." This is a feature of the
2068switch that causes the link state of a particular switch port to be set
2069down (or up) when the state of another switch port goes down (or up).
Matt LaPlante19f59462009-04-27 15:06:31 +02002070Its purpose is to propagate link failures from logically "exterior" ports
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002071to the logically "interior" ports that bonding is able to monitor via
2072miimon. Availability and configuration for trunk failover varies by
2073switch, but this can be a viable alternative to the ARP monitor when using
2074suitable switches.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075
Auke Kok6224e012006-06-08 11:15:35 -0700207612. Configuring Bonding for Maximum Throughput
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002077==============================================
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078
Auke Kok6224e012006-06-08 11:15:35 -0700207912.1 Maximizing Throughput in a Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002080------------------------------------------------------
2081
2082 In a single switch configuration, the best method to maximize
2083throughput depends upon the application and network environment. The
2084various load balancing modes each have strengths and weaknesses in
2085different environments, as detailed below.
2086
2087 For this discussion, we will break down the topologies into
2088two categories. Depending upon the destination of most traffic, we
2089categorize them into either "gatewayed" or "local" configurations.
2090
2091 In a gatewayed configuration, the "switch" is acting primarily
2092as a router, and the majority of traffic passes through this router to
2093other networks. An example would be the following:
2094
2095
2096 +----------+ +----------+
2097 | |eth0 port1| | to other networks
2098 | Host A +---------------------+ router +------------------->
2099 | +---------------------+ | Hosts B and C are out
2100 | |eth1 port2| | here somewhere
2101 +----------+ +----------+
2102
2103 The router may be a dedicated router device, or another host
2104acting as a gateway. For our discussion, the important point is that
2105the majority of traffic from Host A will pass through the router to
2106some other network before reaching its final destination.
2107
2108 In a gatewayed network configuration, although Host A may
2109communicate with many other systems, all of its traffic will be sent
2110and received via one other peer on the local network, the router.
2111
2112 Note that the case of two systems connected directly via
2113multiple physical links is, for purposes of configuring bonding, the
2114same as a gatewayed configuration. In that case, it happens that all
2115traffic is destined for the "gateway" itself, not some other network
2116beyond the gateway.
2117
2118 In a local configuration, the "switch" is acting primarily as
2119a switch, and the majority of traffic passes through this switch to
2120reach other stations on the same network. An example would be the
2121following:
2122
2123 +----------+ +----------+ +--------+
2124 | |eth0 port1| +-------+ Host B |
2125 | Host A +------------+ switch |port3 +--------+
2126 | +------------+ | +--------+
2127 | |eth1 port2| +------------------+ Host C |
2128 +----------+ +----------+port4 +--------+
2129
2130
2131 Again, the switch may be a dedicated switch device, or another
2132host acting as a gateway. For our discussion, the important point is
2133that the majority of traffic from Host A is destined for other hosts
2134on the same local network (Hosts B and C in the above example).
2135
2136 In summary, in a gatewayed configuration, traffic to and from
2137the bonded device will be to the same MAC level peer on the network
2138(the gateway itself, i.e., the router), regardless of its final
2139destination. In a local configuration, traffic flows directly to and
2140from the final destinations, thus, each destination (Host B, Host C)
2141will be addressed directly by their individual MAC addresses.
2142
2143 This distinction between a gatewayed and a local network
2144configuration is important because many of the load balancing modes
2145available use the MAC addresses of the local network source and
2146destination to make load balancing decisions. The behavior of each
2147mode is described below.
2148
2149
Auke Kok6224e012006-06-08 11:15:35 -0700215012.1.1 MT Bonding Mode Selection for Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002151-----------------------------------------------------------
2152
2153 This configuration is the easiest to set up and to understand,
2154although you will have to decide which bonding mode best suits your
2155needs. The trade offs for each mode are detailed below:
2156
2157balance-rr: This mode is the only mode that will permit a single
2158 TCP/IP connection to stripe traffic across multiple
2159 interfaces. It is therefore the only mode that will allow a
2160 single TCP/IP stream to utilize more than one interface's
2161 worth of throughput. This comes at a cost, however: the
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002162 striping generally results in peer systems receiving packets out
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002163 of order, causing TCP/IP's congestion control system to kick
2164 in, often by retransmitting segments.
2165
2166 It is possible to adjust TCP/IP's congestion limits by
2167 altering the net.ipv4.tcp_reordering sysctl parameter. The
2168 usual default value is 3, and the maximum useful value is 127.
2169 For a four interface balance-rr bond, expect that a single
2170 TCP/IP stream will utilize no more than approximately 2.3
2171 interface's worth of throughput, even after adjusting
2172 tcp_reordering.
2173
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002174 Note that the fraction of packets that will be delivered out of
2175 order is highly variable, and is unlikely to be zero. The level
2176 of reordering depends upon a variety of factors, including the
2177 networking interfaces, the switch, and the topology of the
2178 configuration. Speaking in general terms, higher speed network
2179 cards produce more reordering (due to factors such as packet
2180 coalescing), and a "many to many" topology will reorder at a
2181 higher rate than a "many slow to one fast" configuration.
2182
2183 Many switches do not support any modes that stripe traffic
2184 (instead choosing a port based upon IP or MAC level addresses);
2185 for those devices, traffic for a particular connection flowing
2186 through the switch to a balance-rr bond will not utilize greater
2187 than one interface's worth of bandwidth.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002188
2189 If you are utilizing protocols other than TCP/IP, UDP for
2190 example, and your application can tolerate out of order
2191 delivery, then this mode can allow for single stream datagram
2192 performance that scales near linearly as interfaces are added
2193 to the bond.
2194
2195 This mode requires the switch to have the appropriate ports
2196 configured for "etherchannel" or "trunking."
2197
2198active-backup: There is not much advantage in this network topology to
2199 the active-backup mode, as the inactive backup devices are all
2200 connected to the same peer as the primary. In this case, a
2201 load balancing mode (with link monitoring) will provide the
2202 same level of network availability, but with increased
2203 available bandwidth. On the plus side, active-backup mode
2204 does not require any configuration of the switch, so it may
2205 have value if the hardware available does not support any of
2206 the load balance modes.
2207
2208balance-xor: This mode will limit traffic such that packets destined
2209 for specific peers will always be sent over the same
2210 interface. Since the destination is determined by the MAC
2211 addresses involved, this mode works best in a "local" network
2212 configuration (as described above), with destinations all on
2213 the same local network. This mode is likely to be suboptimal
2214 if all your traffic is passed through a single router (i.e., a
2215 "gatewayed" network configuration, as described above).
2216
2217 As with balance-rr, the switch ports need to be configured for
2218 "etherchannel" or "trunking."
2219
2220broadcast: Like active-backup, there is not much advantage to this
2221 mode in this type of network topology.
2222
2223802.3ad: This mode can be a good choice for this type of network
2224 topology. The 802.3ad mode is an IEEE standard, so all peers
2225 that implement 802.3ad should interoperate well. The 802.3ad
2226 protocol includes automatic configuration of the aggregates,
2227 so minimal manual configuration of the switch is needed
2228 (typically only to designate that some set of devices is
2229 available for 802.3ad). The 802.3ad standard also mandates
2230 that frames be delivered in order (within certain limits), so
2231 in general single connections will not see misordering of
2232 packets. The 802.3ad mode does have some drawbacks: the
2233 standard mandates that all devices in the aggregate operate at
2234 the same speed and duplex. Also, as with all bonding load
2235 balance modes other than balance-rr, no single connection will
2236 be able to utilize more than a single interface's worth of
2237 bandwidth.
2238
2239 Additionally, the linux bonding 802.3ad implementation
2240 distributes traffic by peer (using an XOR of MAC addresses),
2241 so in a "gatewayed" configuration, all outgoing traffic will
2242 generally use the same device. Incoming traffic may also end
2243 up on a single device, but that is dependent upon the
2244 balancing policy of the peer's 8023.ad implementation. In a
2245 "local" configuration, traffic will be distributed across the
2246 devices in the bond.
2247
2248 Finally, the 802.3ad mode mandates the use of the MII monitor,
2249 therefore, the ARP monitor is not available in this mode.
2250
2251balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
2252 Since the balancing is done according to MAC address, in a
2253 "gatewayed" configuration (as described above), this mode will
2254 send all traffic across a single device. However, in a
2255 "local" network configuration, this mode balances multiple
2256 local network peers across devices in a vaguely intelligent
2257 manner (not a simple XOR as in balance-xor or 802.3ad mode),
2258 so that mathematically unlucky MAC addresses (i.e., ones that
2259 XOR to the same value) will not all "bunch up" on a single
2260 interface.
2261
2262 Unlike 802.3ad, interfaces may be of differing speeds, and no
2263 special switch configuration is required. On the down side,
2264 in this mode all incoming traffic arrives over a single
2265 interface, this mode requires certain ethtool support in the
2266 network device driver of the slave interfaces, and the ARP
2267 monitor is not available.
2268
2269balance-alb: This mode is everything that balance-tlb is, and more.
2270 It has all of the features (and restrictions) of balance-tlb,
2271 and will also balance incoming traffic from local network
2272 peers (as described in the Bonding Module Options section,
2273 above).
2274
2275 The only additional down side to this mode is that the network
2276 device driver must support changing the hardware address while
2277 the device is open.
2278
Auke Kok6224e012006-06-08 11:15:35 -0700227912.1.2 MT Link Monitoring for Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002280----------------------------------------------------
2281
2282 The choice of link monitoring may largely depend upon which
2283mode you choose to use. The more advanced load balancing modes do not
2284support the use of the ARP monitor, and are thus restricted to using
2285the MII monitor (which does not provide as high a level of end to end
2286assurance as the ARP monitor).
2287
Auke Kok6224e012006-06-08 11:15:35 -0700228812.2 Maximum Throughput in a Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002289-----------------------------------------------------
2290
2291 Multiple switches may be utilized to optimize for throughput
2292when they are configured in parallel as part of an isolated network
2293between two or more systems, for example:
2294
2295 +-----------+
2296 | Host A |
2297 +-+---+---+-+
2298 | | |
2299 +--------+ | +---------+
2300 | | |
2301 +------+---+ +-----+----+ +-----+----+
2302 | Switch A | | Switch B | | Switch C |
2303 +------+---+ +-----+----+ +-----+----+
2304 | | |
2305 +--------+ | +---------+
2306 | | |
2307 +-+---+---+-+
2308 | Host B |
2309 +-----------+
2310
2311 In this configuration, the switches are isolated from one
2312another. One reason to employ a topology such as this is for an
2313isolated network with many hosts (a cluster configured for high
2314performance, for example), using multiple smaller switches can be more
2315cost effective than a single larger switch, e.g., on a network with 24
2316hosts, three 24 port switches can be significantly less expensive than
2317a single 72 port switch.
2318
2319 If access beyond the network is required, an individual host
2320can be equipped with an additional network device connected to an
2321external network; this host then additionally acts as a gateway.
2322
Auke Kok6224e012006-06-08 11:15:35 -0700232312.2.1 MT Bonding Mode Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002324-------------------------------------------------------------
2325
2326 In actual practice, the bonding mode typically employed in
2327configurations of this type is balance-rr. Historically, in this
2328network configuration, the usual caveats about out of order packet
2329delivery are mitigated by the use of network adapters that do not do
2330any kind of packet coalescing (via the use of NAPI, or because the
2331device itself does not generate interrupts until some number of
2332packets has arrived). When employed in this fashion, the balance-rr
2333mode allows individual connections between two hosts to effectively
2334utilize greater than one interface's bandwidth.
2335
Auke Kok6224e012006-06-08 11:15:35 -0700233612.2.2 MT Link Monitoring for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002337------------------------------------------------------
2338
2339 Again, in actual practice, the MII monitor is most often used
2340in this configuration, as performance is given preference over
2341availability. The ARP monitor will function in this topology, but its
2342advantages over the MII monitor are mitigated by the volume of probes
2343needed as the number of systems involved grows (remember that each
2344host in the network is configured with bonding).
2345
Auke Kok6224e012006-06-08 11:15:35 -0700234613. Switch Behavior Issues
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002347==========================
2348
Auke Kok6224e012006-06-08 11:15:35 -0700234913.1 Link Establishment and Failover Delays
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002350-------------------------------------------
2351
2352 Some switches exhibit undesirable behavior with regard to the
2353timing of link up and down reporting by the switch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354
2355 First, when a link comes up, some switches may indicate that
2356the link is up (carrier available), but not pass traffic over the
2357interface for some period of time. This delay is typically due to
2358some type of autonegotiation or routing protocol, but may also occur
2359during switch initialization (e.g., during recovery after a switch
2360failure). If you find this to be a problem, specify an appropriate
2361value to the updelay bonding module option to delay the use of the
2362relevant interface(s).
2363
2364 Second, some switches may "bounce" the link state one or more
2365times while a link is changing state. This occurs most commonly while
2366the switch is initializing. Again, an appropriate updelay value may
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002367help.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368
2369 Note that when a bonding interface has no active links, the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002370driver will immediately reuse the first link that goes up, even if the
2371updelay parameter has been specified (the updelay is ignored in this
2372case). If there are slave interfaces waiting for the updelay timeout
2373to expire, the interface that first went into that state will be
2374immediately reused. This reduces down time of the network if the
2375value of updelay has been overestimated, and since this occurs only in
2376cases with no connectivity, there is no additional penalty for
2377ignoring the updelay.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378
2379 In addition to the concerns about switch timings, if your
2380switches take a long time to go into backup mode, it may be desirable
2381to not activate a backup interface immediately after a link goes down.
2382Failover may be delayed via the downdelay bonding module option.
2383
Auke Kok6224e012006-06-08 11:15:35 -0700238413.2 Duplicated Incoming Packets
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002385--------------------------------
2386
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002387 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2388suppress duplicate packets, which should largely eliminate this problem.
2389The following description is kept for reference.
2390
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002391 It is not uncommon to observe a short burst of duplicated
2392traffic when the bonding device is first used, or after it has been
2393idle for some period of time. This is most easily observed by issuing
2394a "ping" to some other host on the network, and noticing that the
2395output from ping flags duplicates (typically one per slave).
2396
2397 For example, on a bond in active-backup mode with five slaves
2398all connected to one switch, the output may appear as follows:
2399
2400# ping -n 10.0.4.2
2401PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
240264 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
240364 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
240464 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
240564 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
240664 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
240764 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
240864 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
240964 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2410
2411 This is not due to an error in the bonding driver, rather, it
2412is a side effect of how many switches update their MAC forwarding
2413tables. Initially, the switch does not associate the MAC address in
2414the packet with a particular switch port, and so it may send the
2415traffic to all ports until its MAC forwarding table is updated. Since
2416the interfaces attached to the bond may occupy multiple ports on a
2417single switch, when the switch (temporarily) floods the traffic to all
2418ports, the bond device receives multiple copies of the same packet
2419(one per slave device).
2420
2421 The duplicated packet behavior is switch dependent, some
2422switches exhibit this, and some do not. On switches that display this
2423behavior, it can be induced by clearing the MAC forwarding table (on
2424most Cisco switches, the privileged command "clear mac address-table
2425dynamic" will accomplish this).
2426
Auke Kok6224e012006-06-08 11:15:35 -0700242714. Hardware Specific Considerations
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428====================================
2429
2430 This section contains additional information for configuring
2431bonding on specific hardware platforms, or for interfacing bonding
2432with particular switches or other devices.
2433
Auke Kok6224e012006-06-08 11:15:35 -0700243414.1 IBM BladeCenter
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435--------------------
2436
2437 This applies to the JS20 and similar systems.
2438
2439 On the JS20 blades, the bonding driver supports only
2440balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2441largely due to the network topology inside the BladeCenter, detailed
2442below.
2443
2444JS20 network adapter information
2445--------------------------------
2446
2447 All JS20s come with two Broadcom Gigabit Ethernet ports
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002448integrated on the planar (that's "motherboard" in IBM-speak). In the
2449BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2450I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2451An add-on Broadcom daughter card can be installed on a JS20 to provide
2452two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2453wired to I/O Modules 3 and 4, respectively.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454
2455 Each I/O Module may contain either a switch or a passthrough
2456module (which allows ports to be directly connected to an external
2457switch). Some bonding modes require a specific BladeCenter internal
2458network topology in order to function; these are detailed below.
2459
2460 Additional BladeCenter-specific networking information can be
2461found in two IBM Redbooks (www.ibm.com/redbooks):
2462
2463"IBM eServer BladeCenter Networking Options"
2464"IBM eServer BladeCenter Layer 2-7 Network Switching"
2465
2466BladeCenter networking configuration
2467------------------------------------
2468
2469 Because a BladeCenter can be configured in a very large number
2470of ways, this discussion will be confined to describing basic
2471configurations.
2472
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002473 Normally, Ethernet Switch Modules (ESMs) are used in I/O
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2475JS20 will be connected to different internal switches (in the
2476respective I/O modules).
2477
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002478 A passthrough module (OPM or CPM, optical or copper,
2479passthrough module) connects the I/O module directly to an external
2480switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2481interfaces of a JS20 can be redirected to the outside world and
2482connected to a common external switch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002484 Depending upon the mix of ESMs and PMs, the network will
2485appear to bonding as either a single switch topology (all PMs) or as a
2486multiple switch topology (one or more ESMs, zero or more PMs). It is
2487also possible to connect ESMs together, resulting in a configuration
2488much like the example in "High Availability in a Multiple Switch
2489Topology," above.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002490
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002491Requirements for specific modes
2492-------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002494 The balance-rr mode requires the use of passthrough modules
2495for devices in the bond, all connected to an common external switch.
2496That switch must be configured for "etherchannel" or "trunking" on the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497appropriate ports, as is usual for balance-rr.
2498
2499 The balance-alb and balance-tlb modes will function with
2500either switch modules or passthrough modules (or a mix). The only
2501specific requirement for these modes is that all network interfaces
2502must be able to reach all destinations for traffic sent over the
2503bonding device (i.e., the network must converge at some point outside
2504the BladeCenter).
2505
2506 The active-backup mode has no additional requirements.
2507
2508Link monitoring issues
2509----------------------
2510
2511 When an Ethernet Switch Module is in place, only the ARP
2512monitor will reliably detect link loss to an external switch. This is
2513nothing unusual, but examination of the BladeCenter cabinet would
2514suggest that the "external" network ports are the ethernet ports for
2515the system, when it fact there is a switch between these "external"
2516ports and the devices on the JS20 system itself. The MII monitor is
2517only able to detect link failures between the ESM and the JS20 system.
2518
2519 When a passthrough module is in place, the MII monitor does
2520detect failures to the "external" port, which is then directly
2521connected to the JS20 system.
2522
2523Other concerns
2524--------------
2525
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002526 The Serial Over LAN (SoL) link is established over the primary
Linus Torvalds1da177e2005-04-16 15:20:36 -07002527ethernet (eth0) only, therefore, any loss of link to eth0 will result
2528in losing your SoL connection. It will not fail over with other
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002529network traffic, as the SoL system is beyond the control of the
2530bonding driver.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002531
2532 It may be desirable to disable spanning tree on the switch
2533(either the internal Ethernet Switch Module, or an external switch) to
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002534avoid fail-over delay issues when using bonding.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535
2536
Auke Kok6224e012006-06-08 11:15:35 -0700253715. Frequently Asked Questions
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538==============================
2539
25401. Is it SMP safe?
2541
2542 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2543The new driver was designed to be SMP safe from the start.
2544
25452. What type of cards will work with it?
2546
2547 Any Ethernet type cards (you can even mix cards - a Intel
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002548EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2549devices need not be of the same speed.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002551 Starting with version 3.2.1, bonding also supports Infiniband
2552slaves in active-backup mode.
2553
Linus Torvalds1da177e2005-04-16 15:20:36 -070025543. How many bonding devices can I have?
2555
2556 There is no limit.
2557
25584. How many slaves can a bonding device have?
2559
2560 This is limited only by the number of network interfaces Linux
2561supports and/or the number of network cards you can place in your
2562system.
2563
25645. What happens when a slave link dies?
2565
2566 If link monitoring is enabled, then the failing device will be
2567disabled. The active-backup mode will fail over to a backup link, and
2568other modes will ignore the failed link. The link will continue to be
2569monitored, and should it recover, it will rejoin the bond (in whatever
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002570manner is appropriate for the mode). See the sections on High
2571Availability and the documentation for each mode for additional
2572information.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573
2574 Link monitoring can be enabled via either the miimon or
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002575arp_interval parameters (described in the module parameters section,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576above). In general, miimon monitors the carrier state as sensed by
2577the underlying network device, and the arp monitor (arp_interval)
2578monitors connectivity to another host on the local network.
2579
2580 If no link monitoring is configured, the bonding driver will
2581be unable to detect link failures, and will assume that all links are
2582always available. This will likely result in lost packets, and a
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002583resulting degradation of performance. The precise performance loss
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584depends upon the bonding mode and network configuration.
2585
25866. Can bonding be used for High Availability?
2587
2588 Yes. See the section on High Availability for details.
2589
25907. Which switches/systems does it work with?
2591
2592 The full answer to this depends upon the desired mode.
2593
2594 In the basic balance modes (balance-rr and balance-xor), it
2595works with any system that supports etherchannel (also called
2596trunking). Most managed switches currently available have such
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002597support, and many unmanaged switches as well.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002598
2599 The advanced balance modes (balance-tlb and balance-alb) do
2600not have special switch requirements, but do need device drivers that
2601support specific features (described in the appropriate section under
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002602module parameters, above).
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603
Auke Kok6224e012006-06-08 11:15:35 -07002604 In 802.3ad mode, it works with systems that support IEEE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002605802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2606switches currently available support 802.3ad.
2607
2608 The active-backup mode should work with any Layer-II switch.
2609
26108. Where does a bonding device get its MAC address from?
2611
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002612 When using slave devices that have fixed MAC addresses, or when
2613the fail_over_mac option is enabled, the bonding device's MAC address is
2614the MAC address of the active slave.
2615
2616 For other configurations, if not explicitly configured (with
2617ifconfig or ip link), the MAC address of the bonding device is taken from
2618its first slave device. This MAC address is then passed to all following
2619slaves and remains persistent (even if the first slave is removed) until
2620the bonding device is brought down or reconfigured.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002621
2622 If you wish to change the MAC address, you can set it with
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002623ifconfig or ip link:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002624
2625# ifconfig bond0 hw ether 00:11:22:33:44:55
2626
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002627# ip link set bond0 address 66:77:88:99:aa:bb
2628
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 The MAC address can be also changed by bringing down/up the
2630device and then changing its slaves (or their order):
2631
2632# ifconfig bond0 down ; modprobe -r bonding
2633# ifconfig bond0 .... up
2634# ifenslave bond0 eth...
2635
2636 This method will automatically take the address from the next
2637slave that is added.
2638
2639 To restore your slaves' MAC addresses, you need to detach them
2640from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2641then restore the MAC addresses that the slaves had before they were
2642enslaved.
2643
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700264416. Resources and Links
Linus Torvalds1da177e2005-04-16 15:20:36 -07002645=======================
2646
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002647 The latest version of the bonding driver can be found in the latest
Linus Torvalds1da177e2005-04-16 15:20:36 -07002648version of the linux kernel, found on http://kernel.org
2649
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002650 The latest version of this document can be found in the latest kernel
2651source (named Documentation/networking/bonding.txt).
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002652
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002653 Discussions regarding the usage of the bonding driver take place on the
2654bonding-devel mailing list, hosted at sourceforge.net. If you have questions or
2655problems, post them to the list. The list address is:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002656
2657bonding-devel@lists.sourceforge.net
2658
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002659 The administrative interface (to subscribe or unsubscribe) can
2660be found at:
2661
Linus Torvalds1da177e2005-04-16 15:20:36 -07002662https://lists.sourceforge.net/lists/listinfo/bonding-devel
2663
Rick Jonesf8b72d32012-07-20 10:51:37 +00002664 Discussions regarding the development of the bonding driver take place
Nicolas de Pesloüana23c37f12011-03-13 10:34:22 +00002665on the main Linux network mailing list, hosted at vger.kernel.org. The list
2666address is:
2667
2668netdev@vger.kernel.org
2669
2670 The administrative interface (to subscribe or unsubscribe) can
2671be found at:
2672
2673http://vger.kernel.org/vger-lists.html#netdev
2674
Linus Torvalds1da177e2005-04-16 15:20:36 -07002675Donald Becker's Ethernet Drivers and diag programs may be found at :
Justin P. Mattock0ea6e612010-07-23 20:51:24 -07002676 - http://web.archive.org/web/*/http://www.scyld.com/network/
Linus Torvalds1da177e2005-04-16 15:20:36 -07002677
2678You will also find a lot of information regarding Ethernet, NWay, MII,
2679etc. at www.scyld.com.
2680
2681-- END --