blob: a0cda062bc33b6e1de3dd2bb60d5ff62bdf97cd4 [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
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08004 Latest update: 12 November 2007
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
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Auke Kok6224e012006-06-08 11:15:35 -0700534. Querying Bonding Configuration
544.1 Bonding Configuration
554.2 Network Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
Auke Kok6224e012006-06-08 11:15:35 -0700575. Switch Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
Auke Kok6224e012006-06-08 11:15:35 -0700596. 802.1q VLAN Support
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Auke Kok6224e012006-06-08 11:15:35 -0700617. Link Monitoring
627.1 ARP Monitor Operation
637.2 Configuring Multiple ARP Targets
647.3 MII Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -070065
Auke Kok6224e012006-06-08 11:15:35 -0700668. Potential Trouble Sources
678.1 Adventures in Routing
688.2 Ethernet Device Renaming
698.3 Painfully Slow Or No Failed Link Detection By Miimon
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
Auke Kok6224e012006-06-08 11:15:35 -0700719. SNMP agents
Linus Torvalds1da177e2005-04-16 15:20:36 -070072
Auke Kok6224e012006-06-08 11:15:35 -07007310. Promiscuous mode
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Auke Kok6224e012006-06-08 11:15:35 -07007511. Configuring Bonding for High Availability
7611.1 High Availability in a Single Switch Topology
7711.2 High Availability in a Multiple Switch Topology
7811.2.1 HA Bonding Mode Selection for Multiple Switch Topology
7911.2.2 HA Link Monitoring for Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -070080
Auke Kok6224e012006-06-08 11:15:35 -07008112. Configuring Bonding for Maximum Throughput
8212.1 Maximum Throughput in a Single Switch Topology
8312.1.1 MT Bonding Mode Selection for Single Switch Topology
8412.1.2 MT Link Monitoring for Single Switch Topology
8512.2 Maximum Throughput in a Multiple Switch Topology
8612.2.1 MT Bonding Mode Selection for Multiple Switch Topology
8712.2.2 MT Link Monitoring for Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
Auke Kok6224e012006-06-08 11:15:35 -07008913. Switch Behavior Issues
9013.1 Link Establishment and Failover Delays
9113.2 Duplicated Incoming Packets
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
Auke Kok6224e012006-06-08 11:15:35 -07009314. Hardware Specific Considerations
9414.1 IBM BladeCenter
Jay Vosburgh00354cf2005-07-21 12:18:02 -070095
Auke Kok6224e012006-06-08 11:15:35 -07009615. Frequently Asked Questions
Jay Vosburgh00354cf2005-07-21 12:18:02 -070097
Auke Kok6224e012006-06-08 11:15:35 -07009816. Resources and Links
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100
1011. Bonding Driver Installation
102==============================
103
104 Most popular distro kernels ship with the bonding driver
105already available as a module and the ifenslave user level control
106program installed and ready for use. If your distro does not, or you
107have need to compile bonding from source (e.g., configuring and
108installing a mainline kernel from kernel.org), you'll need to perform
109the following steps:
110
1111.1 Configure and build the kernel with bonding
112-----------------------------------------------
113
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700114 The current version of the bonding driver is available in the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115drivers/net/bonding subdirectory of the most recent kernel source
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700116(which is available on http://kernel.org). Most users "rolling their
117own" will want to use the most recent kernel from kernel.org.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118
119 Configure kernel with "make menuconfig" (or "make xconfig" or
120"make config"), then select "Bonding driver support" in the "Network
121device support" section. It is recommended that you configure the
122driver as module since it is currently the only way to pass parameters
123to the driver or configure more than one bonding device.
124
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700125 Build and install the new kernel and modules, then continue
126below to install ifenslave.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
1281.2 Install ifenslave Control Utility
129-------------------------------------
130
131 The ifenslave user level control program is included in the
132kernel source tree, in the file Documentation/networking/ifenslave.c.
133It is generally recommended that you use the ifenslave that
134corresponds to the kernel that you are using (either from the same
135source tree or supplied with the distro), however, ifenslave
136executables from older kernels should function (but features newer
137than the ifenslave release are not supported). Running an ifenslave
138that is newer than the kernel is not supported, and may or may not
139work.
140
141 To install ifenslave, do the following:
142
143# gcc -Wall -O -I/usr/src/linux/include ifenslave.c -o ifenslave
144# cp ifenslave /sbin/ifenslave
145
146 If your kernel source is not in "/usr/src/linux," then replace
147"/usr/src/linux/include" in the above with the location of your kernel
148source include directory.
149
150 You may wish to back up any existing /sbin/ifenslave, or, for
151testing or informal use, tag the ifenslave to the kernel version
152(e.g., name the ifenslave executable /sbin/ifenslave-2.6.10).
153
154IMPORTANT NOTE:
155
156 If you omit the "-I" or specify an incorrect directory, you
157may end up with an ifenslave that is incompatible with the kernel
158you're trying to build it for. Some distros (e.g., Red Hat from 7.1
159onwards) do not have /usr/include/linux symbolically linked to the
160default kernel source include directory.
161
Auke Kok6224e012006-06-08 11:15:35 -0700162SECOND IMPORTANT NOTE:
163 If you plan to configure bonding using sysfs, you do not need
164to use ifenslave.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165
1662. Bonding Driver Options
167=========================
168
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800169 Options for the bonding driver are supplied as parameters to the
170bonding module at load time, or are specified via sysfs.
171
172 Module options may be given as command line arguments to the
173insmod or modprobe command, but are usually specified in either the
174/etc/modules.conf or /etc/modprobe.conf configuration file, or in a
175distro-specific configuration file (some of which are detailed in the next
176section).
177
178 Details on bonding support for sysfs is provided in the
179"Configuring Bonding Manually via Sysfs" section, below.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180
181 The available bonding driver parameters are listed below. If a
182parameter is not specified the default value is used. When initially
183configuring a bond, it is recommended "tail -f /var/log/messages" be
184run in a separate window to watch for bonding driver error messages.
185
186 It is critical that either the miimon or arp_interval and
187arp_ip_target parameters be specified, otherwise serious network
188degradation will occur during link failures. Very few devices do not
189support at least miimon, so there is really no reason not to use it.
190
191 Options with textual values will accept either the text name
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700192or, for backwards compatibility, the option value. E.g.,
193"mode=802.3ad" and "mode=4" set the same mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194
195 The parameters are as follows:
196
197arp_interval
198
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700199 Specifies the ARP link monitoring frequency in milliseconds.
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700200
201 The ARP monitor works by periodically checking the slave
202 devices to determine whether they have sent or received
203 traffic recently (the precise criteria depends upon the
204 bonding mode, and the state of the slave). Regular traffic is
205 generated via ARP probes issued for the addresses specified by
206 the arp_ip_target option.
207
208 This behavior can be modified by the arp_validate option,
209 below.
210
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700211 If ARP monitoring is used in an etherchannel compatible mode
212 (modes 0 and 2), the switch should be configured in a mode
213 that evenly distributes packets across all links. If the
214 switch is configured to distribute the packets in an XOR
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 fashion, all replies from the ARP targets will be received on
216 the same link which could cause the other team members to
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700217 fail. ARP monitoring should not be used in conjunction with
218 miimon. A value of 0 disables ARP monitoring. The default
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219 value is 0.
220
221arp_ip_target
222
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700223 Specifies the IP addresses to use as ARP monitoring peers when
224 arp_interval is > 0. These are the targets of the ARP request
225 sent to determine the health of the link to the targets.
226 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
227 addresses must be separated by a comma. At least one IP
228 address must be given for ARP monitoring to function. The
229 maximum number of targets that can be specified is 16. The
230 default value is no IP addresses.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
Jay Vosburghf5b2b962006-09-22 21:54:53 -0700232arp_validate
233
234 Specifies whether or not ARP probes and replies should be
235 validated in the active-backup mode. This causes the ARP
236 monitor to examine the incoming ARP requests and replies, and
237 only consider a slave to be up if it is receiving the
238 appropriate ARP traffic.
239
240 Possible values are:
241
242 none or 0
243
244 No validation is performed. This is the default.
245
246 active or 1
247
248 Validation is performed only for the active slave.
249
250 backup or 2
251
252 Validation is performed only for backup slaves.
253
254 all or 3
255
256 Validation is performed for all slaves.
257
258 For the active slave, the validation checks ARP replies to
259 confirm that they were generated by an arp_ip_target. Since
260 backup slaves do not typically receive these replies, the
261 validation performed for backup slaves is on the ARP request
262 sent out via the active slave. It is possible that some
263 switch or network configurations may result in situations
264 wherein the backup slaves do not receive the ARP requests; in
265 such a situation, validation of backup slaves must be
266 disabled.
267
268 This option is useful in network configurations in which
269 multiple bonding hosts are concurrently issuing ARPs to one or
270 more targets beyond a common switch. Should the link between
271 the switch and target fail (but not the switch itself), the
272 probe traffic generated by the multiple bonding instances will
273 fool the standard ARP monitor into considering the links as
274 still up. Use of the arp_validate option can resolve this, as
275 the ARP monitor will only consider ARP requests and replies
276 associated with its own instance of bonding.
277
278 This option was added in bonding version 3.1.0.
279
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280downdelay
281
282 Specifies the time, in milliseconds, to wait before disabling
283 a slave after a link failure has been detected. This option
284 is only valid for the miimon link monitor. The downdelay
285 value should be a multiple of the miimon value; if not, it
286 will be rounded down to the nearest multiple. The default
287 value is 0.
288
Jay Vosburghdd957c52007-10-09 19:57:24 -0700289fail_over_mac
290
291 Specifies whether active-backup mode should set all slaves to
292 the same MAC address (the traditional behavior), or, when
293 enabled, change the bond's MAC address when changing the
294 active interface (i.e., fail over the MAC address itself).
295
296 Fail over MAC is useful for devices that cannot ever alter
297 their MAC address, or for devices that refuse incoming
298 broadcasts with their own source MAC (which interferes with
299 the ARP monitor).
300
301 The down side of fail over MAC is that every device on the
302 network must be updated via gratuitous ARP, vs. just updating
303 a switch or set of switches (which often takes place for any
304 traffic, not just ARP traffic, if the switch snoops incoming
305 traffic to update its tables) for the traditional method. If
306 the gratuitous ARP is lost, communication may be disrupted.
307
308 When fail over MAC is used in conjuction with the mii monitor,
309 devices which assert link up prior to being able to actually
310 transmit and receive are particularly susecptible to loss of
311 the gratuitous ARP, and an appropriate updelay setting may be
312 required.
313
314 A value of 0 disables fail over MAC, and is the default. A
315 value of 1 enables fail over MAC. This option is enabled
316 automatically if the first slave added cannot change its MAC
317 address. This option may be modified via sysfs only when no
318 slaves are present in the bond.
319
320 This option was added in bonding version 3.2.0.
321
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322lacp_rate
323
324 Option specifying the rate in which we'll ask our link partner
325 to transmit LACPDU packets in 802.3ad mode. Possible values
326 are:
327
328 slow or 0
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700329 Request partner to transmit LACPDUs every 30 seconds
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330
331 fast or 1
332 Request partner to transmit LACPDUs every 1 second
333
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700334 The default is slow.
335
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336max_bonds
337
338 Specifies the number of bonding devices to create for this
339 instance of the bonding driver. E.g., if max_bonds is 3, and
340 the bonding driver is not already loaded, then bond0, bond1
341 and bond2 will be created. The default value is 1.
342
343miimon
344
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700345 Specifies the MII link monitoring frequency in milliseconds.
346 This determines how often the link state of each slave is
347 inspected for link failures. A value of zero disables MII
348 link monitoring. A value of 100 is a good starting point.
349 The use_carrier option, below, affects how the link state is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 determined. See the High Availability section for additional
351 information. The default value is 0.
352
353mode
354
355 Specifies one of the bonding policies. The default is
356 balance-rr (round robin). Possible values are:
357
358 balance-rr or 0
359
360 Round-robin policy: Transmit packets in sequential
361 order from the first available slave through the
362 last. This mode provides load balancing and fault
363 tolerance.
364
365 active-backup or 1
366
367 Active-backup policy: Only one slave in the bond is
368 active. A different slave becomes active if, and only
369 if, the active slave fails. The bond's MAC address is
370 externally visible on only one port (network adapter)
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700371 to avoid confusing the switch.
372
373 In bonding version 2.6.2 or later, when a failover
374 occurs in active-backup mode, bonding will issue one
375 or more gratuitous ARPs on the newly active slave.
Auke Kok6224e012006-06-08 11:15:35 -0700376 One gratuitous ARP is issued for the bonding master
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700377 interface and each VLAN interfaces configured above
378 it, provided that the interface has at least one IP
379 address configured. Gratuitous ARPs issued for VLAN
380 interfaces are tagged with the appropriate VLAN id.
381
382 This mode provides fault tolerance. The primary
383 option, documented below, affects the behavior of this
384 mode.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385
386 balance-xor or 2
387
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700388 XOR policy: Transmit based on the selected transmit
389 hash policy. The default policy is a simple [(source
390 MAC address XOR'd with destination MAC address) modulo
391 slave count]. Alternate transmit policies may be
392 selected via the xmit_hash_policy option, described
393 below.
394
395 This mode provides load balancing and fault tolerance.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
397 broadcast or 3
398
399 Broadcast policy: transmits everything on all slave
400 interfaces. This mode provides fault tolerance.
401
402 802.3ad or 4
403
404 IEEE 802.3ad Dynamic link aggregation. Creates
405 aggregation groups that share the same speed and
406 duplex settings. Utilizes all slaves in the active
407 aggregator according to the 802.3ad specification.
408
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700409 Slave selection for outgoing traffic is done according
410 to the transmit hash policy, which may be changed from
411 the default simple XOR policy via the xmit_hash_policy
412 option, documented below. Note that not all transmit
413 policies may be 802.3ad compliant, particularly in
414 regards to the packet mis-ordering requirements of
415 section 43.2.4 of the 802.3ad standard. Differing
416 peer implementations will have varying tolerances for
417 noncompliance.
418
419 Prerequisites:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
421 1. Ethtool support in the base drivers for retrieving
422 the speed and duplex of each slave.
423
424 2. A switch that supports IEEE 802.3ad Dynamic link
425 aggregation.
426
427 Most switches will require some type of configuration
428 to enable 802.3ad mode.
429
430 balance-tlb or 5
431
432 Adaptive transmit load balancing: channel bonding that
433 does not require any special switch support. The
434 outgoing traffic is distributed according to the
435 current load (computed relative to the speed) on each
436 slave. Incoming traffic is received by the current
437 slave. If the receiving slave fails, another slave
438 takes over the MAC address of the failed receiving
439 slave.
440
441 Prerequisite:
442
443 Ethtool support in the base drivers for retrieving the
444 speed of each slave.
445
446 balance-alb or 6
447
448 Adaptive load balancing: includes balance-tlb plus
449 receive load balancing (rlb) for IPV4 traffic, and
450 does not require any special switch support. The
451 receive load balancing is achieved by ARP negotiation.
452 The bonding driver intercepts the ARP Replies sent by
453 the local system on their way out and overwrites the
454 source hardware address with the unique hardware
455 address of one of the slaves in the bond such that
456 different peers use different hardware addresses for
457 the server.
458
459 Receive traffic from connections created by the server
460 is also balanced. When the local system sends an ARP
461 Request the bonding driver copies and saves the peer's
462 IP information from the ARP packet. When the ARP
463 Reply arrives from the peer, its hardware address is
464 retrieved and the bonding driver initiates an ARP
465 reply to this peer assigning it to one of the slaves
466 in the bond. A problematic outcome of using ARP
467 negotiation for balancing is that each time that an
468 ARP request is broadcast it uses the hardware address
469 of the bond. Hence, peers learn the hardware address
470 of the bond and the balancing of receive traffic
471 collapses to the current slave. This is handled by
472 sending updates (ARP Replies) to all the peers with
473 their individually assigned hardware address such that
474 the traffic is redistributed. Receive traffic is also
475 redistributed when a new slave is added to the bond
476 and when an inactive slave is re-activated. The
477 receive load is distributed sequentially (round robin)
478 among the group of highest speed slaves in the bond.
479
480 When a link is reconnected or a new slave joins the
481 bond the receive traffic is redistributed among all
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700482 active slaves in the bond by initiating ARP Replies
Auke Kok6224e012006-06-08 11:15:35 -0700483 with the selected MAC address to each of the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 clients. The updelay parameter (detailed below) must
485 be set to a value equal or greater than the switch's
486 forwarding delay so that the ARP Replies sent to the
487 peers will not be blocked by the switch.
488
489 Prerequisites:
490
491 1. Ethtool support in the base drivers for retrieving
492 the speed of each slave.
493
494 2. Base driver support for setting the hardware
495 address of a device while it is open. This is
496 required so that there will always be one slave in the
497 team using the bond hardware address (the
498 curr_active_slave) while having a unique hardware
499 address for each slave in the bond. If the
500 curr_active_slave fails its hardware address is
501 swapped with the new curr_active_slave that was
502 chosen.
503
504primary
505
506 A string (eth0, eth2, etc) specifying which slave is the
507 primary device. The specified device will always be the
508 active slave while it is available. Only when the primary is
509 off-line will alternate devices be used. This is useful when
510 one slave is preferred over another, e.g., when one slave has
511 higher throughput than another.
512
513 The primary option is only valid for active-backup mode.
514
515updelay
516
517 Specifies the time, in milliseconds, to wait before enabling a
518 slave after a link recovery has been detected. This option is
519 only valid for the miimon link monitor. The updelay value
520 should be a multiple of the miimon value; if not, it will be
521 rounded down to the nearest multiple. The default value is 0.
522
523use_carrier
524
525 Specifies whether or not miimon should use MII or ETHTOOL
526 ioctls vs. netif_carrier_ok() to determine the link
527 status. The MII or ETHTOOL ioctls are less efficient and
528 utilize a deprecated calling sequence within the kernel. The
529 netif_carrier_ok() relies on the device driver to maintain its
530 state with netif_carrier_on/off; at this writing, most, but
531 not all, device drivers support this facility.
532
533 If bonding insists that the link is up when it should not be,
534 it may be that your network device driver does not support
535 netif_carrier_on/off. The default state for netif_carrier is
536 "carrier on," so if a driver does not support netif_carrier,
537 it will appear as if the link is always up. In this case,
538 setting use_carrier to 0 will cause bonding to revert to the
539 MII / ETHTOOL ioctl method to determine the link state.
540
541 A value of 1 enables the use of netif_carrier_ok(), a value of
542 0 will use the deprecated MII / ETHTOOL ioctls. The default
543 value is 1.
544
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700545xmit_hash_policy
546
547 Selects the transmit hash policy to use for slave selection in
548 balance-xor and 802.3ad modes. Possible values are:
549
550 layer2
551
552 Uses XOR of hardware MAC addresses to generate the
553 hash. The formula is
554
555 (source MAC XOR destination MAC) modulo slave count
556
557 This algorithm will place all traffic to a particular
558 network peer on the same slave.
559
560 This algorithm is 802.3ad compliant.
561
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800562 layer2+3
563
564 This policy uses a combination of layer2 and layer3
565 protocol information to generate the hash.
566
567 Uses XOR of hardware MAC addresses and IP addresses to
568 generate the hash. The formula is
569
570 (((source IP XOR dest IP) AND 0xffff) XOR
571 ( source MAC XOR destination MAC ))
572 modulo slave count
573
574 This algorithm will place all traffic to a particular
575 network peer on the same slave. For non-IP traffic,
576 the formula is the same as for the layer2 transmit
577 hash policy.
578
579 This policy is intended to provide a more balanced
580 distribution of traffic than layer2 alone, especially
581 in environments where a layer3 gateway device is
582 required to reach most destinations.
583
584 This algorithm is 802.3ad complient.
585
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700586 layer3+4
587
588 This policy uses upper layer protocol information,
589 when available, to generate the hash. This allows for
590 traffic to a particular network peer to span multiple
591 slaves, although a single connection will not span
592 multiple slaves.
593
594 The formula for unfragmented TCP and UDP packets is
595
596 ((source port XOR dest port) XOR
597 ((source IP XOR dest IP) AND 0xffff)
598 modulo slave count
599
600 For fragmented TCP or UDP packets and all other IP
601 protocol traffic, the source and destination port
602 information is omitted. For non-IP traffic, the
603 formula is the same as for the layer2 transmit hash
604 policy.
605
606 This policy is intended to mimic the behavior of
607 certain switches, notably Cisco switches with PFC2 as
608 well as some Foundry and IBM products.
609
610 This algorithm is not fully 802.3ad compliant. A
611 single TCP or UDP conversation containing both
612 fragmented and unfragmented packets will see packets
613 striped across two interfaces. This may result in out
614 of order delivery. Most traffic types will not meet
615 this criteria, as TCP rarely fragments traffic, and
616 most UDP traffic is not involved in extended
617 conversations. Other implementations of 802.3ad may
618 or may not tolerate this noncompliance.
619
620 The default value is layer2. This option was added in bonding
Jay Vosburgh6f6652b2007-12-06 23:40:34 -0800621 version 2.6.3. In earlier versions of bonding, this parameter
622 does not exist, and the layer2 policy is the only policy. The
623 layer2+3 value was added for bonding version 3.2.2.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624
625
6263. Configuring Bonding Devices
627==============================
628
Auke Kok6224e012006-06-08 11:15:35 -0700629 You can configure bonding using either your distro's network
630initialization scripts, or manually using either ifenslave or the
631sysfs interface. Distros generally use one of two packages for the
632network initialization scripts: initscripts or sysconfig. Recent
633versions of these packages have support for bonding, while older
634versions do not.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635
636 We will first describe the options for configuring bonding for
637distros using versions of initscripts and sysconfig with full or
638partial support for bonding, then provide information on enabling
639bonding without support from the network initialization scripts (i.e.,
640older versions of initscripts or sysconfig).
641
642 If you're unsure whether your distro uses sysconfig or
643initscripts, or don't know if it's new enough, have no fear.
644Determining this is fairly straightforward.
645
646 First, issue the command:
647
648$ rpm -qf /sbin/ifup
649
650 It will respond with a line of text starting with either
651"initscripts" or "sysconfig," followed by some numbers. This is the
652package that provides your network initialization scripts.
653
654 Next, to determine if your installation supports bonding,
655issue the command:
656
657$ grep ifenslave /sbin/ifup
658
659 If this returns any matches, then your initscripts or
660sysconfig has support for bonding.
661
Auke Kok6224e012006-06-08 11:15:35 -07006623.1 Configuration with Sysconfig Support
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663----------------------------------------
664
665 This section applies to distros using a version of sysconfig
666with bonding support, for example, SuSE Linux Enterprise Server 9.
667
668 SuSE SLES 9's networking configuration system does support
669bonding, however, at this writing, the YaST system configuration
Auke Kok6224e012006-06-08 11:15:35 -0700670front end does not provide any means to work with bonding devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671Bonding devices can be managed by hand, however, as follows.
672
673 First, if they have not already been configured, configure the
674slave devices. On SLES 9, this is most easily done by running the
675yast2 sysconfig configuration utility. The goal is for to create an
676ifcfg-id file for each slave device. The simplest way to accomplish
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700677this is to configure the devices for DHCP (this is only to get the
678file ifcfg-id file created; see below for some issues with DHCP). The
679name of the configuration file for each device will be of the form:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
681ifcfg-id-xx:xx:xx:xx:xx:xx
682
683 Where the "xx" portion will be replaced with the digits from
684the device's permanent MAC address.
685
686 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
687created, it is necessary to edit the configuration files for the slave
688devices (the MAC addresses correspond to those of the slave devices).
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700689Before editing, the file will contain multiple lines, and will look
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690something like this:
691
692BOOTPROTO='dhcp'
693STARTMODE='on'
694USERCTL='no'
695UNIQUE='XNzu.WeZGOGF+4wE'
696_nm_name='bus-pci-0001:61:01.0'
697
698 Change the BOOTPROTO and STARTMODE lines to the following:
699
700BOOTPROTO='none'
701STARTMODE='off'
702
703 Do not alter the UNIQUE or _nm_name lines. Remove any other
704lines (USERCTL, etc).
705
706 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
707it's time to create the configuration file for the bonding device
708itself. This file is named ifcfg-bondX, where X is the number of the
709bonding device to create, starting at 0. The first such file is
710ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
711network configuration system will correctly start multiple instances
712of bonding.
713
714 The contents of the ifcfg-bondX file is as follows:
715
716BOOTPROTO="static"
717BROADCAST="10.0.2.255"
718IPADDR="10.0.2.10"
719NETMASK="255.255.0.0"
720NETWORK="10.0.2.0"
721REMOTE_IPADDR=""
722STARTMODE="onboot"
723BONDING_MASTER="yes"
724BONDING_MODULE_OPTS="mode=active-backup miimon=100"
725BONDING_SLAVE0="eth0"
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700726BONDING_SLAVE1="bus-pci-0000:06:08.1"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
728 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
729values with the appropriate values for your network.
730
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 The STARTMODE specifies when the device is brought online.
732The possible values are:
733
734 onboot: The device is started at boot time. If you're not
735 sure, this is probably what you want.
736
737 manual: The device is started only when ifup is called
738 manually. Bonding devices may be configured this
739 way if you do not wish them to start automatically
740 at boot for some reason.
741
742 hotplug: The device is started by a hotplug event. This is not
743 a valid choice for a bonding device.
744
745 off or ignore: The device configuration is ignored.
746
747 The line BONDING_MASTER='yes' indicates that the device is a
748bonding master device. The only useful value is "yes."
749
750 The contents of BONDING_MODULE_OPTS are supplied to the
751instance of the bonding module for this device. Specify the options
752for the bonding mode, link monitoring, and so on here. Do not include
753the max_bonds bonding parameter; this will confuse the configuration
754system if you have multiple bonding devices.
755
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700756 Finally, supply one BONDING_SLAVEn="slave device" for each
757slave. where "n" is an increasing value, one for each slave. The
758"slave device" is either an interface name, e.g., "eth0", or a device
759specifier for the network device. The interface name is easier to
760find, but the ethN names are subject to change at boot time if, e.g.,
761a device early in the sequence has failed. The device specifiers
762(bus-pci-0000:06:08.1 in the example above) specify the physical
763network device, and will not change unless the device's bus location
764changes (for example, it is moved from one PCI slot to another). The
765example above uses one of each type for demonstration purposes; most
766configurations will choose one or the other for all slave devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767
768 When all configuration files have been modified or created,
769networking must be restarted for the configuration changes to take
770effect. This can be accomplished via the following:
771
772# /etc/init.d/network restart
773
774 Note that the network control script (/sbin/ifdown) will
775remove the bonding module as part of the network shutdown processing,
776so it is not necessary to remove the module by hand if, e.g., the
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700777module parameters have changed.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
779 Also, at this writing, YaST/YaST2 will not manage bonding
780devices (they do not show bonding interfaces on its list of network
781devices). It is necessary to edit the configuration file by hand to
782change the bonding configuration.
783
784 Additional general options and details of the ifcfg file
785format can be found in an example ifcfg template file:
786
787/etc/sysconfig/network/ifcfg.template
788
789 Note that the template does not document the various BONDING_
790settings described above, but does describe many of the other options.
791
Auke Kok6224e012006-06-08 11:15:35 -07007923.1.1 Using DHCP with Sysconfig
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700793-------------------------------
794
795 Under sysconfig, configuring a device with BOOTPROTO='dhcp'
796will cause it to query DHCP for its IP address information. At this
797writing, this does not function for bonding devices; the scripts
798attempt to obtain the device address from DHCP prior to adding any of
799the slave devices. Without active slaves, the DHCP requests are not
800sent to the network.
801
Auke Kok6224e012006-06-08 11:15:35 -07008023.1.2 Configuring Multiple Bonds with Sysconfig
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700803-----------------------------------------------
804
805 The sysconfig network initialization system is capable of
806handling multiple bonding devices. All that is necessary is for each
807bonding instance to have an appropriately configured ifcfg-bondX file
808(as described above). Do not specify the "max_bonds" parameter to any
809instance of bonding, as this will confuse sysconfig. If you require
810multiple bonding devices with identical parameters, create multiple
811ifcfg-bondX files.
812
813 Because the sysconfig scripts supply the bonding module
814options in the ifcfg-bondX file, it is not necessary to add them to
815the system /etc/modules.conf or /etc/modprobe.conf configuration file.
816
Auke Kok6224e012006-06-08 11:15:35 -07008173.2 Configuration with Initscripts Support
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818------------------------------------------
819
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800820 This section applies to distros using a recent version of
821initscripts with bonding support, for example, Red Hat Enterprise Linux
822version 3 or later, Fedora, etc. On these systems, the network
823initialization scripts have knowledge of bonding, and can be configured to
824control bonding devices. Note that older versions of the initscripts
825package have lower levels of support for bonding; this will be noted where
826applicable.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
828 These distros will not automatically load the network adapter
829driver unless the ethX device is configured with an IP address.
830Because of this constraint, users must manually configure a
831network-script file for all physical adapters that will be members of
832a bondX link. Network script files are located in the directory:
833
834/etc/sysconfig/network-scripts
835
836 The file name must be prefixed with "ifcfg-eth" and suffixed
837with the adapter's physical adapter number. For example, the script
838for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
839Place the following text in the file:
840
841DEVICE=eth0
842USERCTL=no
843ONBOOT=yes
844MASTER=bond0
845SLAVE=yes
846BOOTPROTO=none
847
848 The DEVICE= line will be different for every ethX device and
849must correspond with the name of the file, i.e., ifcfg-eth1 must have
850a device line of DEVICE=eth1. The setting of the MASTER= line will
851also depend on the final bonding interface name chosen for your bond.
852As with other network devices, these typically start at 0, and go up
853one for each device, i.e., the first bonding instance is bond0, the
854second is bond1, and so on.
855
856 Next, create a bond network script. The file name for this
857script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
858the number of the bond. For bond0 the file is named "ifcfg-bond0",
859for bond1 it is named "ifcfg-bond1", and so on. Within that file,
860place the following text:
861
862DEVICE=bond0
863IPADDR=192.168.1.1
864NETMASK=255.255.255.0
865NETWORK=192.168.1.0
866BROADCAST=192.168.1.255
867ONBOOT=yes
868BOOTPROTO=none
869USERCTL=no
870
871 Be sure to change the networking specific lines (IPADDR,
872NETMASK, NETWORK and BROADCAST) to match your network configuration.
873
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800874 For later versions of initscripts, such as that found with Fedora
8757 and Red Hat Enterprise Linux version 5 (or later), it is possible, and,
876indeed, preferable, to specify the bonding options in the ifcfg-bond0
877file, e.g. a line of the format:
878
879BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=+192.168.1.254"
880
881 will configure the bond with the specified options. The options
882specified in BONDING_OPTS are identical to the bonding module parameters
883except for the arp_ip_target field. Each target should be included as a
884separate option and should be preceded by a '+' to indicate it should be
885added to the list of queried targets, e.g.,
886
887 arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
888
889 is the proper syntax to specify multiple targets. When specifying
890options via BONDING_OPTS, it is not necessary to edit /etc/modules.conf or
891/etc/modprobe.conf.
892
893 For older versions of initscripts that do not support
894BONDING_OPTS, it is necessary to edit /etc/modules.conf (or
895/etc/modprobe.conf, depending upon your distro) to load the bonding module
896with your desired options when the bond0 interface is brought up. The
897following lines in /etc/modules.conf (or modprobe.conf) will load the
898bonding module, and select its options:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899
900alias bond0 bonding
901options bond0 mode=balance-alb miimon=100
902
903 Replace the sample parameters with the appropriate set of
904options for your configuration.
905
906 Finally run "/etc/rc.d/init.d/network restart" as root. This
907will restart the networking subsystem and your bond link should be now
908up and running.
909
Auke Kok6224e012006-06-08 11:15:35 -07009103.2.1 Using DHCP with Initscripts
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700911---------------------------------
912
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800913 Recent versions of initscripts (the versions supplied with Fedora
914Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
915work) have support for assigning IP information to bonding devices via
916DHCP.
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700917
918 To configure bonding for DHCP, configure it as described
919above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
920and add a line consisting of "TYPE=Bonding". Note that the TYPE value
921is case sensitive.
922
Auke Kok6224e012006-06-08 11:15:35 -07009233.2.2 Configuring Multiple Bonds with Initscripts
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700924-------------------------------------------------
925
Jay Vosburgh9a6c6862007-11-13 20:25:48 -0800926 Initscripts packages that are included with Fedora 7 and Red Hat
927Enterprise Linux 5 support multiple bonding interfaces by simply
928specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
929number of the bond. This support requires sysfs support in the kernel,
930and a bonding driver of version 3.0.0 or later. Other configurations may
931not support this method for specifying multiple bonding interfaces; for
932those instances, see the "Configuring Multiple Bonds Manually" section,
933below.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934
Auke Kok6224e012006-06-08 11:15:35 -07009353.3 Configuring Bonding Manually with Ifenslave
936-----------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
938 This section applies to distros whose network initialization
939scripts (the sysconfig or initscripts package) do not have specific
940knowledge of bonding. One such distro is SuSE Linux Enterprise Server
941version 8.
942
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700943 The general method for these systems is to place the bonding
944module parameters into /etc/modules.conf or /etc/modprobe.conf (as
945appropriate for the installed distro), then add modprobe and/or
946ifenslave commands to the system's global init script. The name of
947the global init script differs; for sysconfig, it is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948/etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
949
950 For example, if you wanted to make a simple bond of two e100
951devices (presumed to be eth0 and eth1), and have it persist across
952reboots, edit the appropriate file (/etc/init.d/boot.local or
953/etc/rc.d/rc.local), and add the following:
954
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700955modprobe bonding mode=balance-alb miimon=100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956modprobe e100
957ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
958ifenslave bond0 eth0
959ifenslave bond0 eth1
960
961 Replace the example bonding module parameters and bond0
962network configuration (IP address, netmask, etc) with the appropriate
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700963values for your configuration.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964
965 Unfortunately, this method will not provide support for the
966ifup and ifdown scripts on the bond devices. To reload the bonding
967configuration, it is necessary to run the initialization script, e.g.,
968
969# /etc/init.d/boot.local
970
971 or
972
973# /etc/rc.d/rc.local
974
975 It may be desirable in such a case to create a separate script
976which only initializes the bonding configuration, then call that
977separate script from within boot.local. This allows for bonding to be
978enabled without re-running the entire global init script.
979
980 To shut down the bonding devices, it is necessary to first
981mark the bonding device itself as being down, then remove the
982appropriate device driver modules. For our example above, you can do
983the following:
984
985# ifconfig bond0 down
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700986# rmmod bonding
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987# rmmod e100
988
989 Again, for convenience, it may be desirable to create a script
990with these commands.
991
992
Jay Vosburgh00354cf2005-07-21 12:18:02 -07009933.3.1 Configuring Multiple Bonds Manually
994-----------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
996 This section contains information on configuring multiple
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700997bonding devices with differing options for those systems whose network
998initialization scripts lack support for configuring multiple bonds.
999
1000 If you require multiple bonding devices, but all with the same
1001options, you may wish to use the "max_bonds" module parameter,
1002documented above.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001004 To create multiple bonding devices with differing options, it is
1005preferrable to use bonding parameters exported by sysfs, documented in the
1006section below.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001008 For versions of bonding without sysfs support, the only means to
1009provide multiple instances of bonding with differing options is to load
1010the bonding driver multiple times. Note that current versions of the
1011sysconfig network initialization scripts handle this automatically; if
1012your distro uses these scripts, no special action is needed. See the
1013section Configuring Bonding Devices, above, if you're not sure about your
1014network initialization scripts.
1015
1016 To load multiple instances of the module, it is necessary to
1017specify a different name for each instance (the module loading system
1018requires that every loaded module, even multiple instances of the same
1019module, have a unique name). This is accomplished by supplying multiple
1020sets of bonding options in /etc/modprobe.conf, for example:
1021
1022alias bond0 bonding
1023options bond0 -o bond0 mode=balance-rr miimon=100
1024
1025alias bond1 bonding
1026options bond1 -o bond1 mode=balance-alb miimon=50
1027
1028 will load the bonding module two times. The first instance is
1029named "bond0" and creates the bond0 device in balance-rr mode with an
1030miimon of 100. The second instance is named "bond1" and creates the
1031bond1 device in balance-alb mode with an miimon of 50.
1032
1033 In some circumstances (typically with older distributions),
1034the above does not work, and the second bonding instance never sees
1035its options. In that case, the second options line can be substituted
1036as follows:
1037
1038install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1039 mode=balance-alb miimon=50
1040
1041 This may be repeated any number of times, specifying a new and
1042unique name in place of bond1 for each subsequent instance.
1043
1044 It has been observed that some Red Hat supplied kernels are unable
1045to rename modules at load time (the "-o bond1" part). Attempts to pass
1046that option to modprobe will produce an "Operation not permitted" error.
1047This has been reported on some Fedora Core kernels, and has been seen on
1048RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
1049to configure multiple bonds with differing parameters (as they are older
1050kernels, and also lack sysfs support).
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001051
Auke Kok6224e012006-06-08 11:15:35 -070010523.4 Configuring Bonding Manually via Sysfs
1053------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001055 Starting with version 3.0.0, Channel Bonding may be configured
Auke Kok6224e012006-06-08 11:15:35 -07001056via the sysfs interface. This interface allows dynamic configuration
1057of all bonds in the system without unloading the module. It also
1058allows for adding and removing bonds at runtime. Ifenslave is no
1059longer required, though it is still supported.
1060
1061 Use of the sysfs interface allows you to use multiple bonds
1062with different configurations without having to reload the module.
1063It also allows you to use multiple, differently configured bonds when
1064bonding is compiled into the kernel.
1065
1066 You must have the sysfs filesystem mounted to configure
1067bonding this way. The examples in this document assume that you
1068are using the standard mount point for sysfs, e.g. /sys. If your
1069sysfs filesystem is mounted elsewhere, you will need to adjust the
1070example paths accordingly.
1071
1072Creating and Destroying Bonds
1073-----------------------------
1074To add a new bond foo:
1075# echo +foo > /sys/class/net/bonding_masters
1076
1077To remove an existing bond bar:
1078# echo -bar > /sys/class/net/bonding_masters
1079
1080To show all existing bonds:
1081# cat /sys/class/net/bonding_masters
1082
1083NOTE: due to 4K size limitation of sysfs files, this list may be
1084truncated if you have more than a few hundred bonds. This is unlikely
1085to occur under normal operating conditions.
1086
1087Adding and Removing Slaves
1088--------------------------
1089 Interfaces may be enslaved to a bond using the file
1090/sys/class/net/<bond>/bonding/slaves. The semantics for this file
1091are the same as for the bonding_masters file.
1092
1093To enslave interface eth0 to bond bond0:
1094# ifconfig bond0 up
1095# echo +eth0 > /sys/class/net/bond0/bonding/slaves
1096
1097To free slave eth0 from bond bond0:
1098# echo -eth0 > /sys/class/net/bond0/bonding/slaves
1099
Auke Kok6224e012006-06-08 11:15:35 -07001100 When an interface is enslaved to a bond, symlinks between the
1101two are created in the sysfs filesystem. In this case, you would get
1102/sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1103/sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1104
1105 This means that you can tell quickly whether or not an
1106interface is enslaved by looking for the master symlink. Thus:
1107# echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1108will free eth0 from whatever bond it is enslaved to, regardless of
1109the name of the bond interface.
1110
1111Changing a Bond's Configuration
1112-------------------------------
1113 Each bond may be configured individually by manipulating the
1114files located in /sys/class/net/<bond name>/bonding
1115
1116 The names of these files correspond directly with the command-
Paolo Ornati670e9f32006-10-03 22:57:56 +02001117line parameters described elsewhere in this file, and, with the
Auke Kok6224e012006-06-08 11:15:35 -07001118exception of arp_ip_target, they accept the same values. To see the
1119current setting, simply cat the appropriate file.
1120
1121 A few examples will be given here; for specific usage
1122guidelines for each parameter, see the appropriate section in this
1123document.
1124
1125To configure bond0 for balance-alb mode:
1126# ifconfig bond0 down
1127# echo 6 > /sys/class/net/bond0/bonding/mode
1128 - or -
1129# echo balance-alb > /sys/class/net/bond0/bonding/mode
1130 NOTE: The bond interface must be down before the mode can be
1131changed.
1132
1133To enable MII monitoring on bond0 with a 1 second interval:
1134# echo 1000 > /sys/class/net/bond0/bonding/miimon
1135 NOTE: If ARP monitoring is enabled, it will disabled when MII
1136monitoring is enabled, and vice-versa.
1137
1138To add ARP targets:
1139# echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1140# echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1141 NOTE: up to 10 target addresses may be specified.
1142
1143To remove an ARP target:
1144# echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1145
1146Example Configuration
1147---------------------
1148 We begin with the same example that is shown in section 3.3,
1149executed with sysfs, and without using ifenslave.
1150
1151 To make a simple bond of two e100 devices (presumed to be eth0
1152and eth1), and have it persist across reboots, edit the appropriate
1153file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1154following:
1155
1156modprobe bonding
1157modprobe e100
1158echo balance-alb > /sys/class/net/bond0/bonding/mode
1159ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1160echo 100 > /sys/class/net/bond0/bonding/miimon
1161echo +eth0 > /sys/class/net/bond0/bonding/slaves
1162echo +eth1 > /sys/class/net/bond0/bonding/slaves
1163
1164 To add a second bond, with two e1000 interfaces in
1165active-backup mode, using ARP monitoring, add the following lines to
1166your init script:
1167
1168modprobe e1000
1169echo +bond1 > /sys/class/net/bonding_masters
1170echo active-backup > /sys/class/net/bond1/bonding/mode
1171ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1172echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1173echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1174echo +eth2 > /sys/class/net/bond1/bonding/slaves
1175echo +eth3 > /sys/class/net/bond1/bonding/slaves
1176
1177
11784. Querying Bonding Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179=================================
1180
Auke Kok6224e012006-06-08 11:15:35 -070011814.1 Bonding Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182-------------------------
1183
1184 Each bonding device has a read-only file residing in the
1185/proc/net/bonding directory. The file contents include information
1186about the bonding configuration, options and state of each slave.
1187
1188 For example, the contents of /proc/net/bonding/bond0 after the
1189driver is loaded with parameters of mode=0 and miimon=1000 is
1190generally as follows:
1191
1192 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1193 Bonding Mode: load balancing (round-robin)
1194 Currently Active Slave: eth0
1195 MII Status: up
1196 MII Polling Interval (ms): 1000
1197 Up Delay (ms): 0
1198 Down Delay (ms): 0
1199
1200 Slave Interface: eth1
1201 MII Status: up
1202 Link Failure Count: 1
1203
1204 Slave Interface: eth0
1205 MII Status: up
1206 Link Failure Count: 1
1207
1208 The precise format and contents will change depending upon the
1209bonding configuration, state, and version of the bonding driver.
1210
Auke Kok6224e012006-06-08 11:15:35 -070012114.2 Network configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212-------------------------
1213
1214 The network configuration can be inspected using the ifconfig
1215command. Bonding devices will have the MASTER flag set; Bonding slave
1216devices will have the SLAVE flag set. The ifconfig output does not
1217contain information on which slaves are associated with which masters.
1218
1219 In the example below, the bond0 interface is the master
1220(MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1221bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1222TLB and ALB that require a unique MAC address for each slave.
1223
1224# /sbin/ifconfig
1225bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
1226 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
1227 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
1228 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1229 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1230 collisions:0 txqueuelen:0
1231
1232eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1234 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1235 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1236 collisions:0 txqueuelen:100
1237 Interrupt:10 Base address:0x1080
1238
1239eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
1241 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1242 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1243 collisions:0 txqueuelen:100
1244 Interrupt:9 Base address:0x1400
1245
Auke Kok6224e012006-06-08 11:15:35 -070012465. Switch Configuration
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247=======================
1248
1249 For this section, "switch" refers to whatever system the
1250bonded devices are directly connected to (i.e., where the other end of
1251the cable plugs into). This may be an actual dedicated switch device,
1252or it may be another regular system (e.g., another computer running
1253Linux),
1254
1255 The active-backup, balance-tlb and balance-alb modes do not
1256require any specific configuration of the switch.
1257
1258 The 802.3ad mode requires that the switch have the appropriate
1259ports configured as an 802.3ad aggregation. The precise method used
1260to configure this varies from switch to switch, but, for example, a
1261Cisco 3550 series switch requires that the appropriate ports first be
1262grouped together in a single etherchannel instance, then that
1263etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1264standard EtherChannel).
1265
1266 The balance-rr, balance-xor and broadcast modes generally
1267require that the switch have the appropriate ports grouped together.
1268The nomenclature for such a group differs between switches, it may be
1269called an "etherchannel" (as in the Cisco example, above), a "trunk
1270group" or some other similar variation. For these modes, each switch
1271will also have its own configuration options for the switch's transmit
1272policy to the bond. Typical choices include XOR of either the MAC or
1273IP addresses. The transmit policy of the two peers does not need to
1274match. For these three modes, the bonding mode really selects a
1275transmit policy for an EtherChannel group; all three will interoperate
1276with another EtherChannel group.
1277
1278
Auke Kok6224e012006-06-08 11:15:35 -070012796. 802.1q VLAN Support
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280======================
1281
1282 It is possible to configure VLAN devices over a bond interface
1283using the 8021q driver. However, only packets coming from the 8021q
1284driver and passing through bonding will be tagged by default. Self
1285generated packets, for example, bonding's learning packets or ARP
1286packets generated by either ALB mode or the ARP monitor mechanism, are
1287tagged internally by bonding itself. As a result, bonding must
1288"learn" the VLAN IDs configured above it, and use those IDs to tag
1289self generated packets.
1290
1291 For reasons of simplicity, and to support the use of adapters
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001292that can do VLAN hardware acceleration offloading, the bonding
1293interface declares itself as fully hardware offloading capable, it gets
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294the add_vid/kill_vid notifications to gather the necessary
1295information, and it propagates those actions to the slaves. In case
1296of mixed adapter types, hardware accelerated tagged packets that
1297should go through an adapter that is not offloading capable are
1298"un-accelerated" by the bonding driver so the VLAN tag sits in the
1299regular location.
1300
1301 VLAN interfaces *must* be added on top of a bonding interface
1302only after enslaving at least one slave. The bonding interface has a
1303hardware address of 00:00:00:00:00:00 until the first slave is added.
1304If the VLAN interface is created prior to the first enslavement, it
1305would pick up the all-zeroes hardware address. Once the first slave
1306is attached to the bond, the bond device itself will pick up the
1307slave's hardware address, which is then available for the VLAN device.
1308
1309 Also, be aware that a similar problem can occur if all slaves
1310are released from a bond that still has one or more VLAN interfaces on
1311top of it. When a new slave is added, the bonding interface will
1312obtain its hardware address from the first slave, which might not
1313match the hardware address of the VLAN interfaces (which was
1314ultimately copied from an earlier slave).
1315
1316 There are two methods to insure that the VLAN device operates
1317with the correct hardware address if all slaves are removed from a
1318bond interface:
1319
1320 1. Remove all VLAN interfaces then recreate them
1321
1322 2. Set the bonding interface's hardware address so that it
1323matches the hardware address of the VLAN interfaces.
1324
1325 Note that changing a VLAN interface's HW address would set the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001326underlying device -- i.e. the bonding interface -- to promiscuous
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327mode, which might not be what you want.
1328
1329
Auke Kok6224e012006-06-08 11:15:35 -070013307. Link Monitoring
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331==================
1332
1333 The bonding driver at present supports two schemes for
1334monitoring a slave device's link state: the ARP monitor and the MII
1335monitor.
1336
1337 At the present time, due to implementation restrictions in the
1338bonding driver itself, it is not possible to enable both ARP and MII
1339monitoring simultaneously.
1340
Auke Kok6224e012006-06-08 11:15:35 -070013417.1 ARP Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342-------------------------
1343
1344 The ARP monitor operates as its name suggests: it sends ARP
1345queries to one or more designated peer systems on the network, and
1346uses the response as an indication that the link is operating. This
1347gives some assurance that traffic is actually flowing to and from one
1348or more peers on the local network.
1349
1350 The ARP monitor relies on the device driver itself to verify
1351that traffic is flowing. In particular, the driver must keep up to
1352date the last receive time, dev->last_rx, and transmit start time,
1353dev->trans_start. If these are not updated by the driver, then the
1354ARP monitor will immediately fail any slaves using that driver, and
1355those slaves will stay down. If networking monitoring (tcpdump, etc)
1356shows the ARP requests and replies on the network, then it may be that
1357your device driver is not updating last_rx and trans_start.
1358
Auke Kok6224e012006-06-08 11:15:35 -070013597.2 Configuring Multiple ARP Targets
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360------------------------------------
1361
1362 While ARP monitoring can be done with just one target, it can
1363be useful in a High Availability setup to have several targets to
1364monitor. In the case of just one target, the target itself may go
1365down or have a problem making it unresponsive to ARP requests. Having
1366an additional target (or several) increases the reliability of the ARP
1367monitoring.
1368
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001369 Multiple ARP targets must be separated by commas as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370
1371# example options for ARP monitoring with three targets
1372alias bond0 bonding
1373options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1374
1375 For just a single target the options would resemble:
1376
1377# example options for ARP monitoring with one target
1378alias bond0 bonding
1379options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1380
1381
Auke Kok6224e012006-06-08 11:15:35 -070013827.3 MII Monitor Operation
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383-------------------------
1384
1385 The MII monitor monitors only the carrier state of the local
1386network interface. It accomplishes this in one of three ways: by
1387depending upon the device driver to maintain its carrier state, by
1388querying the device's MII registers, or by making an ethtool query to
1389the device.
1390
1391 If the use_carrier module parameter is 1 (the default value),
1392then the MII monitor will rely on the driver for carrier state
1393information (via the netif_carrier subsystem). As explained in the
1394use_carrier parameter information, above, if the MII monitor fails to
1395detect carrier loss on the device (e.g., when the cable is physically
1396disconnected), it may be that the driver does not support
1397netif_carrier.
1398
1399 If use_carrier is 0, then the MII monitor will first query the
1400device's (via ioctl) MII registers and check the link state. If that
1401request fails (not just that it returns carrier down), then the MII
1402monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1403the same information. If both methods fail (i.e., the driver either
1404does not support or had some error in processing both the MII register
1405and ethtool requests), then the MII monitor will assume the link is
1406up.
1407
Auke Kok6224e012006-06-08 11:15:35 -070014088. Potential Sources of Trouble
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409===============================
1410
Auke Kok6224e012006-06-08 11:15:35 -070014118.1 Adventures in Routing
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412-------------------------
1413
1414 When bonding is configured, it is important that the slave
Auke Kok6224e012006-06-08 11:15:35 -07001415devices not have routes that supersede routes of the master (or,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416generally, not have routes at all). For example, suppose the bonding
1417device bond0 has two slaves, eth0 and eth1, and the routing table is
1418as follows:
1419
1420Kernel IP routing table
1421Destination Gateway Genmask Flags MSS Window irtt Iface
142210.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
142310.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
142410.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
1425127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
1426
1427 This routing configuration will likely still update the
1428receive/transmit times in the driver (needed by the ARP monitor), but
1429may bypass the bonding driver (because outgoing traffic to, in this
1430case, another host on network 10 would use eth0 or eth1 before bond0).
1431
1432 The ARP monitor (and ARP itself) may become confused by this
1433configuration, because ARP requests (generated by the ARP monitor)
1434will be sent on one interface (bond0), but the corresponding reply
1435will arrive on a different interface (eth0). This reply looks to ARP
1436as an unsolicited ARP reply (because ARP matches replies on an
1437interface basis), and is discarded. The MII monitor is not affected
1438by the state of the routing table.
1439
1440 The solution here is simply to insure that slaves do not have
1441routes of their own, and if for some reason they must, those routes do
Auke Kok6224e012006-06-08 11:15:35 -07001442not supersede routes of their master. This should generally be the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443case, but unusual configurations or errant manual or automatic static
1444route additions may cause trouble.
1445
Auke Kok6224e012006-06-08 11:15:35 -070014468.2 Ethernet Device Renaming
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447----------------------------
1448
1449 On systems with network configuration scripts that do not
1450associate physical devices directly with network interface names (so
1451that the same physical device always has the same "ethX" name), it may
1452be necessary to add some special logic to either /etc/modules.conf or
1453/etc/modprobe.conf (depending upon which is installed on the system).
1454
1455 For example, given a modules.conf containing the following:
1456
1457alias bond0 bonding
1458options bond0 mode=some-mode miimon=50
1459alias eth0 tg3
1460alias eth1 tg3
1461alias eth2 e1000
1462alias eth3 e1000
1463
1464 If neither eth0 and eth1 are slaves to bond0, then when the
1465bond0 interface comes up, the devices may end up reordered. This
1466happens because bonding is loaded first, then its slave device's
1467drivers are loaded next. Since no other drivers have been loaded,
1468when the e1000 driver loads, it will receive eth0 and eth1 for its
1469devices, but the bonding configuration tries to enslave eth2 and eth3
1470(which may later be assigned to the tg3 devices).
1471
1472 Adding the following:
1473
1474add above bonding e1000 tg3
1475
1476 causes modprobe to load e1000 then tg3, in that order, when
1477bonding is loaded. This command is fully documented in the
1478modules.conf manual page.
1479
1480 On systems utilizing modprobe.conf (or modprobe.conf.local),
1481an equivalent problem can occur. In this case, the following can be
1482added to modprobe.conf (or modprobe.conf.local, as appropriate), as
1483follows (all on one line; it has been split here for clarity):
1484
1485install bonding /sbin/modprobe tg3; /sbin/modprobe e1000;
1486 /sbin/modprobe --ignore-install bonding
1487
1488 This will, when loading the bonding module, rather than
1489performing the normal action, instead execute the provided command.
1490This command loads the device drivers in the order needed, then calls
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001491modprobe with --ignore-install to cause the normal action to then take
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492place. Full documentation on this can be found in the modprobe.conf
1493and modprobe manual pages.
1494
Auke Kok6224e012006-06-08 11:15:35 -070014958.3. Painfully Slow Or No Failed Link Detection By Miimon
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496---------------------------------------------------------
1497
1498 By default, bonding enables the use_carrier option, which
1499instructs bonding to trust the driver to maintain carrier state.
1500
1501 As discussed in the options section, above, some drivers do
1502not support the netif_carrier_on/_off link state tracking system.
1503With use_carrier enabled, bonding will always see these links as up,
1504regardless of their actual state.
1505
1506 Additionally, other drivers do support netif_carrier, but do
1507not maintain it in real time, e.g., only polling the link state at
1508some fixed interval. In this case, miimon will detect failures, but
1509only after some long period of time has expired. If it appears that
1510miimon is very slow in detecting link failures, try specifying
1511use_carrier=0 to see if that improves the failure detection time. If
1512it does, then it may be that the driver checks the carrier state at a
1513fixed interval, but does not cache the MII register values (so the
1514use_carrier=0 method of querying the registers directly works). If
1515use_carrier=0 does not improve the failover, then the driver may cache
1516the registers, or the problem may be elsewhere.
1517
1518 Also, remember that miimon only checks for the device's
1519carrier state. It has no way to determine the state of devices on or
1520beyond other ports of a switch, or if a switch is refusing to pass
1521traffic while still maintaining carrier on.
1522
Auke Kok6224e012006-06-08 11:15:35 -070015239. SNMP agents
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524===============
1525
1526 If running SNMP agents, the bonding driver should be loaded
1527before any network drivers participating in a bond. This requirement
Tobias Klauserd533f672005-09-10 00:26:46 -07001528is due to the interface index (ipAdEntIfIndex) being associated to
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529the first interface found with a given IP address. That is, there is
1530only one ipAdEntIfIndex for each IP address. For example, if eth0 and
1531eth1 are slaves of bond0 and the driver for eth0 is loaded before the
1532bonding driver, the interface for the IP address will be associated
1533with the eth0 interface. This configuration is shown below, the IP
1534address 192.168.1.1 has an interface index of 2 which indexes to eth0
1535in the ifDescr table (ifDescr.2).
1536
1537 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1538 interfaces.ifTable.ifEntry.ifDescr.2 = eth0
1539 interfaces.ifTable.ifEntry.ifDescr.3 = eth1
1540 interfaces.ifTable.ifEntry.ifDescr.4 = eth2
1541 interfaces.ifTable.ifEntry.ifDescr.5 = eth3
1542 interfaces.ifTable.ifEntry.ifDescr.6 = bond0
1543 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
1544 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1545 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
1546 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1547
1548 This problem is avoided by loading the bonding driver before
1549any network drivers participating in a bond. Below is an example of
1550loading the bonding driver first, the IP address 192.168.1.1 is
1551correctly associated with ifDescr.2.
1552
1553 interfaces.ifTable.ifEntry.ifDescr.1 = lo
1554 interfaces.ifTable.ifEntry.ifDescr.2 = bond0
1555 interfaces.ifTable.ifEntry.ifDescr.3 = eth0
1556 interfaces.ifTable.ifEntry.ifDescr.4 = eth1
1557 interfaces.ifTable.ifEntry.ifDescr.5 = eth2
1558 interfaces.ifTable.ifEntry.ifDescr.6 = eth3
1559 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
1560 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1561 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
1562 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1563
1564 While some distributions may not report the interface name in
1565ifDescr, the association between the IP address and IfIndex remains
1566and SNMP functions such as Interface_Scan_Next will report that
1567association.
1568
Auke Kok6224e012006-06-08 11:15:35 -0700156910. Promiscuous mode
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570====================
1571
1572 When running network monitoring tools, e.g., tcpdump, it is
1573common to enable promiscuous mode on the device, so that all traffic
1574is seen (instead of seeing only traffic destined for the local host).
1575The bonding driver handles promiscuous mode changes to the bonding
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001576master device (e.g., bond0), and propagates the setting to the slave
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577devices.
1578
1579 For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001580the promiscuous mode setting is propagated to all slaves.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581
1582 For the active-backup, balance-tlb and balance-alb modes, the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001583promiscuous mode setting is propagated only to the active slave.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
1585 For balance-tlb mode, the active slave is the slave currently
1586receiving inbound traffic.
1587
1588 For balance-alb mode, the active slave is the slave used as a
1589"primary." This slave is used for mode-specific control traffic, for
1590sending to peers that are unassigned or if the load is unbalanced.
1591
1592 For the active-backup, balance-tlb and balance-alb modes, when
1593the active slave changes (e.g., due to a link failure), the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001594promiscuous setting will be propagated to the new active slave.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595
Auke Kok6224e012006-06-08 11:15:35 -0700159611. Configuring Bonding for High Availability
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001597=============================================
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
1599 High Availability refers to configurations that provide
1600maximum network availability by having redundant or backup devices,
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001601links or switches between the host and the rest of the world. The
1602goal is to provide the maximum availability of network connectivity
1603(i.e., the network always works), even though other configurations
1604could provide higher throughput.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
Auke Kok6224e012006-06-08 11:15:35 -0700160611.1 High Availability in a Single Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607--------------------------------------------------
1608
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001609 If two hosts (or a host and a single switch) are directly
1610connected via multiple physical links, then there is no availability
1611penalty to optimizing for maximum bandwidth. In this case, there is
1612only one switch (or peer), so if it fails, there is no alternative
1613access to fail over to. Additionally, the bonding load balance modes
1614support link monitoring of their members, so if individual links fail,
1615the load will be rebalanced across the remaining devices.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001617 See Section 13, "Configuring Bonding for Maximum Throughput"
1618for information on configuring bonding with one peer device.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619
Auke Kok6224e012006-06-08 11:15:35 -0700162011.2 High Availability in a Multiple Switch Topology
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621----------------------------------------------------
1622
1623 With multiple switches, the configuration of bonding and the
1624network changes dramatically. In multiple switch topologies, there is
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001625a trade off between network availability and usable bandwidth.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626
1627 Below is a sample network, configured to maximize the
1628availability of the network:
1629
1630 | |
1631 |port3 port3|
1632 +-----+----+ +-----+----+
1633 | |port2 ISL port2| |
1634 | switch A +--------------------------+ switch B |
1635 | | | |
1636 +-----+----+ +-----++---+
1637 |port1 port1|
1638 | +-------+ |
1639 +-------------+ host1 +---------------+
1640 eth0 +-------+ eth1
1641
1642 In this configuration, there is a link between the two
1643switches (ISL, or inter switch link), and multiple ports connecting to
1644the outside world ("port3" on each switch). There is no technical
1645reason that this could not be extended to a third switch.
1646
Auke Kok6224e012006-06-08 11:15:35 -0700164711.2.1 HA Bonding Mode Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001648-------------------------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001650 In a topology such as the example above, the active-backup and
1651broadcast modes are the only useful bonding modes when optimizing for
1652availability; the other modes require all links to terminate on the
1653same peer for them to behave rationally.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654
1655active-backup: This is generally the preferred mode, particularly if
1656 the switches have an ISL and play together well. If the
1657 network configuration is such that one switch is specifically
1658 a backup switch (e.g., has lower capacity, higher cost, etc),
1659 then the primary option can be used to insure that the
1660 preferred link is always used when it is available.
1661
1662broadcast: This mode is really a special purpose mode, and is suitable
1663 only for very specific needs. For example, if the two
1664 switches are not connected (no ISL), and the networks beyond
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001665 them are totally independent. In this case, if it is
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 necessary for some specific one-way traffic to reach both
1667 independent networks, then the broadcast mode may be suitable.
1668
Auke Kok6224e012006-06-08 11:15:35 -0700166911.2.2 HA Link Monitoring Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001670----------------------------------------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671
1672 The choice of link monitoring ultimately depends upon your
1673switch. If the switch can reliably fail ports in response to other
1674failures, then either the MII or ARP monitors should work. For
1675example, in the above example, if the "port3" link fails at the remote
1676end, the MII monitor has no direct means to detect this. The ARP
1677monitor could be configured with a target at the remote end of port3,
1678thus detecting that failure without switch support.
1679
1680 In general, however, in a multiple switch topology, the ARP
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001681monitor can provide a higher level of reliability in detecting end to
1682end connectivity failures (which may be caused by the failure of any
1683individual component to pass traffic for any reason). Additionally,
1684the ARP monitor should be configured with multiple targets (at least
1685one for each switch in the network). This will insure that,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686regardless of which switch is active, the ARP monitor has a suitable
1687target to query.
1688
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001689 Note, also, that of late many switches now support a functionality
1690generally referred to as "trunk failover." This is a feature of the
1691switch that causes the link state of a particular switch port to be set
1692down (or up) when the state of another switch port goes down (or up).
1693It's purpose is to propogate link failures from logically "exterior" ports
1694to the logically "interior" ports that bonding is able to monitor via
1695miimon. Availability and configuration for trunk failover varies by
1696switch, but this can be a viable alternative to the ARP monitor when using
1697suitable switches.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698
Auke Kok6224e012006-06-08 11:15:35 -0700169912. Configuring Bonding for Maximum Throughput
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001700==============================================
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701
Auke Kok6224e012006-06-08 11:15:35 -0700170212.1 Maximizing Throughput in a Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001703------------------------------------------------------
1704
1705 In a single switch configuration, the best method to maximize
1706throughput depends upon the application and network environment. The
1707various load balancing modes each have strengths and weaknesses in
1708different environments, as detailed below.
1709
1710 For this discussion, we will break down the topologies into
1711two categories. Depending upon the destination of most traffic, we
1712categorize them into either "gatewayed" or "local" configurations.
1713
1714 In a gatewayed configuration, the "switch" is acting primarily
1715as a router, and the majority of traffic passes through this router to
1716other networks. An example would be the following:
1717
1718
1719 +----------+ +----------+
1720 | |eth0 port1| | to other networks
1721 | Host A +---------------------+ router +------------------->
1722 | +---------------------+ | Hosts B and C are out
1723 | |eth1 port2| | here somewhere
1724 +----------+ +----------+
1725
1726 The router may be a dedicated router device, or another host
1727acting as a gateway. For our discussion, the important point is that
1728the majority of traffic from Host A will pass through the router to
1729some other network before reaching its final destination.
1730
1731 In a gatewayed network configuration, although Host A may
1732communicate with many other systems, all of its traffic will be sent
1733and received via one other peer on the local network, the router.
1734
1735 Note that the case of two systems connected directly via
1736multiple physical links is, for purposes of configuring bonding, the
1737same as a gatewayed configuration. In that case, it happens that all
1738traffic is destined for the "gateway" itself, not some other network
1739beyond the gateway.
1740
1741 In a local configuration, the "switch" is acting primarily as
1742a switch, and the majority of traffic passes through this switch to
1743reach other stations on the same network. An example would be the
1744following:
1745
1746 +----------+ +----------+ +--------+
1747 | |eth0 port1| +-------+ Host B |
1748 | Host A +------------+ switch |port3 +--------+
1749 | +------------+ | +--------+
1750 | |eth1 port2| +------------------+ Host C |
1751 +----------+ +----------+port4 +--------+
1752
1753
1754 Again, the switch may be a dedicated switch device, or another
1755host acting as a gateway. For our discussion, the important point is
1756that the majority of traffic from Host A is destined for other hosts
1757on the same local network (Hosts B and C in the above example).
1758
1759 In summary, in a gatewayed configuration, traffic to and from
1760the bonded device will be to the same MAC level peer on the network
1761(the gateway itself, i.e., the router), regardless of its final
1762destination. In a local configuration, traffic flows directly to and
1763from the final destinations, thus, each destination (Host B, Host C)
1764will be addressed directly by their individual MAC addresses.
1765
1766 This distinction between a gatewayed and a local network
1767configuration is important because many of the load balancing modes
1768available use the MAC addresses of the local network source and
1769destination to make load balancing decisions. The behavior of each
1770mode is described below.
1771
1772
Auke Kok6224e012006-06-08 11:15:35 -0700177312.1.1 MT Bonding Mode Selection for Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001774-----------------------------------------------------------
1775
1776 This configuration is the easiest to set up and to understand,
1777although you will have to decide which bonding mode best suits your
1778needs. The trade offs for each mode are detailed below:
1779
1780balance-rr: This mode is the only mode that will permit a single
1781 TCP/IP connection to stripe traffic across multiple
1782 interfaces. It is therefore the only mode that will allow a
1783 single TCP/IP stream to utilize more than one interface's
1784 worth of throughput. This comes at a cost, however: the
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001785 striping generally results in peer systems receiving packets out
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001786 of order, causing TCP/IP's congestion control system to kick
1787 in, often by retransmitting segments.
1788
1789 It is possible to adjust TCP/IP's congestion limits by
1790 altering the net.ipv4.tcp_reordering sysctl parameter. The
1791 usual default value is 3, and the maximum useful value is 127.
1792 For a four interface balance-rr bond, expect that a single
1793 TCP/IP stream will utilize no more than approximately 2.3
1794 interface's worth of throughput, even after adjusting
1795 tcp_reordering.
1796
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08001797 Note that the fraction of packets that will be delivered out of
1798 order is highly variable, and is unlikely to be zero. The level
1799 of reordering depends upon a variety of factors, including the
1800 networking interfaces, the switch, and the topology of the
1801 configuration. Speaking in general terms, higher speed network
1802 cards produce more reordering (due to factors such as packet
1803 coalescing), and a "many to many" topology will reorder at a
1804 higher rate than a "many slow to one fast" configuration.
1805
1806 Many switches do not support any modes that stripe traffic
1807 (instead choosing a port based upon IP or MAC level addresses);
1808 for those devices, traffic for a particular connection flowing
1809 through the switch to a balance-rr bond will not utilize greater
1810 than one interface's worth of bandwidth.
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001811
1812 If you are utilizing protocols other than TCP/IP, UDP for
1813 example, and your application can tolerate out of order
1814 delivery, then this mode can allow for single stream datagram
1815 performance that scales near linearly as interfaces are added
1816 to the bond.
1817
1818 This mode requires the switch to have the appropriate ports
1819 configured for "etherchannel" or "trunking."
1820
1821active-backup: There is not much advantage in this network topology to
1822 the active-backup mode, as the inactive backup devices are all
1823 connected to the same peer as the primary. In this case, a
1824 load balancing mode (with link monitoring) will provide the
1825 same level of network availability, but with increased
1826 available bandwidth. On the plus side, active-backup mode
1827 does not require any configuration of the switch, so it may
1828 have value if the hardware available does not support any of
1829 the load balance modes.
1830
1831balance-xor: This mode will limit traffic such that packets destined
1832 for specific peers will always be sent over the same
1833 interface. Since the destination is determined by the MAC
1834 addresses involved, this mode works best in a "local" network
1835 configuration (as described above), with destinations all on
1836 the same local network. This mode is likely to be suboptimal
1837 if all your traffic is passed through a single router (i.e., a
1838 "gatewayed" network configuration, as described above).
1839
1840 As with balance-rr, the switch ports need to be configured for
1841 "etherchannel" or "trunking."
1842
1843broadcast: Like active-backup, there is not much advantage to this
1844 mode in this type of network topology.
1845
1846802.3ad: This mode can be a good choice for this type of network
1847 topology. The 802.3ad mode is an IEEE standard, so all peers
1848 that implement 802.3ad should interoperate well. The 802.3ad
1849 protocol includes automatic configuration of the aggregates,
1850 so minimal manual configuration of the switch is needed
1851 (typically only to designate that some set of devices is
1852 available for 802.3ad). The 802.3ad standard also mandates
1853 that frames be delivered in order (within certain limits), so
1854 in general single connections will not see misordering of
1855 packets. The 802.3ad mode does have some drawbacks: the
1856 standard mandates that all devices in the aggregate operate at
1857 the same speed and duplex. Also, as with all bonding load
1858 balance modes other than balance-rr, no single connection will
1859 be able to utilize more than a single interface's worth of
1860 bandwidth.
1861
1862 Additionally, the linux bonding 802.3ad implementation
1863 distributes traffic by peer (using an XOR of MAC addresses),
1864 so in a "gatewayed" configuration, all outgoing traffic will
1865 generally use the same device. Incoming traffic may also end
1866 up on a single device, but that is dependent upon the
1867 balancing policy of the peer's 8023.ad implementation. In a
1868 "local" configuration, traffic will be distributed across the
1869 devices in the bond.
1870
1871 Finally, the 802.3ad mode mandates the use of the MII monitor,
1872 therefore, the ARP monitor is not available in this mode.
1873
1874balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
1875 Since the balancing is done according to MAC address, in a
1876 "gatewayed" configuration (as described above), this mode will
1877 send all traffic across a single device. However, in a
1878 "local" network configuration, this mode balances multiple
1879 local network peers across devices in a vaguely intelligent
1880 manner (not a simple XOR as in balance-xor or 802.3ad mode),
1881 so that mathematically unlucky MAC addresses (i.e., ones that
1882 XOR to the same value) will not all "bunch up" on a single
1883 interface.
1884
1885 Unlike 802.3ad, interfaces may be of differing speeds, and no
1886 special switch configuration is required. On the down side,
1887 in this mode all incoming traffic arrives over a single
1888 interface, this mode requires certain ethtool support in the
1889 network device driver of the slave interfaces, and the ARP
1890 monitor is not available.
1891
1892balance-alb: This mode is everything that balance-tlb is, and more.
1893 It has all of the features (and restrictions) of balance-tlb,
1894 and will also balance incoming traffic from local network
1895 peers (as described in the Bonding Module Options section,
1896 above).
1897
1898 The only additional down side to this mode is that the network
1899 device driver must support changing the hardware address while
1900 the device is open.
1901
Auke Kok6224e012006-06-08 11:15:35 -0700190212.1.2 MT Link Monitoring for Single Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001903----------------------------------------------------
1904
1905 The choice of link monitoring may largely depend upon which
1906mode you choose to use. The more advanced load balancing modes do not
1907support the use of the ARP monitor, and are thus restricted to using
1908the MII monitor (which does not provide as high a level of end to end
1909assurance as the ARP monitor).
1910
Auke Kok6224e012006-06-08 11:15:35 -0700191112.2 Maximum Throughput in a Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001912-----------------------------------------------------
1913
1914 Multiple switches may be utilized to optimize for throughput
1915when they are configured in parallel as part of an isolated network
1916between two or more systems, for example:
1917
1918 +-----------+
1919 | Host A |
1920 +-+---+---+-+
1921 | | |
1922 +--------+ | +---------+
1923 | | |
1924 +------+---+ +-----+----+ +-----+----+
1925 | Switch A | | Switch B | | Switch C |
1926 +------+---+ +-----+----+ +-----+----+
1927 | | |
1928 +--------+ | +---------+
1929 | | |
1930 +-+---+---+-+
1931 | Host B |
1932 +-----------+
1933
1934 In this configuration, the switches are isolated from one
1935another. One reason to employ a topology such as this is for an
1936isolated network with many hosts (a cluster configured for high
1937performance, for example), using multiple smaller switches can be more
1938cost effective than a single larger switch, e.g., on a network with 24
1939hosts, three 24 port switches can be significantly less expensive than
1940a single 72 port switch.
1941
1942 If access beyond the network is required, an individual host
1943can be equipped with an additional network device connected to an
1944external network; this host then additionally acts as a gateway.
1945
Auke Kok6224e012006-06-08 11:15:35 -0700194612.2.1 MT Bonding Mode Selection for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001947-------------------------------------------------------------
1948
1949 In actual practice, the bonding mode typically employed in
1950configurations of this type is balance-rr. Historically, in this
1951network configuration, the usual caveats about out of order packet
1952delivery are mitigated by the use of network adapters that do not do
1953any kind of packet coalescing (via the use of NAPI, or because the
1954device itself does not generate interrupts until some number of
1955packets has arrived). When employed in this fashion, the balance-rr
1956mode allows individual connections between two hosts to effectively
1957utilize greater than one interface's bandwidth.
1958
Auke Kok6224e012006-06-08 11:15:35 -0700195912.2.2 MT Link Monitoring for Multiple Switch Topology
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001960------------------------------------------------------
1961
1962 Again, in actual practice, the MII monitor is most often used
1963in this configuration, as performance is given preference over
1964availability. The ARP monitor will function in this topology, but its
1965advantages over the MII monitor are mitigated by the volume of probes
1966needed as the number of systems involved grows (remember that each
1967host in the network is configured with bonding).
1968
Auke Kok6224e012006-06-08 11:15:35 -0700196913. Switch Behavior Issues
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001970==========================
1971
Auke Kok6224e012006-06-08 11:15:35 -0700197213.1 Link Establishment and Failover Delays
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001973-------------------------------------------
1974
1975 Some switches exhibit undesirable behavior with regard to the
1976timing of link up and down reporting by the switch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977
1978 First, when a link comes up, some switches may indicate that
1979the link is up (carrier available), but not pass traffic over the
1980interface for some period of time. This delay is typically due to
1981some type of autonegotiation or routing protocol, but may also occur
1982during switch initialization (e.g., during recovery after a switch
1983failure). If you find this to be a problem, specify an appropriate
1984value to the updelay bonding module option to delay the use of the
1985relevant interface(s).
1986
1987 Second, some switches may "bounce" the link state one or more
1988times while a link is changing state. This occurs most commonly while
1989the switch is initializing. Again, an appropriate updelay value may
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001990help.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991
1992 Note that when a bonding interface has no active links, the
Jay Vosburgh00354cf2005-07-21 12:18:02 -07001993driver will immediately reuse the first link that goes up, even if the
1994updelay parameter has been specified (the updelay is ignored in this
1995case). If there are slave interfaces waiting for the updelay timeout
1996to expire, the interface that first went into that state will be
1997immediately reused. This reduces down time of the network if the
1998value of updelay has been overestimated, and since this occurs only in
1999cases with no connectivity, there is no additional penalty for
2000ignoring the updelay.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001
2002 In addition to the concerns about switch timings, if your
2003switches take a long time to go into backup mode, it may be desirable
2004to not activate a backup interface immediately after a link goes down.
2005Failover may be delayed via the downdelay bonding module option.
2006
Auke Kok6224e012006-06-08 11:15:35 -0700200713.2 Duplicated Incoming Packets
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002008--------------------------------
2009
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002010 NOTE: Starting with version 3.0.2, the bonding driver has logic to
2011suppress duplicate packets, which should largely eliminate this problem.
2012The following description is kept for reference.
2013
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002014 It is not uncommon to observe a short burst of duplicated
2015traffic when the bonding device is first used, or after it has been
2016idle for some period of time. This is most easily observed by issuing
2017a "ping" to some other host on the network, and noticing that the
2018output from ping flags duplicates (typically one per slave).
2019
2020 For example, on a bond in active-backup mode with five slaves
2021all connected to one switch, the output may appear as follows:
2022
2023# ping -n 10.0.4.2
2024PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
202564 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
202664 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
202764 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
202864 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
202964 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
203064 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
203164 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
203264 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2033
2034 This is not due to an error in the bonding driver, rather, it
2035is a side effect of how many switches update their MAC forwarding
2036tables. Initially, the switch does not associate the MAC address in
2037the packet with a particular switch port, and so it may send the
2038traffic to all ports until its MAC forwarding table is updated. Since
2039the interfaces attached to the bond may occupy multiple ports on a
2040single switch, when the switch (temporarily) floods the traffic to all
2041ports, the bond device receives multiple copies of the same packet
2042(one per slave device).
2043
2044 The duplicated packet behavior is switch dependent, some
2045switches exhibit this, and some do not. On switches that display this
2046behavior, it can be induced by clearing the MAC forwarding table (on
2047most Cisco switches, the privileged command "clear mac address-table
2048dynamic" will accomplish this).
2049
Auke Kok6224e012006-06-08 11:15:35 -0700205014. Hardware Specific Considerations
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051====================================
2052
2053 This section contains additional information for configuring
2054bonding on specific hardware platforms, or for interfacing bonding
2055with particular switches or other devices.
2056
Auke Kok6224e012006-06-08 11:15:35 -0700205714.1 IBM BladeCenter
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058--------------------
2059
2060 This applies to the JS20 and similar systems.
2061
2062 On the JS20 blades, the bonding driver supports only
2063balance-rr, active-backup, balance-tlb and balance-alb modes. This is
2064largely due to the network topology inside the BladeCenter, detailed
2065below.
2066
2067JS20 network adapter information
2068--------------------------------
2069
2070 All JS20s come with two Broadcom Gigabit Ethernet ports
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002071integrated on the planar (that's "motherboard" in IBM-speak). In the
2072BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2073I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2074An add-on Broadcom daughter card can be installed on a JS20 to provide
2075two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
2076wired to I/O Modules 3 and 4, respectively.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077
2078 Each I/O Module may contain either a switch or a passthrough
2079module (which allows ports to be directly connected to an external
2080switch). Some bonding modes require a specific BladeCenter internal
2081network topology in order to function; these are detailed below.
2082
2083 Additional BladeCenter-specific networking information can be
2084found in two IBM Redbooks (www.ibm.com/redbooks):
2085
2086"IBM eServer BladeCenter Networking Options"
2087"IBM eServer BladeCenter Layer 2-7 Network Switching"
2088
2089BladeCenter networking configuration
2090------------------------------------
2091
2092 Because a BladeCenter can be configured in a very large number
2093of ways, this discussion will be confined to describing basic
2094configurations.
2095
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002096 Normally, Ethernet Switch Modules (ESMs) are used in I/O
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
2098JS20 will be connected to different internal switches (in the
2099respective I/O modules).
2100
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002101 A passthrough module (OPM or CPM, optical or copper,
2102passthrough module) connects the I/O module directly to an external
2103switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
2104interfaces of a JS20 can be redirected to the outside world and
2105connected to a common external switch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002107 Depending upon the mix of ESMs and PMs, the network will
2108appear to bonding as either a single switch topology (all PMs) or as a
2109multiple switch topology (one or more ESMs, zero or more PMs). It is
2110also possible to connect ESMs together, resulting in a configuration
2111much like the example in "High Availability in a Multiple Switch
2112Topology," above.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002114Requirements for specific modes
2115-------------------------------
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002117 The balance-rr mode requires the use of passthrough modules
2118for devices in the bond, all connected to an common external switch.
2119That switch must be configured for "etherchannel" or "trunking" on the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120appropriate ports, as is usual for balance-rr.
2121
2122 The balance-alb and balance-tlb modes will function with
2123either switch modules or passthrough modules (or a mix). The only
2124specific requirement for these modes is that all network interfaces
2125must be able to reach all destinations for traffic sent over the
2126bonding device (i.e., the network must converge at some point outside
2127the BladeCenter).
2128
2129 The active-backup mode has no additional requirements.
2130
2131Link monitoring issues
2132----------------------
2133
2134 When an Ethernet Switch Module is in place, only the ARP
2135monitor will reliably detect link loss to an external switch. This is
2136nothing unusual, but examination of the BladeCenter cabinet would
2137suggest that the "external" network ports are the ethernet ports for
2138the system, when it fact there is a switch between these "external"
2139ports and the devices on the JS20 system itself. The MII monitor is
2140only able to detect link failures between the ESM and the JS20 system.
2141
2142 When a passthrough module is in place, the MII monitor does
2143detect failures to the "external" port, which is then directly
2144connected to the JS20 system.
2145
2146Other concerns
2147--------------
2148
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002149 The Serial Over LAN (SoL) link is established over the primary
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150ethernet (eth0) only, therefore, any loss of link to eth0 will result
2151in losing your SoL connection. It will not fail over with other
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002152network traffic, as the SoL system is beyond the control of the
2153bonding driver.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154
2155 It may be desirable to disable spanning tree on the switch
2156(either the internal Ethernet Switch Module, or an external switch) to
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002157avoid fail-over delay issues when using bonding.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158
2159
Auke Kok6224e012006-06-08 11:15:35 -0700216015. Frequently Asked Questions
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161==============================
2162
21631. Is it SMP safe?
2164
2165 Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2166The new driver was designed to be SMP safe from the start.
2167
21682. What type of cards will work with it?
2169
2170 Any Ethernet type cards (you can even mix cards - a Intel
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002171EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
2172devices need not be of the same speed.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002174 Starting with version 3.2.1, bonding also supports Infiniband
2175slaves in active-backup mode.
2176
Linus Torvalds1da177e2005-04-16 15:20:36 -070021773. How many bonding devices can I have?
2178
2179 There is no limit.
2180
21814. How many slaves can a bonding device have?
2182
2183 This is limited only by the number of network interfaces Linux
2184supports and/or the number of network cards you can place in your
2185system.
2186
21875. What happens when a slave link dies?
2188
2189 If link monitoring is enabled, then the failing device will be
2190disabled. The active-backup mode will fail over to a backup link, and
2191other modes will ignore the failed link. The link will continue to be
2192monitored, and should it recover, it will rejoin the bond (in whatever
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002193manner is appropriate for the mode). See the sections on High
2194Availability and the documentation for each mode for additional
2195information.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196
2197 Link monitoring can be enabled via either the miimon or
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002198arp_interval parameters (described in the module parameters section,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199above). In general, miimon monitors the carrier state as sensed by
2200the underlying network device, and the arp monitor (arp_interval)
2201monitors connectivity to another host on the local network.
2202
2203 If no link monitoring is configured, the bonding driver will
2204be unable to detect link failures, and will assume that all links are
2205always available. This will likely result in lost packets, and a
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002206resulting degradation of performance. The precise performance loss
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207depends upon the bonding mode and network configuration.
2208
22096. Can bonding be used for High Availability?
2210
2211 Yes. See the section on High Availability for details.
2212
22137. Which switches/systems does it work with?
2214
2215 The full answer to this depends upon the desired mode.
2216
2217 In the basic balance modes (balance-rr and balance-xor), it
2218works with any system that supports etherchannel (also called
2219trunking). Most managed switches currently available have such
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002220support, and many unmanaged switches as well.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221
2222 The advanced balance modes (balance-tlb and balance-alb) do
2223not have special switch requirements, but do need device drivers that
2224support specific features (described in the appropriate section under
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002225module parameters, above).
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226
Auke Kok6224e012006-06-08 11:15:35 -07002227 In 802.3ad mode, it works with systems that support IEEE
Linus Torvalds1da177e2005-04-16 15:20:36 -07002228802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
2229switches currently available support 802.3ad.
2230
2231 The active-backup mode should work with any Layer-II switch.
2232
22338. Where does a bonding device get its MAC address from?
2234
Jay Vosburgh9a6c6862007-11-13 20:25:48 -08002235 When using slave devices that have fixed MAC addresses, or when
2236the fail_over_mac option is enabled, the bonding device's MAC address is
2237the MAC address of the active slave.
2238
2239 For other configurations, if not explicitly configured (with
2240ifconfig or ip link), the MAC address of the bonding device is taken from
2241its first slave device. This MAC address is then passed to all following
2242slaves and remains persistent (even if the first slave is removed) until
2243the bonding device is brought down or reconfigured.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244
2245 If you wish to change the MAC address, you can set it with
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002246ifconfig or ip link:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247
2248# ifconfig bond0 hw ether 00:11:22:33:44:55
2249
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002250# ip link set bond0 address 66:77:88:99:aa:bb
2251
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252 The MAC address can be also changed by bringing down/up the
2253device and then changing its slaves (or their order):
2254
2255# ifconfig bond0 down ; modprobe -r bonding
2256# ifconfig bond0 .... up
2257# ifenslave bond0 eth...
2258
2259 This method will automatically take the address from the next
2260slave that is added.
2261
2262 To restore your slaves' MAC addresses, you need to detach them
2263from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2264then restore the MAC addresses that the slaves had before they were
2265enslaved.
2266
Jay Vosburgh00354cf2005-07-21 12:18:02 -0700226716. Resources and Links
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268=======================
2269
2270The latest version of the bonding driver can be found in the latest
2271version of the linux kernel, found on http://kernel.org
2272
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002273The latest version of this document can be found in either the latest
2274kernel source (named Documentation/networking/bonding.txt), or on the
2275bonding sourceforge site:
2276
2277http://www.sourceforge.net/projects/bonding
2278
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279Discussions regarding the bonding driver take place primarily on the
2280bonding-devel mailing list, hosted at sourceforge.net. If you have
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002281questions or problems, post them to the list. The list address is:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282
2283bonding-devel@lists.sourceforge.net
2284
Jay Vosburgh00354cf2005-07-21 12:18:02 -07002285 The administrative interface (to subscribe or unsubscribe) can
2286be found at:
2287
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288https://lists.sourceforge.net/lists/listinfo/bonding-devel
2289
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290Donald Becker's Ethernet Drivers and diag programs may be found at :
2291 - http://www.scyld.com/network/
2292
2293You will also find a lot of information regarding Ethernet, NWay, MII,
2294etc. at www.scyld.com.
2295
2296-- END --