]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | |
00354cfb JV |
2 | Linux Ethernet Bonding Driver HOWTO |
3 | ||
a549952a | 4 | Latest update: 23 September 2009 |
1da177e4 LT |
5 | |
6 | Initial release : Thomas Davis <tadavis at lbl.gov> | |
7 | Corrections, HA extensions : 2000/10/03-15 : | |
8 | - Willy Tarreau <willy at meta-x.org> | |
9 | - Constantine Gavrilov <const-g at xpert.com> | |
10 | - Chad N. Tindel <ctindel at ieee dot org> | |
11 | - Janice Girouard <girouard at us dot ibm dot com> | |
12 | - Jay Vosburgh <fubar at us dot ibm dot com> | |
13 | ||
14 | Reorganized and updated Feb 2005 by Jay Vosburgh | |
6224e01d AK |
15 | Added Sysfs information: 2006/04/24 |
16 | - Mitch Williams <mitch.a.williams at intel.com> | |
1da177e4 | 17 | |
00354cfb JV |
18 | Introduction |
19 | ============ | |
20 | ||
21 | The Linux bonding driver provides a method for aggregating | |
22 | multiple network interfaces into a single logical "bonded" interface. | |
23 | The behavior of the bonded interfaces depends upon the mode; generally | |
24 | speaking, modes provide either hot standby or load balancing services. | |
25 | Additionally, link integrity monitoring may be performed. | |
1da177e4 | 26 | |
00354cfb JV |
27 | The bonding driver originally came from Donald Becker's |
28 | beowulf patches for kernel 2.0. It has changed quite a bit since, and | |
29 | the original tools from extreme-linux and beowulf sites will not work | |
30 | with this version of the driver. | |
1da177e4 | 31 | |
00354cfb JV |
32 | For new versions of the driver, updated userspace tools, and |
33 | who to ask for help, please follow the links at the end of this file. | |
1da177e4 LT |
34 | |
35 | Table of Contents | |
36 | ================= | |
37 | ||
38 | 1. Bonding Driver Installation | |
39 | ||
40 | 2. Bonding Driver Options | |
41 | ||
42 | 3. Configuring Bonding Devices | |
6224e01d AK |
43 | 3.1 Configuration with Sysconfig Support |
44 | 3.1.1 Using DHCP with Sysconfig | |
45 | 3.1.2 Configuring Multiple Bonds with Sysconfig | |
46 | 3.2 Configuration with Initscripts Support | |
47 | 3.2.1 Using DHCP with Initscripts | |
48 | 3.2.2 Configuring Multiple Bonds with Initscripts | |
49 | 3.3 Configuring Bonding Manually with Ifenslave | |
00354cfb | 50 | 3.3.1 Configuring Multiple Bonds Manually |
6224e01d | 51 | 3.4 Configuring Bonding Manually via Sysfs |
bb1d9123 | 52 | 3.5 Overriding Configuration for Special Cases |
1da177e4 | 53 | |
6224e01d AK |
54 | 4. Querying Bonding Configuration |
55 | 4.1 Bonding Configuration | |
56 | 4.2 Network Configuration | |
1da177e4 | 57 | |
6224e01d | 58 | 5. Switch Configuration |
1da177e4 | 59 | |
6224e01d | 60 | 6. 802.1q VLAN Support |
1da177e4 | 61 | |
6224e01d AK |
62 | 7. Link Monitoring |
63 | 7.1 ARP Monitor Operation | |
64 | 7.2 Configuring Multiple ARP Targets | |
65 | 7.3 MII Monitor Operation | |
1da177e4 | 66 | |
6224e01d AK |
67 | 8. Potential Trouble Sources |
68 | 8.1 Adventures in Routing | |
69 | 8.2 Ethernet Device Renaming | |
70 | 8.3 Painfully Slow Or No Failed Link Detection By Miimon | |
1da177e4 | 71 | |
6224e01d | 72 | 9. SNMP agents |
1da177e4 | 73 | |
6224e01d | 74 | 10. Promiscuous mode |
1da177e4 | 75 | |
6224e01d AK |
76 | 11. Configuring Bonding for High Availability |
77 | 11.1 High Availability in a Single Switch Topology | |
78 | 11.2 High Availability in a Multiple Switch Topology | |
79 | 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology | |
80 | 11.2.2 HA Link Monitoring for Multiple Switch Topology | |
00354cfb | 81 | |
6224e01d AK |
82 | 12. Configuring Bonding for Maximum Throughput |
83 | 12.1 Maximum Throughput in a Single Switch Topology | |
84 | 12.1.1 MT Bonding Mode Selection for Single Switch Topology | |
85 | 12.1.2 MT Link Monitoring for Single Switch Topology | |
86 | 12.2 Maximum Throughput in a Multiple Switch Topology | |
87 | 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology | |
88 | 12.2.2 MT Link Monitoring for Multiple Switch Topology | |
1da177e4 | 89 | |
6224e01d AK |
90 | 13. Switch Behavior Issues |
91 | 13.1 Link Establishment and Failover Delays | |
92 | 13.2 Duplicated Incoming Packets | |
1da177e4 | 93 | |
6224e01d AK |
94 | 14. Hardware Specific Considerations |
95 | 14.1 IBM BladeCenter | |
1da177e4 | 96 | |
6224e01d | 97 | 15. Frequently Asked Questions |
00354cfb | 98 | |
6224e01d | 99 | 16. Resources and Links |
1da177e4 LT |
100 | |
101 | ||
102 | 1. Bonding Driver Installation | |
103 | ============================== | |
104 | ||
105 | Most popular distro kernels ship with the bonding driver | |
106 | already available as a module and the ifenslave user level control | |
107 | program installed and ready for use. If your distro does not, or you | |
108 | have need to compile bonding from source (e.g., configuring and | |
109 | installing a mainline kernel from kernel.org), you'll need to perform | |
110 | the following steps: | |
111 | ||
112 | 1.1 Configure and build the kernel with bonding | |
113 | ----------------------------------------------- | |
114 | ||
00354cfb | 115 | The current version of the bonding driver is available in the |
1da177e4 | 116 | drivers/net/bonding subdirectory of the most recent kernel source |
00354cfb JV |
117 | (which is available on http://kernel.org). Most users "rolling their |
118 | own" will want to use the most recent kernel from kernel.org. | |
1da177e4 LT |
119 | |
120 | Configure kernel with "make menuconfig" (or "make xconfig" or | |
121 | "make config"), then select "Bonding driver support" in the "Network | |
122 | device support" section. It is recommended that you configure the | |
123 | driver as module since it is currently the only way to pass parameters | |
124 | to the driver or configure more than one bonding device. | |
125 | ||
00354cfb JV |
126 | Build and install the new kernel and modules, then continue |
127 | below to install ifenslave. | |
1da177e4 LT |
128 | |
129 | 1.2 Install ifenslave Control Utility | |
130 | ------------------------------------- | |
131 | ||
132 | The ifenslave user level control program is included in the | |
133 | kernel source tree, in the file Documentation/networking/ifenslave.c. | |
134 | It is generally recommended that you use the ifenslave that | |
135 | corresponds to the kernel that you are using (either from the same | |
136 | source tree or supplied with the distro), however, ifenslave | |
137 | executables from older kernels should function (but features newer | |
138 | than the ifenslave release are not supported). Running an ifenslave | |
139 | that is newer than the kernel is not supported, and may or may not | |
140 | work. | |
141 | ||
142 | To install ifenslave, do the following: | |
143 | ||
144 | # gcc -Wall -O -I/usr/src/linux/include ifenslave.c -o ifenslave | |
145 | # cp ifenslave /sbin/ifenslave | |
146 | ||
147 | If your kernel source is not in "/usr/src/linux," then replace | |
148 | "/usr/src/linux/include" in the above with the location of your kernel | |
149 | source include directory. | |
150 | ||
151 | You may wish to back up any existing /sbin/ifenslave, or, for | |
152 | testing or informal use, tag the ifenslave to the kernel version | |
153 | (e.g., name the ifenslave executable /sbin/ifenslave-2.6.10). | |
154 | ||
155 | IMPORTANT NOTE: | |
156 | ||
157 | If you omit the "-I" or specify an incorrect directory, you | |
158 | may end up with an ifenslave that is incompatible with the kernel | |
159 | you're trying to build it for. Some distros (e.g., Red Hat from 7.1 | |
160 | onwards) do not have /usr/include/linux symbolically linked to the | |
161 | default kernel source include directory. | |
162 | ||
6224e01d AK |
163 | SECOND IMPORTANT NOTE: |
164 | If you plan to configure bonding using sysfs, you do not need | |
165 | to use ifenslave. | |
1da177e4 LT |
166 | |
167 | 2. Bonding Driver Options | |
168 | ========================= | |
169 | ||
9a6c6867 JV |
170 | Options for the bonding driver are supplied as parameters to the |
171 | bonding module at load time, or are specified via sysfs. | |
172 | ||
173 | Module options may be given as command line arguments to the | |
174 | insmod or modprobe command, but are usually specified in either the | |
175 | /etc/modules.conf or /etc/modprobe.conf configuration file, or in a | |
176 | distro-specific configuration file (some of which are detailed in the next | |
177 | section). | |
178 | ||
179 | Details on bonding support for sysfs is provided in the | |
180 | "Configuring Bonding Manually via Sysfs" section, below. | |
1da177e4 LT |
181 | |
182 | The available bonding driver parameters are listed below. If a | |
183 | parameter is not specified the default value is used. When initially | |
184 | configuring a bond, it is recommended "tail -f /var/log/messages" be | |
185 | run in a separate window to watch for bonding driver error messages. | |
186 | ||
187 | It is critical that either the miimon or arp_interval and | |
188 | arp_ip_target parameters be specified, otherwise serious network | |
189 | degradation will occur during link failures. Very few devices do not | |
190 | support at least miimon, so there is really no reason not to use it. | |
191 | ||
192 | Options with textual values will accept either the text name | |
00354cfb JV |
193 | or, for backwards compatibility, the option value. E.g., |
194 | "mode=802.3ad" and "mode=4" set the same mode. | |
1da177e4 LT |
195 | |
196 | The parameters are as follows: | |
197 | ||
fd989c83 JV |
198 | ad_select |
199 | ||
200 | Specifies the 802.3ad aggregation selection logic to use. The | |
201 | possible values and their effects are: | |
202 | ||
203 | stable or 0 | |
204 | ||
205 | The active aggregator is chosen by largest aggregate | |
206 | bandwidth. | |
207 | ||
208 | Reselection of the active aggregator occurs only when all | |
209 | slaves of the active aggregator are down or the active | |
210 | aggregator has no slaves. | |
211 | ||
212 | This is the default value. | |
213 | ||
214 | bandwidth or 1 | |
215 | ||
216 | The active aggregator is chosen by largest aggregate | |
217 | bandwidth. Reselection occurs if: | |
218 | ||
219 | - A slave is added to or removed from the bond | |
220 | ||
221 | - Any slave's link state changes | |
222 | ||
223 | - Any slave's 802.3ad association state changes | |
224 | ||
19f59460 | 225 | - The bond's administrative state changes to up |
fd989c83 JV |
226 | |
227 | count or 2 | |
228 | ||
229 | The active aggregator is chosen by the largest number of | |
230 | ports (slaves). Reselection occurs as described under the | |
231 | "bandwidth" setting, above. | |
232 | ||
233 | The bandwidth and count selection policies permit failover of | |
234 | 802.3ad aggregations when partial failure of the active aggregator | |
235 | occurs. This keeps the aggregator with the highest availability | |
236 | (either in bandwidth or in number of ports) active at all times. | |
237 | ||
238 | This option was added in bonding version 3.4.0. | |
239 | ||
1da177e4 LT |
240 | arp_interval |
241 | ||
00354cfb | 242 | Specifies the ARP link monitoring frequency in milliseconds. |
f5b2b966 JV |
243 | |
244 | The ARP monitor works by periodically checking the slave | |
245 | devices to determine whether they have sent or received | |
246 | traffic recently (the precise criteria depends upon the | |
247 | bonding mode, and the state of the slave). Regular traffic is | |
248 | generated via ARP probes issued for the addresses specified by | |
249 | the arp_ip_target option. | |
250 | ||
251 | This behavior can be modified by the arp_validate option, | |
252 | below. | |
253 | ||
00354cfb JV |
254 | If ARP monitoring is used in an etherchannel compatible mode |
255 | (modes 0 and 2), the switch should be configured in a mode | |
256 | that evenly distributes packets across all links. If the | |
257 | switch is configured to distribute the packets in an XOR | |
1da177e4 LT |
258 | fashion, all replies from the ARP targets will be received on |
259 | the same link which could cause the other team members to | |
00354cfb JV |
260 | fail. ARP monitoring should not be used in conjunction with |
261 | miimon. A value of 0 disables ARP monitoring. The default | |
1da177e4 LT |
262 | value is 0. |
263 | ||
264 | arp_ip_target | |
265 | ||
00354cfb JV |
266 | Specifies the IP addresses to use as ARP monitoring peers when |
267 | arp_interval is > 0. These are the targets of the ARP request | |
268 | sent to determine the health of the link to the targets. | |
269 | Specify these values in ddd.ddd.ddd.ddd format. Multiple IP | |
270 | addresses must be separated by a comma. At least one IP | |
271 | address must be given for ARP monitoring to function. The | |
272 | maximum number of targets that can be specified is 16. The | |
273 | default value is no IP addresses. | |
1da177e4 | 274 | |
f5b2b966 JV |
275 | arp_validate |
276 | ||
277 | Specifies whether or not ARP probes and replies should be | |
278 | validated in the active-backup mode. This causes the ARP | |
279 | monitor to examine the incoming ARP requests and replies, and | |
280 | only consider a slave to be up if it is receiving the | |
281 | appropriate ARP traffic. | |
282 | ||
283 | Possible values are: | |
284 | ||
285 | none or 0 | |
286 | ||
287 | No validation is performed. This is the default. | |
288 | ||
289 | active or 1 | |
290 | ||
291 | Validation is performed only for the active slave. | |
292 | ||
293 | backup or 2 | |
294 | ||
295 | Validation is performed only for backup slaves. | |
296 | ||
297 | all or 3 | |
298 | ||
299 | Validation is performed for all slaves. | |
300 | ||
301 | For the active slave, the validation checks ARP replies to | |
302 | confirm that they were generated by an arp_ip_target. Since | |
303 | backup slaves do not typically receive these replies, the | |
304 | validation performed for backup slaves is on the ARP request | |
305 | sent out via the active slave. It is possible that some | |
306 | switch or network configurations may result in situations | |
307 | wherein the backup slaves do not receive the ARP requests; in | |
308 | such a situation, validation of backup slaves must be | |
309 | disabled. | |
310 | ||
311 | This option is useful in network configurations in which | |
312 | multiple bonding hosts are concurrently issuing ARPs to one or | |
313 | more targets beyond a common switch. Should the link between | |
314 | the switch and target fail (but not the switch itself), the | |
315 | probe traffic generated by the multiple bonding instances will | |
316 | fool the standard ARP monitor into considering the links as | |
317 | still up. Use of the arp_validate option can resolve this, as | |
318 | the ARP monitor will only consider ARP requests and replies | |
319 | associated with its own instance of bonding. | |
320 | ||
321 | This option was added in bonding version 3.1.0. | |
322 | ||
1da177e4 LT |
323 | downdelay |
324 | ||
325 | Specifies the time, in milliseconds, to wait before disabling | |
326 | a slave after a link failure has been detected. This option | |
327 | is only valid for the miimon link monitor. The downdelay | |
328 | value should be a multiple of the miimon value; if not, it | |
329 | will be rounded down to the nearest multiple. The default | |
330 | value is 0. | |
331 | ||
dd957c57 JV |
332 | fail_over_mac |
333 | ||
334 | Specifies whether active-backup mode should set all slaves to | |
3915c1e8 JV |
335 | the same MAC address at enslavement (the traditional |
336 | behavior), or, when enabled, perform special handling of the | |
337 | bond's MAC address in accordance with the selected policy. | |
338 | ||
339 | Possible values are: | |
340 | ||
341 | none or 0 | |
342 | ||
343 | This setting disables fail_over_mac, and causes | |
344 | bonding to set all slaves of an active-backup bond to | |
345 | the same MAC address at enslavement time. This is the | |
346 | default. | |
347 | ||
348 | active or 1 | |
349 | ||
350 | The "active" fail_over_mac policy indicates that the | |
351 | MAC address of the bond should always be the MAC | |
352 | address of the currently active slave. The MAC | |
353 | address of the slaves is not changed; instead, the MAC | |
354 | address of the bond changes during a failover. | |
355 | ||
356 | This policy is useful for devices that cannot ever | |
357 | alter their MAC address, or for devices that refuse | |
358 | incoming broadcasts with their own source MAC (which | |
359 | interferes with the ARP monitor). | |
360 | ||
361 | The down side of this policy is that every device on | |
362 | the network must be updated via gratuitous ARP, | |
363 | vs. just updating a switch or set of switches (which | |
364 | often takes place for any traffic, not just ARP | |
365 | traffic, if the switch snoops incoming traffic to | |
366 | update its tables) for the traditional method. If the | |
367 | gratuitous ARP is lost, communication may be | |
368 | disrupted. | |
369 | ||
370 | When this policy is used in conjuction with the mii | |
371 | monitor, devices which assert link up prior to being | |
372 | able to actually transmit and receive are particularly | |
19f59460 | 373 | susceptible to loss of the gratuitous ARP, and an |
3915c1e8 JV |
374 | appropriate updelay setting may be required. |
375 | ||
376 | follow or 2 | |
377 | ||
378 | The "follow" fail_over_mac policy causes the MAC | |
379 | address of the bond to be selected normally (normally | |
380 | the MAC address of the first slave added to the bond). | |
381 | However, the second and subsequent slaves are not set | |
382 | to this MAC address while they are in a backup role; a | |
383 | slave is programmed with the bond's MAC address at | |
384 | failover time (and the formerly active slave receives | |
385 | the newly active slave's MAC address). | |
386 | ||
387 | This policy is useful for multiport devices that | |
388 | either become confused or incur a performance penalty | |
389 | when multiple ports are programmed with the same MAC | |
390 | address. | |
391 | ||
392 | ||
393 | The default policy is none, unless the first slave cannot | |
394 | change its MAC address, in which case the active policy is | |
395 | selected by default. | |
396 | ||
397 | This option may be modified via sysfs only when no slaves are | |
398 | present in the bond. | |
399 | ||
400 | This option was added in bonding version 3.2.0. The "follow" | |
401 | policy was added in bonding version 3.3.0. | |
dd957c57 | 402 | |
1da177e4 LT |
403 | lacp_rate |
404 | ||
405 | Option specifying the rate in which we'll ask our link partner | |
406 | to transmit LACPDU packets in 802.3ad mode. Possible values | |
407 | are: | |
408 | ||
409 | slow or 0 | |
00354cfb | 410 | Request partner to transmit LACPDUs every 30 seconds |
1da177e4 LT |
411 | |
412 | fast or 1 | |
413 | Request partner to transmit LACPDUs every 1 second | |
414 | ||
00354cfb JV |
415 | The default is slow. |
416 | ||
1da177e4 LT |
417 | max_bonds |
418 | ||
419 | Specifies the number of bonding devices to create for this | |
420 | instance of the bonding driver. E.g., if max_bonds is 3, and | |
421 | the bonding driver is not already loaded, then bond0, bond1 | |
b8a9787e JV |
422 | and bond2 will be created. The default value is 1. Specifying |
423 | a value of 0 will load bonding, but will not create any devices. | |
1da177e4 LT |
424 | |
425 | miimon | |
426 | ||
00354cfb JV |
427 | Specifies the MII link monitoring frequency in milliseconds. |
428 | This determines how often the link state of each slave is | |
429 | inspected for link failures. A value of zero disables MII | |
430 | link monitoring. A value of 100 is a good starting point. | |
431 | The use_carrier option, below, affects how the link state is | |
1da177e4 LT |
432 | determined. See the High Availability section for additional |
433 | information. The default value is 0. | |
434 | ||
435 | mode | |
436 | ||
437 | Specifies one of the bonding policies. The default is | |
438 | balance-rr (round robin). Possible values are: | |
439 | ||
440 | balance-rr or 0 | |
441 | ||
442 | Round-robin policy: Transmit packets in sequential | |
443 | order from the first available slave through the | |
444 | last. This mode provides load balancing and fault | |
445 | tolerance. | |
446 | ||
447 | active-backup or 1 | |
448 | ||
449 | Active-backup policy: Only one slave in the bond is | |
450 | active. A different slave becomes active if, and only | |
451 | if, the active slave fails. The bond's MAC address is | |
452 | externally visible on only one port (network adapter) | |
00354cfb JV |
453 | to avoid confusing the switch. |
454 | ||
455 | In bonding version 2.6.2 or later, when a failover | |
456 | occurs in active-backup mode, bonding will issue one | |
457 | or more gratuitous ARPs on the newly active slave. | |
6224e01d | 458 | One gratuitous ARP is issued for the bonding master |
00354cfb JV |
459 | interface and each VLAN interfaces configured above |
460 | it, provided that the interface has at least one IP | |
461 | address configured. Gratuitous ARPs issued for VLAN | |
462 | interfaces are tagged with the appropriate VLAN id. | |
463 | ||
464 | This mode provides fault tolerance. The primary | |
465 | option, documented below, affects the behavior of this | |
466 | mode. | |
1da177e4 LT |
467 | |
468 | balance-xor or 2 | |
469 | ||
00354cfb JV |
470 | XOR policy: Transmit based on the selected transmit |
471 | hash policy. The default policy is a simple [(source | |
472 | MAC address XOR'd with destination MAC address) modulo | |
473 | slave count]. Alternate transmit policies may be | |
474 | selected via the xmit_hash_policy option, described | |
475 | below. | |
476 | ||
477 | This mode provides load balancing and fault tolerance. | |
1da177e4 LT |
478 | |
479 | broadcast or 3 | |
480 | ||
481 | Broadcast policy: transmits everything on all slave | |
482 | interfaces. This mode provides fault tolerance. | |
483 | ||
484 | 802.3ad or 4 | |
485 | ||
486 | IEEE 802.3ad Dynamic link aggregation. Creates | |
487 | aggregation groups that share the same speed and | |
488 | duplex settings. Utilizes all slaves in the active | |
489 | aggregator according to the 802.3ad specification. | |
490 | ||
00354cfb JV |
491 | Slave selection for outgoing traffic is done according |
492 | to the transmit hash policy, which may be changed from | |
493 | the default simple XOR policy via the xmit_hash_policy | |
494 | option, documented below. Note that not all transmit | |
495 | policies may be 802.3ad compliant, particularly in | |
496 | regards to the packet mis-ordering requirements of | |
497 | section 43.2.4 of the 802.3ad standard. Differing | |
498 | peer implementations will have varying tolerances for | |
499 | noncompliance. | |
500 | ||
501 | Prerequisites: | |
1da177e4 LT |
502 | |
503 | 1. Ethtool support in the base drivers for retrieving | |
504 | the speed and duplex of each slave. | |
505 | ||
506 | 2. A switch that supports IEEE 802.3ad Dynamic link | |
507 | aggregation. | |
508 | ||
509 | Most switches will require some type of configuration | |
510 | to enable 802.3ad mode. | |
511 | ||
512 | balance-tlb or 5 | |
513 | ||
514 | Adaptive transmit load balancing: channel bonding that | |
515 | does not require any special switch support. The | |
516 | outgoing traffic is distributed according to the | |
517 | current load (computed relative to the speed) on each | |
518 | slave. Incoming traffic is received by the current | |
519 | slave. If the receiving slave fails, another slave | |
520 | takes over the MAC address of the failed receiving | |
521 | slave. | |
522 | ||
523 | Prerequisite: | |
524 | ||
525 | Ethtool support in the base drivers for retrieving the | |
526 | speed of each slave. | |
527 | ||
528 | balance-alb or 6 | |
529 | ||
530 | Adaptive load balancing: includes balance-tlb plus | |
531 | receive load balancing (rlb) for IPV4 traffic, and | |
532 | does not require any special switch support. The | |
533 | receive load balancing is achieved by ARP negotiation. | |
534 | The bonding driver intercepts the ARP Replies sent by | |
535 | the local system on their way out and overwrites the | |
536 | source hardware address with the unique hardware | |
537 | address of one of the slaves in the bond such that | |
538 | different peers use different hardware addresses for | |
539 | the server. | |
540 | ||
541 | Receive traffic from connections created by the server | |
542 | is also balanced. When the local system sends an ARP | |
543 | Request the bonding driver copies and saves the peer's | |
544 | IP information from the ARP packet. When the ARP | |
545 | Reply arrives from the peer, its hardware address is | |
546 | retrieved and the bonding driver initiates an ARP | |
547 | reply to this peer assigning it to one of the slaves | |
548 | in the bond. A problematic outcome of using ARP | |
549 | negotiation for balancing is that each time that an | |
550 | ARP request is broadcast it uses the hardware address | |
551 | of the bond. Hence, peers learn the hardware address | |
552 | of the bond and the balancing of receive traffic | |
553 | collapses to the current slave. This is handled by | |
554 | sending updates (ARP Replies) to all the peers with | |
555 | their individually assigned hardware address such that | |
556 | the traffic is redistributed. Receive traffic is also | |
557 | redistributed when a new slave is added to the bond | |
558 | and when an inactive slave is re-activated. The | |
559 | receive load is distributed sequentially (round robin) | |
560 | among the group of highest speed slaves in the bond. | |
561 | ||
562 | When a link is reconnected or a new slave joins the | |
563 | bond the receive traffic is redistributed among all | |
00354cfb | 564 | active slaves in the bond by initiating ARP Replies |
6224e01d | 565 | with the selected MAC address to each of the |
1da177e4 LT |
566 | clients. The updelay parameter (detailed below) must |
567 | be set to a value equal or greater than the switch's | |
568 | forwarding delay so that the ARP Replies sent to the | |
569 | peers will not be blocked by the switch. | |
570 | ||
571 | Prerequisites: | |
572 | ||
573 | 1. Ethtool support in the base drivers for retrieving | |
574 | the speed of each slave. | |
575 | ||
576 | 2. Base driver support for setting the hardware | |
577 | address of a device while it is open. This is | |
578 | required so that there will always be one slave in the | |
579 | team using the bond hardware address (the | |
580 | curr_active_slave) while having a unique hardware | |
581 | address for each slave in the bond. If the | |
582 | curr_active_slave fails its hardware address is | |
583 | swapped with the new curr_active_slave that was | |
584 | chosen. | |
585 | ||
b59f9f74 JV |
586 | num_grat_arp |
587 | ||
588 | Specifies the number of gratuitous ARPs to be issued after a | |
589 | failover event. One gratuitous ARP is issued immediately after | |
590 | the failover, subsequent ARPs are sent at a rate of one per link | |
591 | monitor interval (arp_interval or miimon, whichever is active). | |
592 | ||
593 | The valid range is 0 - 255; the default value is 1. This option | |
594 | affects only the active-backup mode. This option was added for | |
595 | bonding version 3.3.0. | |
596 | ||
305d552a BH |
597 | num_unsol_na |
598 | ||
599 | Specifies the number of unsolicited IPv6 Neighbor Advertisements | |
600 | to be issued after a failover event. One unsolicited NA is issued | |
601 | immediately after the failover. | |
602 | ||
603 | The valid range is 0 - 255; the default value is 1. This option | |
604 | affects only the active-backup mode. This option was added for | |
605 | bonding version 3.4.0. | |
606 | ||
1da177e4 LT |
607 | primary |
608 | ||
609 | A string (eth0, eth2, etc) specifying which slave is the | |
610 | primary device. The specified device will always be the | |
611 | active slave while it is available. Only when the primary is | |
612 | off-line will alternate devices be used. This is useful when | |
613 | one slave is preferred over another, e.g., when one slave has | |
614 | higher throughput than another. | |
615 | ||
616 | The primary option is only valid for active-backup mode. | |
617 | ||
a549952a JP |
618 | primary_reselect |
619 | ||
620 | Specifies the reselection policy for the primary slave. This | |
621 | affects how the primary slave is chosen to become the active slave | |
622 | when failure of the active slave or recovery of the primary slave | |
623 | occurs. This option is designed to prevent flip-flopping between | |
624 | the primary slave and other slaves. Possible values are: | |
625 | ||
626 | always or 0 (default) | |
627 | ||
628 | The primary slave becomes the active slave whenever it | |
629 | comes back up. | |
630 | ||
631 | better or 1 | |
632 | ||
633 | The primary slave becomes the active slave when it comes | |
634 | back up, if the speed and duplex of the primary slave is | |
635 | better than the speed and duplex of the current active | |
636 | slave. | |
637 | ||
638 | failure or 2 | |
639 | ||
640 | The primary slave becomes the active slave only if the | |
641 | current active slave fails and the primary slave is up. | |
642 | ||
643 | The primary_reselect setting is ignored in two cases: | |
644 | ||
645 | If no slaves are active, the first slave to recover is | |
646 | made the active slave. | |
647 | ||
648 | When initially enslaved, the primary slave is always made | |
649 | the active slave. | |
650 | ||
651 | Changing the primary_reselect policy via sysfs will cause an | |
652 | immediate selection of the best active slave according to the new | |
653 | policy. This may or may not result in a change of the active | |
654 | slave, depending upon the circumstances. | |
655 | ||
656 | This option was added for bonding version 3.6.0. | |
657 | ||
1da177e4 LT |
658 | updelay |
659 | ||
660 | Specifies the time, in milliseconds, to wait before enabling a | |
661 | slave after a link recovery has been detected. This option is | |
662 | only valid for the miimon link monitor. The updelay value | |
663 | should be a multiple of the miimon value; if not, it will be | |
664 | rounded down to the nearest multiple. The default value is 0. | |
665 | ||
666 | use_carrier | |
667 | ||
668 | Specifies whether or not miimon should use MII or ETHTOOL | |
669 | ioctls vs. netif_carrier_ok() to determine the link | |
670 | status. The MII or ETHTOOL ioctls are less efficient and | |
671 | utilize a deprecated calling sequence within the kernel. The | |
672 | netif_carrier_ok() relies on the device driver to maintain its | |
673 | state with netif_carrier_on/off; at this writing, most, but | |
674 | not all, device drivers support this facility. | |
675 | ||
676 | If bonding insists that the link is up when it should not be, | |
677 | it may be that your network device driver does not support | |
678 | netif_carrier_on/off. The default state for netif_carrier is | |
679 | "carrier on," so if a driver does not support netif_carrier, | |
680 | it will appear as if the link is always up. In this case, | |
681 | setting use_carrier to 0 will cause bonding to revert to the | |
682 | MII / ETHTOOL ioctl method to determine the link state. | |
683 | ||
684 | A value of 1 enables the use of netif_carrier_ok(), a value of | |
685 | 0 will use the deprecated MII / ETHTOOL ioctls. The default | |
686 | value is 1. | |
687 | ||
00354cfb JV |
688 | xmit_hash_policy |
689 | ||
690 | Selects the transmit hash policy to use for slave selection in | |
691 | balance-xor and 802.3ad modes. Possible values are: | |
692 | ||
693 | layer2 | |
694 | ||
695 | Uses XOR of hardware MAC addresses to generate the | |
696 | hash. The formula is | |
697 | ||
698 | (source MAC XOR destination MAC) modulo slave count | |
699 | ||
700 | This algorithm will place all traffic to a particular | |
701 | network peer on the same slave. | |
702 | ||
703 | This algorithm is 802.3ad compliant. | |
704 | ||
6f6652be JV |
705 | layer2+3 |
706 | ||
707 | This policy uses a combination of layer2 and layer3 | |
708 | protocol information to generate the hash. | |
709 | ||
710 | Uses XOR of hardware MAC addresses and IP addresses to | |
711 | generate the hash. The formula is | |
712 | ||
713 | (((source IP XOR dest IP) AND 0xffff) XOR | |
714 | ( source MAC XOR destination MAC )) | |
715 | modulo slave count | |
716 | ||
717 | This algorithm will place all traffic to a particular | |
718 | network peer on the same slave. For non-IP traffic, | |
719 | the formula is the same as for the layer2 transmit | |
720 | hash policy. | |
721 | ||
722 | This policy is intended to provide a more balanced | |
723 | distribution of traffic than layer2 alone, especially | |
724 | in environments where a layer3 gateway device is | |
725 | required to reach most destinations. | |
726 | ||
d9195881 | 727 | This algorithm is 802.3ad compliant. |
6f6652be | 728 | |
00354cfb JV |
729 | layer3+4 |
730 | ||
731 | This policy uses upper layer protocol information, | |
732 | when available, to generate the hash. This allows for | |
733 | traffic to a particular network peer to span multiple | |
734 | slaves, although a single connection will not span | |
735 | multiple slaves. | |
736 | ||
737 | The formula for unfragmented TCP and UDP packets is | |
738 | ||
739 | ((source port XOR dest port) XOR | |
740 | ((source IP XOR dest IP) AND 0xffff) | |
741 | modulo slave count | |
742 | ||
743 | For fragmented TCP or UDP packets and all other IP | |
744 | protocol traffic, the source and destination port | |
745 | information is omitted. For non-IP traffic, the | |
746 | formula is the same as for the layer2 transmit hash | |
747 | policy. | |
748 | ||
749 | This policy is intended to mimic the behavior of | |
750 | certain switches, notably Cisco switches with PFC2 as | |
751 | well as some Foundry and IBM products. | |
752 | ||
753 | This algorithm is not fully 802.3ad compliant. A | |
754 | single TCP or UDP conversation containing both | |
755 | fragmented and unfragmented packets will see packets | |
756 | striped across two interfaces. This may result in out | |
757 | of order delivery. Most traffic types will not meet | |
758 | this criteria, as TCP rarely fragments traffic, and | |
759 | most UDP traffic is not involved in extended | |
760 | conversations. Other implementations of 802.3ad may | |
761 | or may not tolerate this noncompliance. | |
762 | ||
763 | The default value is layer2. This option was added in bonding | |
6f6652be JV |
764 | version 2.6.3. In earlier versions of bonding, this parameter |
765 | does not exist, and the layer2 policy is the only policy. The | |
766 | layer2+3 value was added for bonding version 3.2.2. | |
1da177e4 LT |
767 | |
768 | ||
769 | 3. Configuring Bonding Devices | |
770 | ============================== | |
771 | ||
6224e01d AK |
772 | You can configure bonding using either your distro's network |
773 | initialization scripts, or manually using either ifenslave or the | |
774 | sysfs interface. Distros generally use one of two packages for the | |
775 | network initialization scripts: initscripts or sysconfig. Recent | |
776 | versions of these packages have support for bonding, while older | |
777 | versions do not. | |
1da177e4 LT |
778 | |
779 | We will first describe the options for configuring bonding for | |
780 | distros using versions of initscripts and sysconfig with full or | |
781 | partial support for bonding, then provide information on enabling | |
782 | bonding without support from the network initialization scripts (i.e., | |
783 | older versions of initscripts or sysconfig). | |
784 | ||
785 | If you're unsure whether your distro uses sysconfig or | |
786 | initscripts, or don't know if it's new enough, have no fear. | |
787 | Determining this is fairly straightforward. | |
788 | ||
789 | First, issue the command: | |
790 | ||
791 | $ rpm -qf /sbin/ifup | |
792 | ||
793 | It will respond with a line of text starting with either | |
794 | "initscripts" or "sysconfig," followed by some numbers. This is the | |
795 | package that provides your network initialization scripts. | |
796 | ||
797 | Next, to determine if your installation supports bonding, | |
798 | issue the command: | |
799 | ||
800 | $ grep ifenslave /sbin/ifup | |
801 | ||
802 | If this returns any matches, then your initscripts or | |
803 | sysconfig has support for bonding. | |
804 | ||
6224e01d | 805 | 3.1 Configuration with Sysconfig Support |
1da177e4 LT |
806 | ---------------------------------------- |
807 | ||
808 | This section applies to distros using a version of sysconfig | |
809 | with bonding support, for example, SuSE Linux Enterprise Server 9. | |
810 | ||
811 | SuSE SLES 9's networking configuration system does support | |
812 | bonding, however, at this writing, the YaST system configuration | |
6224e01d | 813 | front end does not provide any means to work with bonding devices. |
1da177e4 LT |
814 | Bonding devices can be managed by hand, however, as follows. |
815 | ||
816 | First, if they have not already been configured, configure the | |
817 | slave devices. On SLES 9, this is most easily done by running the | |
818 | yast2 sysconfig configuration utility. The goal is for to create an | |
819 | ifcfg-id file for each slave device. The simplest way to accomplish | |
00354cfb JV |
820 | this is to configure the devices for DHCP (this is only to get the |
821 | file ifcfg-id file created; see below for some issues with DHCP). The | |
822 | name of the configuration file for each device will be of the form: | |
1da177e4 LT |
823 | |
824 | ifcfg-id-xx:xx:xx:xx:xx:xx | |
825 | ||
826 | Where the "xx" portion will be replaced with the digits from | |
827 | the device's permanent MAC address. | |
828 | ||
829 | Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been | |
830 | created, it is necessary to edit the configuration files for the slave | |
831 | devices (the MAC addresses correspond to those of the slave devices). | |
00354cfb | 832 | Before editing, the file will contain multiple lines, and will look |
1da177e4 LT |
833 | something like this: |
834 | ||
835 | BOOTPROTO='dhcp' | |
836 | STARTMODE='on' | |
837 | USERCTL='no' | |
838 | UNIQUE='XNzu.WeZGOGF+4wE' | |
839 | _nm_name='bus-pci-0001:61:01.0' | |
840 | ||
841 | Change the BOOTPROTO and STARTMODE lines to the following: | |
842 | ||
843 | BOOTPROTO='none' | |
844 | STARTMODE='off' | |
845 | ||
846 | Do not alter the UNIQUE or _nm_name lines. Remove any other | |
847 | lines (USERCTL, etc). | |
848 | ||
849 | Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified, | |
850 | it's time to create the configuration file for the bonding device | |
851 | itself. This file is named ifcfg-bondX, where X is the number of the | |
852 | bonding device to create, starting at 0. The first such file is | |
853 | ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig | |
854 | network configuration system will correctly start multiple instances | |
855 | of bonding. | |
856 | ||
857 | The contents of the ifcfg-bondX file is as follows: | |
858 | ||
859 | BOOTPROTO="static" | |
860 | BROADCAST="10.0.2.255" | |
861 | IPADDR="10.0.2.10" | |
862 | NETMASK="255.255.0.0" | |
863 | NETWORK="10.0.2.0" | |
864 | REMOTE_IPADDR="" | |
865 | STARTMODE="onboot" | |
866 | BONDING_MASTER="yes" | |
867 | BONDING_MODULE_OPTS="mode=active-backup miimon=100" | |
868 | BONDING_SLAVE0="eth0" | |
00354cfb | 869 | BONDING_SLAVE1="bus-pci-0000:06:08.1" |
1da177e4 LT |
870 | |
871 | Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK | |
872 | values with the appropriate values for your network. | |
873 | ||
1da177e4 LT |
874 | The STARTMODE specifies when the device is brought online. |
875 | The possible values are: | |
876 | ||
877 | onboot: The device is started at boot time. If you're not | |
878 | sure, this is probably what you want. | |
879 | ||
880 | manual: The device is started only when ifup is called | |
881 | manually. Bonding devices may be configured this | |
882 | way if you do not wish them to start automatically | |
883 | at boot for some reason. | |
884 | ||
885 | hotplug: The device is started by a hotplug event. This is not | |
886 | a valid choice for a bonding device. | |
887 | ||
888 | off or ignore: The device configuration is ignored. | |
889 | ||
890 | The line BONDING_MASTER='yes' indicates that the device is a | |
891 | bonding master device. The only useful value is "yes." | |
892 | ||
893 | The contents of BONDING_MODULE_OPTS are supplied to the | |
894 | instance of the bonding module for this device. Specify the options | |
895 | for the bonding mode, link monitoring, and so on here. Do not include | |
896 | the max_bonds bonding parameter; this will confuse the configuration | |
897 | system if you have multiple bonding devices. | |
898 | ||
00354cfb JV |
899 | Finally, supply one BONDING_SLAVEn="slave device" for each |
900 | slave. where "n" is an increasing value, one for each slave. The | |
901 | "slave device" is either an interface name, e.g., "eth0", or a device | |
902 | specifier for the network device. The interface name is easier to | |
903 | find, but the ethN names are subject to change at boot time if, e.g., | |
904 | a device early in the sequence has failed. The device specifiers | |
905 | (bus-pci-0000:06:08.1 in the example above) specify the physical | |
906 | network device, and will not change unless the device's bus location | |
907 | changes (for example, it is moved from one PCI slot to another). The | |
908 | example above uses one of each type for demonstration purposes; most | |
909 | configurations will choose one or the other for all slave devices. | |
1da177e4 LT |
910 | |
911 | When all configuration files have been modified or created, | |
912 | networking must be restarted for the configuration changes to take | |
913 | effect. This can be accomplished via the following: | |
914 | ||
915 | # /etc/init.d/network restart | |
916 | ||
917 | Note that the network control script (/sbin/ifdown) will | |
918 | remove the bonding module as part of the network shutdown processing, | |
919 | so it is not necessary to remove the module by hand if, e.g., the | |
00354cfb | 920 | module parameters have changed. |
1da177e4 LT |
921 | |
922 | Also, at this writing, YaST/YaST2 will not manage bonding | |
923 | devices (they do not show bonding interfaces on its list of network | |
924 | devices). It is necessary to edit the configuration file by hand to | |
925 | change the bonding configuration. | |
926 | ||
927 | Additional general options and details of the ifcfg file | |
928 | format can be found in an example ifcfg template file: | |
929 | ||
930 | /etc/sysconfig/network/ifcfg.template | |
931 | ||
932 | Note that the template does not document the various BONDING_ | |
933 | settings described above, but does describe many of the other options. | |
934 | ||
6224e01d | 935 | 3.1.1 Using DHCP with Sysconfig |
00354cfb JV |
936 | ------------------------------- |
937 | ||
938 | Under sysconfig, configuring a device with BOOTPROTO='dhcp' | |
939 | will cause it to query DHCP for its IP address information. At this | |
940 | writing, this does not function for bonding devices; the scripts | |
941 | attempt to obtain the device address from DHCP prior to adding any of | |
942 | the slave devices. Without active slaves, the DHCP requests are not | |
943 | sent to the network. | |
944 | ||
6224e01d | 945 | 3.1.2 Configuring Multiple Bonds with Sysconfig |
00354cfb JV |
946 | ----------------------------------------------- |
947 | ||
948 | The sysconfig network initialization system is capable of | |
949 | handling multiple bonding devices. All that is necessary is for each | |
950 | bonding instance to have an appropriately configured ifcfg-bondX file | |
951 | (as described above). Do not specify the "max_bonds" parameter to any | |
952 | instance of bonding, as this will confuse sysconfig. If you require | |
953 | multiple bonding devices with identical parameters, create multiple | |
954 | ifcfg-bondX files. | |
955 | ||
956 | Because the sysconfig scripts supply the bonding module | |
957 | options in the ifcfg-bondX file, it is not necessary to add them to | |
958 | the system /etc/modules.conf or /etc/modprobe.conf configuration file. | |
959 | ||
6224e01d | 960 | 3.2 Configuration with Initscripts Support |
1da177e4 LT |
961 | ------------------------------------------ |
962 | ||
9a6c6867 JV |
963 | This section applies to distros using a recent version of |
964 | initscripts with bonding support, for example, Red Hat Enterprise Linux | |
965 | version 3 or later, Fedora, etc. On these systems, the network | |
966 | initialization scripts have knowledge of bonding, and can be configured to | |
967 | control bonding devices. Note that older versions of the initscripts | |
968 | package have lower levels of support for bonding; this will be noted where | |
969 | applicable. | |
1da177e4 LT |
970 | |
971 | These distros will not automatically load the network adapter | |
972 | driver unless the ethX device is configured with an IP address. | |
973 | Because of this constraint, users must manually configure a | |
974 | network-script file for all physical adapters that will be members of | |
975 | a bondX link. Network script files are located in the directory: | |
976 | ||
977 | /etc/sysconfig/network-scripts | |
978 | ||
979 | The file name must be prefixed with "ifcfg-eth" and suffixed | |
980 | with the adapter's physical adapter number. For example, the script | |
981 | for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0. | |
982 | Place the following text in the file: | |
983 | ||
984 | DEVICE=eth0 | |
985 | USERCTL=no | |
986 | ONBOOT=yes | |
987 | MASTER=bond0 | |
988 | SLAVE=yes | |
989 | BOOTPROTO=none | |
990 | ||
991 | The DEVICE= line will be different for every ethX device and | |
992 | must correspond with the name of the file, i.e., ifcfg-eth1 must have | |
993 | a device line of DEVICE=eth1. The setting of the MASTER= line will | |
994 | also depend on the final bonding interface name chosen for your bond. | |
995 | As with other network devices, these typically start at 0, and go up | |
996 | one for each device, i.e., the first bonding instance is bond0, the | |
997 | second is bond1, and so on. | |
998 | ||
999 | Next, create a bond network script. The file name for this | |
1000 | script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is | |
1001 | the number of the bond. For bond0 the file is named "ifcfg-bond0", | |
1002 | for bond1 it is named "ifcfg-bond1", and so on. Within that file, | |
1003 | place the following text: | |
1004 | ||
1005 | DEVICE=bond0 | |
1006 | IPADDR=192.168.1.1 | |
1007 | NETMASK=255.255.255.0 | |
1008 | NETWORK=192.168.1.0 | |
1009 | BROADCAST=192.168.1.255 | |
1010 | ONBOOT=yes | |
1011 | BOOTPROTO=none | |
1012 | USERCTL=no | |
1013 | ||
1014 | Be sure to change the networking specific lines (IPADDR, | |
1015 | NETMASK, NETWORK and BROADCAST) to match your network configuration. | |
1016 | ||
9a6c6867 | 1017 | For later versions of initscripts, such as that found with Fedora |
3f8b4b13 AG |
1018 | 7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible, |
1019 | and, indeed, preferable, to specify the bonding options in the ifcfg-bond0 | |
9a6c6867 JV |
1020 | file, e.g. a line of the format: |
1021 | ||
3f8b4b13 | 1022 | BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254" |
9a6c6867 JV |
1023 | |
1024 | will configure the bond with the specified options. The options | |
1025 | specified in BONDING_OPTS are identical to the bonding module parameters | |
3f8b4b13 AG |
1026 | except for the arp_ip_target field when using versions of initscripts older |
1027 | than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When | |
1028 | using older versions each target should be included as a separate option and | |
1029 | should be preceded by a '+' to indicate it should be added to the list of | |
1030 | queried targets, e.g., | |
9a6c6867 JV |
1031 | |
1032 | arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2 | |
1033 | ||
1034 | is the proper syntax to specify multiple targets. When specifying | |
1035 | options via BONDING_OPTS, it is not necessary to edit /etc/modules.conf or | |
1036 | /etc/modprobe.conf. | |
1037 | ||
3f8b4b13 | 1038 | For even older versions of initscripts that do not support |
9a6c6867 JV |
1039 | BONDING_OPTS, it is necessary to edit /etc/modules.conf (or |
1040 | /etc/modprobe.conf, depending upon your distro) to load the bonding module | |
1041 | with your desired options when the bond0 interface is brought up. The | |
1042 | following lines in /etc/modules.conf (or modprobe.conf) will load the | |
1043 | bonding module, and select its options: | |
1da177e4 LT |
1044 | |
1045 | alias bond0 bonding | |
1046 | options bond0 mode=balance-alb miimon=100 | |
1047 | ||
1048 | Replace the sample parameters with the appropriate set of | |
1049 | options for your configuration. | |
1050 | ||
1051 | Finally run "/etc/rc.d/init.d/network restart" as root. This | |
1052 | will restart the networking subsystem and your bond link should be now | |
1053 | up and running. | |
1054 | ||
6224e01d | 1055 | 3.2.1 Using DHCP with Initscripts |
00354cfb JV |
1056 | --------------------------------- |
1057 | ||
9a6c6867 JV |
1058 | Recent versions of initscripts (the versions supplied with Fedora |
1059 | Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to | |
1060 | work) have support for assigning IP information to bonding devices via | |
1061 | DHCP. | |
00354cfb JV |
1062 | |
1063 | To configure bonding for DHCP, configure it as described | |
1064 | above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp" | |
1065 | and add a line consisting of "TYPE=Bonding". Note that the TYPE value | |
1066 | is case sensitive. | |
1067 | ||
6224e01d | 1068 | 3.2.2 Configuring Multiple Bonds with Initscripts |
00354cfb JV |
1069 | ------------------------------------------------- |
1070 | ||
9a6c6867 JV |
1071 | Initscripts packages that are included with Fedora 7 and Red Hat |
1072 | Enterprise Linux 5 support multiple bonding interfaces by simply | |
1073 | specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the | |
1074 | number of the bond. This support requires sysfs support in the kernel, | |
1075 | and a bonding driver of version 3.0.0 or later. Other configurations may | |
1076 | not support this method for specifying multiple bonding interfaces; for | |
1077 | those instances, see the "Configuring Multiple Bonds Manually" section, | |
1078 | below. | |
1da177e4 | 1079 | |
6224e01d AK |
1080 | 3.3 Configuring Bonding Manually with Ifenslave |
1081 | ----------------------------------------------- | |
1da177e4 LT |
1082 | |
1083 | This section applies to distros whose network initialization | |
1084 | scripts (the sysconfig or initscripts package) do not have specific | |
1085 | knowledge of bonding. One such distro is SuSE Linux Enterprise Server | |
1086 | version 8. | |
1087 | ||
00354cfb JV |
1088 | The general method for these systems is to place the bonding |
1089 | module parameters into /etc/modules.conf or /etc/modprobe.conf (as | |
1090 | appropriate for the installed distro), then add modprobe and/or | |
1091 | ifenslave commands to the system's global init script. The name of | |
1092 | the global init script differs; for sysconfig, it is | |
1da177e4 LT |
1093 | /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local. |
1094 | ||
1095 | For example, if you wanted to make a simple bond of two e100 | |
1096 | devices (presumed to be eth0 and eth1), and have it persist across | |
1097 | reboots, edit the appropriate file (/etc/init.d/boot.local or | |
1098 | /etc/rc.d/rc.local), and add the following: | |
1099 | ||
00354cfb | 1100 | modprobe bonding mode=balance-alb miimon=100 |
1da177e4 LT |
1101 | modprobe e100 |
1102 | ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up | |
1103 | ifenslave bond0 eth0 | |
1104 | ifenslave bond0 eth1 | |
1105 | ||
1106 | Replace the example bonding module parameters and bond0 | |
1107 | network configuration (IP address, netmask, etc) with the appropriate | |
00354cfb | 1108 | values for your configuration. |
1da177e4 LT |
1109 | |
1110 | Unfortunately, this method will not provide support for the | |
1111 | ifup and ifdown scripts on the bond devices. To reload the bonding | |
1112 | configuration, it is necessary to run the initialization script, e.g., | |
1113 | ||
1114 | # /etc/init.d/boot.local | |
1115 | ||
1116 | or | |
1117 | ||
1118 | # /etc/rc.d/rc.local | |
1119 | ||
1120 | It may be desirable in such a case to create a separate script | |
1121 | which only initializes the bonding configuration, then call that | |
1122 | separate script from within boot.local. This allows for bonding to be | |
1123 | enabled without re-running the entire global init script. | |
1124 | ||
1125 | To shut down the bonding devices, it is necessary to first | |
1126 | mark the bonding device itself as being down, then remove the | |
1127 | appropriate device driver modules. For our example above, you can do | |
1128 | the following: | |
1129 | ||
1130 | # ifconfig bond0 down | |
00354cfb | 1131 | # rmmod bonding |
1da177e4 LT |
1132 | # rmmod e100 |
1133 | ||
1134 | Again, for convenience, it may be desirable to create a script | |
1135 | with these commands. | |
1136 | ||
1137 | ||
00354cfb JV |
1138 | 3.3.1 Configuring Multiple Bonds Manually |
1139 | ----------------------------------------- | |
1da177e4 LT |
1140 | |
1141 | This section contains information on configuring multiple | |
00354cfb JV |
1142 | bonding devices with differing options for those systems whose network |
1143 | initialization scripts lack support for configuring multiple bonds. | |
1144 | ||
1145 | If you require multiple bonding devices, but all with the same | |
1146 | options, you may wish to use the "max_bonds" module parameter, | |
1147 | documented above. | |
1da177e4 | 1148 | |
9a6c6867 JV |
1149 | To create multiple bonding devices with differing options, it is |
1150 | preferrable to use bonding parameters exported by sysfs, documented in the | |
1151 | section below. | |
1152 | ||
1153 | For versions of bonding without sysfs support, the only means to | |
1154 | provide multiple instances of bonding with differing options is to load | |
1155 | the bonding driver multiple times. Note that current versions of the | |
1156 | sysconfig network initialization scripts handle this automatically; if | |
1157 | your distro uses these scripts, no special action is needed. See the | |
1158 | section Configuring Bonding Devices, above, if you're not sure about your | |
1159 | network initialization scripts. | |
1160 | ||
1161 | To load multiple instances of the module, it is necessary to | |
1162 | specify a different name for each instance (the module loading system | |
1163 | requires that every loaded module, even multiple instances of the same | |
1164 | module, have a unique name). This is accomplished by supplying multiple | |
1165 | sets of bonding options in /etc/modprobe.conf, for example: | |
1166 | ||
1167 | alias bond0 bonding | |
1168 | options bond0 -o bond0 mode=balance-rr miimon=100 | |
1169 | ||
1170 | alias bond1 bonding | |
1171 | options bond1 -o bond1 mode=balance-alb miimon=50 | |
1172 | ||
1173 | will load the bonding module two times. The first instance is | |
1174 | named "bond0" and creates the bond0 device in balance-rr mode with an | |
1175 | miimon of 100. The second instance is named "bond1" and creates the | |
1176 | bond1 device in balance-alb mode with an miimon of 50. | |
1177 | ||
1178 | In some circumstances (typically with older distributions), | |
1179 | the above does not work, and the second bonding instance never sees | |
1180 | its options. In that case, the second options line can be substituted | |
1181 | as follows: | |
1182 | ||
1183 | install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \ | |
1184 | mode=balance-alb miimon=50 | |
00354cfb | 1185 | |
9a6c6867 JV |
1186 | This may be repeated any number of times, specifying a new and |
1187 | unique name in place of bond1 for each subsequent instance. | |
1188 | ||
1189 | It has been observed that some Red Hat supplied kernels are unable | |
1190 | to rename modules at load time (the "-o bond1" part). Attempts to pass | |
1191 | that option to modprobe will produce an "Operation not permitted" error. | |
1192 | This has been reported on some Fedora Core kernels, and has been seen on | |
1193 | RHEL 4 as well. On kernels exhibiting this problem, it will be impossible | |
1194 | to configure multiple bonds with differing parameters (as they are older | |
1195 | kernels, and also lack sysfs support). | |
1da177e4 | 1196 | |
6224e01d AK |
1197 | 3.4 Configuring Bonding Manually via Sysfs |
1198 | ------------------------------------------ | |
1199 | ||
9a6c6867 | 1200 | Starting with version 3.0.0, Channel Bonding may be configured |
6224e01d AK |
1201 | via the sysfs interface. This interface allows dynamic configuration |
1202 | of all bonds in the system without unloading the module. It also | |
1203 | allows for adding and removing bonds at runtime. Ifenslave is no | |
1204 | longer required, though it is still supported. | |
1205 | ||
1206 | Use of the sysfs interface allows you to use multiple bonds | |
1207 | with different configurations without having to reload the module. | |
1208 | It also allows you to use multiple, differently configured bonds when | |
1209 | bonding is compiled into the kernel. | |
1210 | ||
1211 | You must have the sysfs filesystem mounted to configure | |
1212 | bonding this way. The examples in this document assume that you | |
1213 | are using the standard mount point for sysfs, e.g. /sys. If your | |
1214 | sysfs filesystem is mounted elsewhere, you will need to adjust the | |
1215 | example paths accordingly. | |
1216 | ||
1217 | Creating and Destroying Bonds | |
1218 | ----------------------------- | |
1219 | To add a new bond foo: | |
1220 | # echo +foo > /sys/class/net/bonding_masters | |
1221 | ||
1222 | To remove an existing bond bar: | |
1223 | # echo -bar > /sys/class/net/bonding_masters | |
1224 | ||
1225 | To show all existing bonds: | |
1226 | # cat /sys/class/net/bonding_masters | |
1227 | ||
1228 | NOTE: due to 4K size limitation of sysfs files, this list may be | |
1229 | truncated if you have more than a few hundred bonds. This is unlikely | |
1230 | to occur under normal operating conditions. | |
1231 | ||
1232 | Adding and Removing Slaves | |
1233 | -------------------------- | |
1234 | Interfaces may be enslaved to a bond using the file | |
1235 | /sys/class/net/<bond>/bonding/slaves. The semantics for this file | |
1236 | are the same as for the bonding_masters file. | |
1237 | ||
1238 | To enslave interface eth0 to bond bond0: | |
1239 | # ifconfig bond0 up | |
1240 | # echo +eth0 > /sys/class/net/bond0/bonding/slaves | |
1241 | ||
1242 | To free slave eth0 from bond bond0: | |
1243 | # echo -eth0 > /sys/class/net/bond0/bonding/slaves | |
1244 | ||
6224e01d AK |
1245 | When an interface is enslaved to a bond, symlinks between the |
1246 | two are created in the sysfs filesystem. In this case, you would get | |
1247 | /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and | |
1248 | /sys/class/net/eth0/master pointing to /sys/class/net/bond0. | |
1249 | ||
1250 | This means that you can tell quickly whether or not an | |
1251 | interface is enslaved by looking for the master symlink. Thus: | |
1252 | # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves | |
1253 | will free eth0 from whatever bond it is enslaved to, regardless of | |
1254 | the name of the bond interface. | |
1255 | ||
1256 | Changing a Bond's Configuration | |
1257 | ------------------------------- | |
1258 | Each bond may be configured individually by manipulating the | |
1259 | files located in /sys/class/net/<bond name>/bonding | |
1260 | ||
1261 | The names of these files correspond directly with the command- | |
670e9f34 | 1262 | line parameters described elsewhere in this file, and, with the |
6224e01d AK |
1263 | exception of arp_ip_target, they accept the same values. To see the |
1264 | current setting, simply cat the appropriate file. | |
1265 | ||
1266 | A few examples will be given here; for specific usage | |
1267 | guidelines for each parameter, see the appropriate section in this | |
1268 | document. | |
1269 | ||
1270 | To configure bond0 for balance-alb mode: | |
1271 | # ifconfig bond0 down | |
1272 | # echo 6 > /sys/class/net/bond0/bonding/mode | |
1273 | - or - | |
1274 | # echo balance-alb > /sys/class/net/bond0/bonding/mode | |
1275 | NOTE: The bond interface must be down before the mode can be | |
1276 | changed. | |
1277 | ||
1278 | To enable MII monitoring on bond0 with a 1 second interval: | |
1279 | # echo 1000 > /sys/class/net/bond0/bonding/miimon | |
1280 | NOTE: If ARP monitoring is enabled, it will disabled when MII | |
1281 | monitoring is enabled, and vice-versa. | |
1282 | ||
1283 | To add ARP targets: | |
1284 | # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target | |
1285 | # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target | |
5a31bec0 | 1286 | NOTE: up to 16 target addresses may be specified. |
6224e01d AK |
1287 | |
1288 | To remove an ARP target: | |
1289 | # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target | |
1290 | ||
1291 | Example Configuration | |
1292 | --------------------- | |
1293 | We begin with the same example that is shown in section 3.3, | |
1294 | executed with sysfs, and without using ifenslave. | |
1295 | ||
1296 | To make a simple bond of two e100 devices (presumed to be eth0 | |
1297 | and eth1), and have it persist across reboots, edit the appropriate | |
1298 | file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the | |
1299 | following: | |
1300 | ||
1301 | modprobe bonding | |
1302 | modprobe e100 | |
1303 | echo balance-alb > /sys/class/net/bond0/bonding/mode | |
1304 | ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up | |
1305 | echo 100 > /sys/class/net/bond0/bonding/miimon | |
1306 | echo +eth0 > /sys/class/net/bond0/bonding/slaves | |
1307 | echo +eth1 > /sys/class/net/bond0/bonding/slaves | |
1308 | ||
1309 | To add a second bond, with two e1000 interfaces in | |
1310 | active-backup mode, using ARP monitoring, add the following lines to | |
1311 | your init script: | |
1312 | ||
1313 | modprobe e1000 | |
1314 | echo +bond1 > /sys/class/net/bonding_masters | |
1315 | echo active-backup > /sys/class/net/bond1/bonding/mode | |
1316 | ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up | |
1317 | echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target | |
1318 | echo 2000 > /sys/class/net/bond1/bonding/arp_interval | |
1319 | echo +eth2 > /sys/class/net/bond1/bonding/slaves | |
1320 | echo +eth3 > /sys/class/net/bond1/bonding/slaves | |
1321 | ||
bb1d9123 AG |
1322 | 3.5 Overriding Configuration for Special Cases |
1323 | ---------------------------------------------- | |
1324 | When using the bonding driver, the physical port which transmits a frame is | |
1325 | typically selected by the bonding driver, and is not relevant to the user or | |
1326 | system administrator. The output port is simply selected using the policies of | |
1327 | the selected bonding mode. On occasion however, it is helpful to direct certain | |
1328 | classes of traffic to certain physical interfaces on output to implement | |
1329 | slightly more complex policies. For example, to reach a web server over a | |
1330 | bonded interface in which eth0 connects to a private network, while eth1 | |
1331 | connects via a public network, it may be desirous to bias the bond to send said | |
1332 | traffic over eth0 first, using eth1 only as a fall back, while all other traffic | |
1333 | can safely be sent over either interface. Such configurations may be achieved | |
1334 | using the traffic control utilities inherent in linux. | |
1335 | ||
1336 | By default the bonding driver is multiqueue aware and 16 queues are created | |
1337 | when the driver initializes (see Documentation/networking/multiqueue.txt | |
1338 | for details). If more or less queues are desired the module parameter | |
1339 | tx_queues can be used to change this value. There is no sysfs parameter | |
1340 | available as the allocation is done at module init time. | |
1341 | ||
1342 | The output of the file /proc/net/bonding/bondX has changed so the output Queue | |
1343 | ID is now printed for each slave: | |
1344 | ||
1345 | Bonding Mode: fault-tolerance (active-backup) | |
1346 | Primary Slave: None | |
1347 | Currently Active Slave: eth0 | |
1348 | MII Status: up | |
1349 | MII Polling Interval (ms): 0 | |
1350 | Up Delay (ms): 0 | |
1351 | Down Delay (ms): 0 | |
1352 | ||
1353 | Slave Interface: eth0 | |
1354 | MII Status: up | |
1355 | Link Failure Count: 0 | |
1356 | Permanent HW addr: 00:1a:a0:12:8f:cb | |
1357 | Slave queue ID: 0 | |
1358 | ||
1359 | Slave Interface: eth1 | |
1360 | MII Status: up | |
1361 | Link Failure Count: 0 | |
1362 | Permanent HW addr: 00:1a:a0:12:8f:cc | |
1363 | Slave queue ID: 2 | |
1364 | ||
1365 | The queue_id for a slave can be set using the command: | |
1366 | ||
1367 | # echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id | |
1368 | ||
1369 | Any interface that needs a queue_id set should set it with multiple calls | |
1370 | like the one above until proper priorities are set for all interfaces. On | |
1371 | distributions that allow configuration via initscripts, multiple 'queue_id' | |
1372 | arguments can be added to BONDING_OPTS to set all needed slave queues. | |
1373 | ||
1374 | These queue id's can be used in conjunction with the tc utility to configure | |
1375 | a multiqueue qdisc and filters to bias certain traffic to transmit on certain | |
1376 | slave devices. For instance, say we wanted, in the above configuration to | |
1377 | force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output | |
1378 | device. The following commands would accomplish this: | |
1379 | ||
1380 | # tc qdisc add dev bond0 handle 1 root multiq | |
1381 | ||
1382 | # tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip dst \ | |
1383 | 192.168.1.100 action skbedit queue_mapping 2 | |
1384 | ||
1385 | These commands tell the kernel to attach a multiqueue queue discipline to the | |
1386 | bond0 interface and filter traffic enqueued to it, such that packets with a dst | |
1387 | ip of 192.168.1.100 have their output queue mapping value overwritten to 2. | |
1388 | This value is then passed into the driver, causing the normal output path | |
1389 | selection policy to be overridden, selecting instead qid 2, which maps to eth1. | |
1390 | ||
1391 | Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver | |
1392 | that normal output policy selection should take place. One benefit to simply | |
1393 | leaving the qid for a slave to 0 is the multiqueue awareness in the bonding | |
1394 | driver that is now present. This awareness allows tc filters to be placed on | |
1395 | slave devices as well as bond devices and the bonding driver will simply act as | |
1396 | a pass-through for selecting output queues on the slave device rather than | |
1397 | output port selection. | |
1398 | ||
1399 | This feature first appeared in bonding driver version 3.7.0 and support for | |
1400 | output slave selection was limited to round-robin and active-backup modes. | |
1401 | ||
1402 | 4 Querying Bonding Configuration | |
1da177e4 LT |
1403 | ================================= |
1404 | ||
6224e01d | 1405 | 4.1 Bonding Configuration |
1da177e4 LT |
1406 | ------------------------- |
1407 | ||
1408 | Each bonding device has a read-only file residing in the | |
1409 | /proc/net/bonding directory. The file contents include information | |
1410 | about the bonding configuration, options and state of each slave. | |
1411 | ||
1412 | For example, the contents of /proc/net/bonding/bond0 after the | |
1413 | driver is loaded with parameters of mode=0 and miimon=1000 is | |
1414 | generally as follows: | |
1415 | ||
1416 | Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004) | |
1417 | Bonding Mode: load balancing (round-robin) | |
1418 | Currently Active Slave: eth0 | |
1419 | MII Status: up | |
1420 | MII Polling Interval (ms): 1000 | |
1421 | Up Delay (ms): 0 | |
1422 | Down Delay (ms): 0 | |
1423 | ||
1424 | Slave Interface: eth1 | |
1425 | MII Status: up | |
1426 | Link Failure Count: 1 | |
1427 | ||
1428 | Slave Interface: eth0 | |
1429 | MII Status: up | |
1430 | Link Failure Count: 1 | |
1431 | ||
1432 | The precise format and contents will change depending upon the | |
1433 | bonding configuration, state, and version of the bonding driver. | |
1434 | ||
6224e01d | 1435 | 4.2 Network configuration |
1da177e4 LT |
1436 | ------------------------- |
1437 | ||
1438 | The network configuration can be inspected using the ifconfig | |
1439 | command. Bonding devices will have the MASTER flag set; Bonding slave | |
1440 | devices will have the SLAVE flag set. The ifconfig output does not | |
1441 | contain information on which slaves are associated with which masters. | |
1442 | ||
1443 | In the example below, the bond0 interface is the master | |
1444 | (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of | |
1445 | bond0 have the same MAC address (HWaddr) as bond0 for all modes except | |
1446 | TLB and ALB that require a unique MAC address for each slave. | |
1447 | ||
1448 | # /sbin/ifconfig | |
1449 | bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4 | |
1450 | inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0 | |
1451 | UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1 | |
1452 | RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0 | |
1453 | TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0 | |
1454 | collisions:0 txqueuelen:0 | |
1455 | ||
1456 | eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4 | |
1da177e4 LT |
1457 | UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1 |
1458 | RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0 | |
1459 | TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0 | |
1460 | collisions:0 txqueuelen:100 | |
1461 | Interrupt:10 Base address:0x1080 | |
1462 | ||
1463 | eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4 | |
1da177e4 LT |
1464 | UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1 |
1465 | RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0 | |
1466 | TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0 | |
1467 | collisions:0 txqueuelen:100 | |
1468 | Interrupt:9 Base address:0x1400 | |
1469 | ||
6224e01d | 1470 | 5. Switch Configuration |
1da177e4 LT |
1471 | ======================= |
1472 | ||
1473 | For this section, "switch" refers to whatever system the | |
1474 | bonded devices are directly connected to (i.e., where the other end of | |
1475 | the cable plugs into). This may be an actual dedicated switch device, | |
1476 | or it may be another regular system (e.g., another computer running | |
1477 | Linux), | |
1478 | ||
1479 | The active-backup, balance-tlb and balance-alb modes do not | |
1480 | require any specific configuration of the switch. | |
1481 | ||
1482 | The 802.3ad mode requires that the switch have the appropriate | |
1483 | ports configured as an 802.3ad aggregation. The precise method used | |
1484 | to configure this varies from switch to switch, but, for example, a | |
1485 | Cisco 3550 series switch requires that the appropriate ports first be | |
1486 | grouped together in a single etherchannel instance, then that | |
1487 | etherchannel is set to mode "lacp" to enable 802.3ad (instead of | |
1488 | standard EtherChannel). | |
1489 | ||
1490 | The balance-rr, balance-xor and broadcast modes generally | |
1491 | require that the switch have the appropriate ports grouped together. | |
1492 | The nomenclature for such a group differs between switches, it may be | |
1493 | called an "etherchannel" (as in the Cisco example, above), a "trunk | |
1494 | group" or some other similar variation. For these modes, each switch | |
1495 | will also have its own configuration options for the switch's transmit | |
1496 | policy to the bond. Typical choices include XOR of either the MAC or | |
1497 | IP addresses. The transmit policy of the two peers does not need to | |
1498 | match. For these three modes, the bonding mode really selects a | |
1499 | transmit policy for an EtherChannel group; all three will interoperate | |
1500 | with another EtherChannel group. | |
1501 | ||
1502 | ||
6224e01d | 1503 | 6. 802.1q VLAN Support |
1da177e4 LT |
1504 | ====================== |
1505 | ||
1506 | It is possible to configure VLAN devices over a bond interface | |
1507 | using the 8021q driver. However, only packets coming from the 8021q | |
1508 | driver and passing through bonding will be tagged by default. Self | |
1509 | generated packets, for example, bonding's learning packets or ARP | |
1510 | packets generated by either ALB mode or the ARP monitor mechanism, are | |
1511 | tagged internally by bonding itself. As a result, bonding must | |
1512 | "learn" the VLAN IDs configured above it, and use those IDs to tag | |
1513 | self generated packets. | |
1514 | ||
1515 | For reasons of simplicity, and to support the use of adapters | |
00354cfb JV |
1516 | that can do VLAN hardware acceleration offloading, the bonding |
1517 | interface declares itself as fully hardware offloading capable, it gets | |
1da177e4 LT |
1518 | the add_vid/kill_vid notifications to gather the necessary |
1519 | information, and it propagates those actions to the slaves. In case | |
1520 | of mixed adapter types, hardware accelerated tagged packets that | |
1521 | should go through an adapter that is not offloading capable are | |
1522 | "un-accelerated" by the bonding driver so the VLAN tag sits in the | |
1523 | regular location. | |
1524 | ||
1525 | VLAN interfaces *must* be added on top of a bonding interface | |
1526 | only after enslaving at least one slave. The bonding interface has a | |
1527 | hardware address of 00:00:00:00:00:00 until the first slave is added. | |
1528 | If the VLAN interface is created prior to the first enslavement, it | |
1529 | would pick up the all-zeroes hardware address. Once the first slave | |
1530 | is attached to the bond, the bond device itself will pick up the | |
1531 | slave's hardware address, which is then available for the VLAN device. | |
1532 | ||
1533 | Also, be aware that a similar problem can occur if all slaves | |
1534 | are released from a bond that still has one or more VLAN interfaces on | |
1535 | top of it. When a new slave is added, the bonding interface will | |
1536 | obtain its hardware address from the first slave, which might not | |
1537 | match the hardware address of the VLAN interfaces (which was | |
1538 | ultimately copied from an earlier slave). | |
1539 | ||
1540 | There are two methods to insure that the VLAN device operates | |
1541 | with the correct hardware address if all slaves are removed from a | |
1542 | bond interface: | |
1543 | ||
1544 | 1. Remove all VLAN interfaces then recreate them | |
1545 | ||
1546 | 2. Set the bonding interface's hardware address so that it | |
1547 | matches the hardware address of the VLAN interfaces. | |
1548 | ||
1549 | Note that changing a VLAN interface's HW address would set the | |
00354cfb | 1550 | underlying device -- i.e. the bonding interface -- to promiscuous |
1da177e4 LT |
1551 | mode, which might not be what you want. |
1552 | ||
1553 | ||
6224e01d | 1554 | 7. Link Monitoring |
1da177e4 LT |
1555 | ================== |
1556 | ||
1557 | The bonding driver at present supports two schemes for | |
1558 | monitoring a slave device's link state: the ARP monitor and the MII | |
1559 | monitor. | |
1560 | ||
1561 | At the present time, due to implementation restrictions in the | |
1562 | bonding driver itself, it is not possible to enable both ARP and MII | |
1563 | monitoring simultaneously. | |
1564 | ||
6224e01d | 1565 | 7.1 ARP Monitor Operation |
1da177e4 LT |
1566 | ------------------------- |
1567 | ||
1568 | The ARP monitor operates as its name suggests: it sends ARP | |
1569 | queries to one or more designated peer systems on the network, and | |
1570 | uses the response as an indication that the link is operating. This | |
1571 | gives some assurance that traffic is actually flowing to and from one | |
1572 | or more peers on the local network. | |
1573 | ||
1574 | The ARP monitor relies on the device driver itself to verify | |
1575 | that traffic is flowing. In particular, the driver must keep up to | |
1576 | date the last receive time, dev->last_rx, and transmit start time, | |
1577 | dev->trans_start. If these are not updated by the driver, then the | |
1578 | ARP monitor will immediately fail any slaves using that driver, and | |
1579 | those slaves will stay down. If networking monitoring (tcpdump, etc) | |
1580 | shows the ARP requests and replies on the network, then it may be that | |
1581 | your device driver is not updating last_rx and trans_start. | |
1582 | ||
6224e01d | 1583 | 7.2 Configuring Multiple ARP Targets |
1da177e4 LT |
1584 | ------------------------------------ |
1585 | ||
1586 | While ARP monitoring can be done with just one target, it can | |
1587 | be useful in a High Availability setup to have several targets to | |
1588 | monitor. In the case of just one target, the target itself may go | |
1589 | down or have a problem making it unresponsive to ARP requests. Having | |
1590 | an additional target (or several) increases the reliability of the ARP | |
1591 | monitoring. | |
1592 | ||
00354cfb | 1593 | Multiple ARP targets must be separated by commas as follows: |
1da177e4 LT |
1594 | |
1595 | # example options for ARP monitoring with three targets | |
1596 | alias bond0 bonding | |
1597 | options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9 | |
1598 | ||
1599 | For just a single target the options would resemble: | |
1600 | ||
1601 | # example options for ARP monitoring with one target | |
1602 | alias bond0 bonding | |
1603 | options bond0 arp_interval=60 arp_ip_target=192.168.0.100 | |
1604 | ||
1605 | ||
6224e01d | 1606 | 7.3 MII Monitor Operation |
1da177e4 LT |
1607 | ------------------------- |
1608 | ||
1609 | The MII monitor monitors only the carrier state of the local | |
1610 | network interface. It accomplishes this in one of three ways: by | |
1611 | depending upon the device driver to maintain its carrier state, by | |
1612 | querying the device's MII registers, or by making an ethtool query to | |
1613 | the device. | |
1614 | ||
1615 | If the use_carrier module parameter is 1 (the default value), | |
1616 | then the MII monitor will rely on the driver for carrier state | |
1617 | information (via the netif_carrier subsystem). As explained in the | |
1618 | use_carrier parameter information, above, if the MII monitor fails to | |
1619 | detect carrier loss on the device (e.g., when the cable is physically | |
1620 | disconnected), it may be that the driver does not support | |
1621 | netif_carrier. | |
1622 | ||
1623 | If use_carrier is 0, then the MII monitor will first query the | |
1624 | device's (via ioctl) MII registers and check the link state. If that | |
1625 | request fails (not just that it returns carrier down), then the MII | |
1626 | monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain | |
1627 | the same information. If both methods fail (i.e., the driver either | |
1628 | does not support or had some error in processing both the MII register | |
1629 | and ethtool requests), then the MII monitor will assume the link is | |
1630 | up. | |
1631 | ||
6224e01d | 1632 | 8. Potential Sources of Trouble |
1da177e4 LT |
1633 | =============================== |
1634 | ||
6224e01d | 1635 | 8.1 Adventures in Routing |
1da177e4 LT |
1636 | ------------------------- |
1637 | ||
1638 | When bonding is configured, it is important that the slave | |
6224e01d | 1639 | devices not have routes that supersede routes of the master (or, |
1da177e4 LT |
1640 | generally, not have routes at all). For example, suppose the bonding |
1641 | device bond0 has two slaves, eth0 and eth1, and the routing table is | |
1642 | as follows: | |
1643 | ||
1644 | Kernel IP routing table | |
1645 | Destination Gateway Genmask Flags MSS Window irtt Iface | |
1646 | 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0 | |
1647 | 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1 | |
1648 | 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0 | |
1649 | 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo | |
1650 | ||
1651 | This routing configuration will likely still update the | |
1652 | receive/transmit times in the driver (needed by the ARP monitor), but | |
1653 | may bypass the bonding driver (because outgoing traffic to, in this | |
1654 | case, another host on network 10 would use eth0 or eth1 before bond0). | |
1655 | ||
1656 | The ARP monitor (and ARP itself) may become confused by this | |
1657 | configuration, because ARP requests (generated by the ARP monitor) | |
1658 | will be sent on one interface (bond0), but the corresponding reply | |
1659 | will arrive on a different interface (eth0). This reply looks to ARP | |
1660 | as an unsolicited ARP reply (because ARP matches replies on an | |
1661 | interface basis), and is discarded. The MII monitor is not affected | |
1662 | by the state of the routing table. | |
1663 | ||
1664 | The solution here is simply to insure that slaves do not have | |
1665 | routes of their own, and if for some reason they must, those routes do | |
6224e01d | 1666 | not supersede routes of their master. This should generally be the |
1da177e4 LT |
1667 | case, but unusual configurations or errant manual or automatic static |
1668 | route additions may cause trouble. | |
1669 | ||
6224e01d | 1670 | 8.2 Ethernet Device Renaming |
1da177e4 LT |
1671 | ---------------------------- |
1672 | ||
1673 | On systems with network configuration scripts that do not | |
1674 | associate physical devices directly with network interface names (so | |
1675 | that the same physical device always has the same "ethX" name), it may | |
1676 | be necessary to add some special logic to either /etc/modules.conf or | |
1677 | /etc/modprobe.conf (depending upon which is installed on the system). | |
1678 | ||
1679 | For example, given a modules.conf containing the following: | |
1680 | ||
1681 | alias bond0 bonding | |
1682 | options bond0 mode=some-mode miimon=50 | |
1683 | alias eth0 tg3 | |
1684 | alias eth1 tg3 | |
1685 | alias eth2 e1000 | |
1686 | alias eth3 e1000 | |
1687 | ||
1688 | If neither eth0 and eth1 are slaves to bond0, then when the | |
1689 | bond0 interface comes up, the devices may end up reordered. This | |
1690 | happens because bonding is loaded first, then its slave device's | |
1691 | drivers are loaded next. Since no other drivers have been loaded, | |
1692 | when the e1000 driver loads, it will receive eth0 and eth1 for its | |
1693 | devices, but the bonding configuration tries to enslave eth2 and eth3 | |
1694 | (which may later be assigned to the tg3 devices). | |
1695 | ||
1696 | Adding the following: | |
1697 | ||
1698 | add above bonding e1000 tg3 | |
1699 | ||
1700 | causes modprobe to load e1000 then tg3, in that order, when | |
1701 | bonding is loaded. This command is fully documented in the | |
1702 | modules.conf manual page. | |
1703 | ||
1704 | On systems utilizing modprobe.conf (or modprobe.conf.local), | |
1705 | an equivalent problem can occur. In this case, the following can be | |
1706 | added to modprobe.conf (or modprobe.conf.local, as appropriate), as | |
1707 | follows (all on one line; it has been split here for clarity): | |
1708 | ||
1709 | install bonding /sbin/modprobe tg3; /sbin/modprobe e1000; | |
1710 | /sbin/modprobe --ignore-install bonding | |
1711 | ||
1712 | This will, when loading the bonding module, rather than | |
1713 | performing the normal action, instead execute the provided command. | |
1714 | This command loads the device drivers in the order needed, then calls | |
00354cfb | 1715 | modprobe with --ignore-install to cause the normal action to then take |
1da177e4 LT |
1716 | place. Full documentation on this can be found in the modprobe.conf |
1717 | and modprobe manual pages. | |
1718 | ||
6224e01d | 1719 | 8.3. Painfully Slow Or No Failed Link Detection By Miimon |
1da177e4 LT |
1720 | --------------------------------------------------------- |
1721 | ||
1722 | By default, bonding enables the use_carrier option, which | |
1723 | instructs bonding to trust the driver to maintain carrier state. | |
1724 | ||
1725 | As discussed in the options section, above, some drivers do | |
1726 | not support the netif_carrier_on/_off link state tracking system. | |
1727 | With use_carrier enabled, bonding will always see these links as up, | |
1728 | regardless of their actual state. | |
1729 | ||
1730 | Additionally, other drivers do support netif_carrier, but do | |
1731 | not maintain it in real time, e.g., only polling the link state at | |
1732 | some fixed interval. In this case, miimon will detect failures, but | |
1733 | only after some long period of time has expired. If it appears that | |
1734 | miimon is very slow in detecting link failures, try specifying | |
1735 | use_carrier=0 to see if that improves the failure detection time. If | |
1736 | it does, then it may be that the driver checks the carrier state at a | |
1737 | fixed interval, but does not cache the MII register values (so the | |
1738 | use_carrier=0 method of querying the registers directly works). If | |
1739 | use_carrier=0 does not improve the failover, then the driver may cache | |
1740 | the registers, or the problem may be elsewhere. | |
1741 | ||
1742 | Also, remember that miimon only checks for the device's | |
1743 | carrier state. It has no way to determine the state of devices on or | |
1744 | beyond other ports of a switch, or if a switch is refusing to pass | |
1745 | traffic while still maintaining carrier on. | |
1746 | ||
6224e01d | 1747 | 9. SNMP agents |
1da177e4 LT |
1748 | =============== |
1749 | ||
1750 | If running SNMP agents, the bonding driver should be loaded | |
1751 | before any network drivers participating in a bond. This requirement | |
d533f671 | 1752 | is due to the interface index (ipAdEntIfIndex) being associated to |
1da177e4 LT |
1753 | the first interface found with a given IP address. That is, there is |
1754 | only one ipAdEntIfIndex for each IP address. For example, if eth0 and | |
1755 | eth1 are slaves of bond0 and the driver for eth0 is loaded before the | |
1756 | bonding driver, the interface for the IP address will be associated | |
1757 | with the eth0 interface. This configuration is shown below, the IP | |
1758 | address 192.168.1.1 has an interface index of 2 which indexes to eth0 | |
1759 | in the ifDescr table (ifDescr.2). | |
1760 | ||
1761 | interfaces.ifTable.ifEntry.ifDescr.1 = lo | |
1762 | interfaces.ifTable.ifEntry.ifDescr.2 = eth0 | |
1763 | interfaces.ifTable.ifEntry.ifDescr.3 = eth1 | |
1764 | interfaces.ifTable.ifEntry.ifDescr.4 = eth2 | |
1765 | interfaces.ifTable.ifEntry.ifDescr.5 = eth3 | |
1766 | interfaces.ifTable.ifEntry.ifDescr.6 = bond0 | |
1767 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5 | |
1768 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2 | |
1769 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4 | |
1770 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1 | |
1771 | ||
1772 | This problem is avoided by loading the bonding driver before | |
1773 | any network drivers participating in a bond. Below is an example of | |
1774 | loading the bonding driver first, the IP address 192.168.1.1 is | |
1775 | correctly associated with ifDescr.2. | |
1776 | ||
1777 | interfaces.ifTable.ifEntry.ifDescr.1 = lo | |
1778 | interfaces.ifTable.ifEntry.ifDescr.2 = bond0 | |
1779 | interfaces.ifTable.ifEntry.ifDescr.3 = eth0 | |
1780 | interfaces.ifTable.ifEntry.ifDescr.4 = eth1 | |
1781 | interfaces.ifTable.ifEntry.ifDescr.5 = eth2 | |
1782 | interfaces.ifTable.ifEntry.ifDescr.6 = eth3 | |
1783 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6 | |
1784 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2 | |
1785 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5 | |
1786 | ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1 | |
1787 | ||
1788 | While some distributions may not report the interface name in | |
1789 | ifDescr, the association between the IP address and IfIndex remains | |
1790 | and SNMP functions such as Interface_Scan_Next will report that | |
1791 | association. | |
1792 | ||
6224e01d | 1793 | 10. Promiscuous mode |
1da177e4 LT |
1794 | ==================== |
1795 | ||
1796 | When running network monitoring tools, e.g., tcpdump, it is | |
1797 | common to enable promiscuous mode on the device, so that all traffic | |
1798 | is seen (instead of seeing only traffic destined for the local host). | |
1799 | The bonding driver handles promiscuous mode changes to the bonding | |
00354cfb | 1800 | master device (e.g., bond0), and propagates the setting to the slave |
1da177e4 LT |
1801 | devices. |
1802 | ||
1803 | For the balance-rr, balance-xor, broadcast, and 802.3ad modes, | |
00354cfb | 1804 | the promiscuous mode setting is propagated to all slaves. |
1da177e4 LT |
1805 | |
1806 | For the active-backup, balance-tlb and balance-alb modes, the | |
00354cfb | 1807 | promiscuous mode setting is propagated only to the active slave. |
1da177e4 LT |
1808 | |
1809 | For balance-tlb mode, the active slave is the slave currently | |
1810 | receiving inbound traffic. | |
1811 | ||
1812 | For balance-alb mode, the active slave is the slave used as a | |
1813 | "primary." This slave is used for mode-specific control traffic, for | |
1814 | sending to peers that are unassigned or if the load is unbalanced. | |
1815 | ||
1816 | For the active-backup, balance-tlb and balance-alb modes, when | |
1817 | the active slave changes (e.g., due to a link failure), the | |
00354cfb | 1818 | promiscuous setting will be propagated to the new active slave. |
1da177e4 | 1819 | |
6224e01d | 1820 | 11. Configuring Bonding for High Availability |
00354cfb | 1821 | ============================================= |
1da177e4 LT |
1822 | |
1823 | High Availability refers to configurations that provide | |
1824 | maximum network availability by having redundant or backup devices, | |
00354cfb JV |
1825 | links or switches between the host and the rest of the world. The |
1826 | goal is to provide the maximum availability of network connectivity | |
1827 | (i.e., the network always works), even though other configurations | |
1828 | could provide higher throughput. | |
1da177e4 | 1829 | |
6224e01d | 1830 | 11.1 High Availability in a Single Switch Topology |
1da177e4 LT |
1831 | -------------------------------------------------- |
1832 | ||
00354cfb JV |
1833 | If two hosts (or a host and a single switch) are directly |
1834 | connected via multiple physical links, then there is no availability | |
1835 | penalty to optimizing for maximum bandwidth. In this case, there is | |
1836 | only one switch (or peer), so if it fails, there is no alternative | |
1837 | access to fail over to. Additionally, the bonding load balance modes | |
1838 | support link monitoring of their members, so if individual links fail, | |
1839 | the load will be rebalanced across the remaining devices. | |
1840 | ||
1841 | See Section 13, "Configuring Bonding for Maximum Throughput" | |
1842 | for information on configuring bonding with one peer device. | |
1843 | ||
6224e01d | 1844 | 11.2 High Availability in a Multiple Switch Topology |
00354cfb JV |
1845 | ---------------------------------------------------- |
1846 | ||
1847 | With multiple switches, the configuration of bonding and the | |
1848 | network changes dramatically. In multiple switch topologies, there is | |
1849 | a trade off between network availability and usable bandwidth. | |
1850 | ||
1851 | Below is a sample network, configured to maximize the | |
1852 | availability of the network: | |
1da177e4 | 1853 | |
00354cfb JV |
1854 | | | |
1855 | |port3 port3| | |
1856 | +-----+----+ +-----+----+ | |
1857 | | |port2 ISL port2| | | |
1858 | | switch A +--------------------------+ switch B | | |
1859 | | | | | | |
1860 | +-----+----+ +-----++---+ | |
1861 | |port1 port1| | |
1862 | | +-------+ | | |
1863 | +-------------+ host1 +---------------+ | |
1864 | eth0 +-------+ eth1 | |
1da177e4 | 1865 | |
00354cfb JV |
1866 | In this configuration, there is a link between the two |
1867 | switches (ISL, or inter switch link), and multiple ports connecting to | |
1868 | the outside world ("port3" on each switch). There is no technical | |
1869 | reason that this could not be extended to a third switch. | |
1da177e4 | 1870 | |
6224e01d | 1871 | 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology |
00354cfb | 1872 | ------------------------------------------------------------- |
1da177e4 | 1873 | |
00354cfb JV |
1874 | In a topology such as the example above, the active-backup and |
1875 | broadcast modes are the only useful bonding modes when optimizing for | |
1876 | availability; the other modes require all links to terminate on the | |
1877 | same peer for them to behave rationally. | |
1878 | ||
1879 | active-backup: This is generally the preferred mode, particularly if | |
1880 | the switches have an ISL and play together well. If the | |
1881 | network configuration is such that one switch is specifically | |
1882 | a backup switch (e.g., has lower capacity, higher cost, etc), | |
1883 | then the primary option can be used to insure that the | |
1884 | preferred link is always used when it is available. | |
1885 | ||
1886 | broadcast: This mode is really a special purpose mode, and is suitable | |
1887 | only for very specific needs. For example, if the two | |
1888 | switches are not connected (no ISL), and the networks beyond | |
1889 | them are totally independent. In this case, if it is | |
1890 | necessary for some specific one-way traffic to reach both | |
1891 | independent networks, then the broadcast mode may be suitable. | |
1892 | ||
6224e01d | 1893 | 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology |
00354cfb JV |
1894 | ---------------------------------------------------------------- |
1895 | ||
1896 | The choice of link monitoring ultimately depends upon your | |
1897 | switch. If the switch can reliably fail ports in response to other | |
1898 | failures, then either the MII or ARP monitors should work. For | |
1899 | example, in the above example, if the "port3" link fails at the remote | |
1900 | end, the MII monitor has no direct means to detect this. The ARP | |
1901 | monitor could be configured with a target at the remote end of port3, | |
1902 | thus detecting that failure without switch support. | |
1903 | ||
1904 | In general, however, in a multiple switch topology, the ARP | |
1905 | monitor can provide a higher level of reliability in detecting end to | |
1906 | end connectivity failures (which may be caused by the failure of any | |
1907 | individual component to pass traffic for any reason). Additionally, | |
1908 | the ARP monitor should be configured with multiple targets (at least | |
1909 | one for each switch in the network). This will insure that, | |
1910 | regardless of which switch is active, the ARP monitor has a suitable | |
1911 | target to query. | |
1912 | ||
9a6c6867 JV |
1913 | Note, also, that of late many switches now support a functionality |
1914 | generally referred to as "trunk failover." This is a feature of the | |
1915 | switch that causes the link state of a particular switch port to be set | |
1916 | down (or up) when the state of another switch port goes down (or up). | |
19f59460 | 1917 | Its purpose is to propagate link failures from logically "exterior" ports |
9a6c6867 JV |
1918 | to the logically "interior" ports that bonding is able to monitor via |
1919 | miimon. Availability and configuration for trunk failover varies by | |
1920 | switch, but this can be a viable alternative to the ARP monitor when using | |
1921 | suitable switches. | |
00354cfb | 1922 | |
6224e01d | 1923 | 12. Configuring Bonding for Maximum Throughput |
00354cfb JV |
1924 | ============================================== |
1925 | ||
6224e01d | 1926 | 12.1 Maximizing Throughput in a Single Switch Topology |
00354cfb JV |
1927 | ------------------------------------------------------ |
1928 | ||
1929 | In a single switch configuration, the best method to maximize | |
1930 | throughput depends upon the application and network environment. The | |
1931 | various load balancing modes each have strengths and weaknesses in | |
1932 | different environments, as detailed below. | |
1933 | ||
1934 | For this discussion, we will break down the topologies into | |
1935 | two categories. Depending upon the destination of most traffic, we | |
1936 | categorize them into either "gatewayed" or "local" configurations. | |
1937 | ||
1938 | In a gatewayed configuration, the "switch" is acting primarily | |
1939 | as a router, and the majority of traffic passes through this router to | |
1940 | other networks. An example would be the following: | |
1941 | ||
1942 | ||
1943 | +----------+ +----------+ | |
1944 | | |eth0 port1| | to other networks | |
1945 | | Host A +---------------------+ router +-------------------> | |
1946 | | +---------------------+ | Hosts B and C are out | |
1947 | | |eth1 port2| | here somewhere | |
1948 | +----------+ +----------+ | |
1949 | ||
1950 | The router may be a dedicated router device, or another host | |
1951 | acting as a gateway. For our discussion, the important point is that | |
1952 | the majority of traffic from Host A will pass through the router to | |
1953 | some other network before reaching its final destination. | |
1954 | ||
1955 | In a gatewayed network configuration, although Host A may | |
1956 | communicate with many other systems, all of its traffic will be sent | |
1957 | and received via one other peer on the local network, the router. | |
1958 | ||
1959 | Note that the case of two systems connected directly via | |
1960 | multiple physical links is, for purposes of configuring bonding, the | |
1961 | same as a gatewayed configuration. In that case, it happens that all | |
1962 | traffic is destined for the "gateway" itself, not some other network | |
1963 | beyond the gateway. | |
1964 | ||
1965 | In a local configuration, the "switch" is acting primarily as | |
1966 | a switch, and the majority of traffic passes through this switch to | |
1967 | reach other stations on the same network. An example would be the | |
1968 | following: | |
1969 | ||
1970 | +----------+ +----------+ +--------+ | |
1971 | | |eth0 port1| +-------+ Host B | | |
1972 | | Host A +------------+ switch |port3 +--------+ | |
1973 | | +------------+ | +--------+ | |
1974 | | |eth1 port2| +------------------+ Host C | | |
1975 | +----------+ +----------+port4 +--------+ | |
1976 | ||
1977 | ||
1978 | Again, the switch may be a dedicated switch device, or another | |
1979 | host acting as a gateway. For our discussion, the important point is | |
1980 | that the majority of traffic from Host A is destined for other hosts | |
1981 | on the same local network (Hosts B and C in the above example). | |
1982 | ||
1983 | In summary, in a gatewayed configuration, traffic to and from | |
1984 | the bonded device will be to the same MAC level peer on the network | |
1985 | (the gateway itself, i.e., the router), regardless of its final | |
1986 | destination. In a local configuration, traffic flows directly to and | |
1987 | from the final destinations, thus, each destination (Host B, Host C) | |
1988 | will be addressed directly by their individual MAC addresses. | |
1989 | ||
1990 | This distinction between a gatewayed and a local network | |
1991 | configuration is important because many of the load balancing modes | |
1992 | available use the MAC addresses of the local network source and | |
1993 | destination to make load balancing decisions. The behavior of each | |
1994 | mode is described below. | |
1995 | ||
1996 | ||
6224e01d | 1997 | 12.1.1 MT Bonding Mode Selection for Single Switch Topology |
00354cfb | 1998 | ----------------------------------------------------------- |
1da177e4 LT |
1999 | |
2000 | This configuration is the easiest to set up and to understand, | |
2001 | although you will have to decide which bonding mode best suits your | |
00354cfb | 2002 | needs. The trade offs for each mode are detailed below: |
1da177e4 LT |
2003 | |
2004 | balance-rr: This mode is the only mode that will permit a single | |
2005 | TCP/IP connection to stripe traffic across multiple | |
2006 | interfaces. It is therefore the only mode that will allow a | |
2007 | single TCP/IP stream to utilize more than one interface's | |
2008 | worth of throughput. This comes at a cost, however: the | |
9a6c6867 | 2009 | striping generally results in peer systems receiving packets out |
1da177e4 LT |
2010 | of order, causing TCP/IP's congestion control system to kick |
2011 | in, often by retransmitting segments. | |
2012 | ||
2013 | It is possible to adjust TCP/IP's congestion limits by | |
2014 | altering the net.ipv4.tcp_reordering sysctl parameter. The | |
2015 | usual default value is 3, and the maximum useful value is 127. | |
2016 | For a four interface balance-rr bond, expect that a single | |
2017 | TCP/IP stream will utilize no more than approximately 2.3 | |
2018 | interface's worth of throughput, even after adjusting | |
2019 | tcp_reordering. | |
2020 | ||
9a6c6867 JV |
2021 | Note that the fraction of packets that will be delivered out of |
2022 | order is highly variable, and is unlikely to be zero. The level | |
2023 | of reordering depends upon a variety of factors, including the | |
2024 | networking interfaces, the switch, and the topology of the | |
2025 | configuration. Speaking in general terms, higher speed network | |
2026 | cards produce more reordering (due to factors such as packet | |
2027 | coalescing), and a "many to many" topology will reorder at a | |
2028 | higher rate than a "many slow to one fast" configuration. | |
2029 | ||
2030 | Many switches do not support any modes that stripe traffic | |
2031 | (instead choosing a port based upon IP or MAC level addresses); | |
2032 | for those devices, traffic for a particular connection flowing | |
2033 | through the switch to a balance-rr bond will not utilize greater | |
2034 | than one interface's worth of bandwidth. | |
00354cfb | 2035 | |
1da177e4 LT |
2036 | If you are utilizing protocols other than TCP/IP, UDP for |
2037 | example, and your application can tolerate out of order | |
2038 | delivery, then this mode can allow for single stream datagram | |
2039 | performance that scales near linearly as interfaces are added | |
2040 | to the bond. | |
2041 | ||
2042 | This mode requires the switch to have the appropriate ports | |
2043 | configured for "etherchannel" or "trunking." | |
2044 | ||
2045 | active-backup: There is not much advantage in this network topology to | |
2046 | the active-backup mode, as the inactive backup devices are all | |
2047 | connected to the same peer as the primary. In this case, a | |
2048 | load balancing mode (with link monitoring) will provide the | |
2049 | same level of network availability, but with increased | |
00354cfb JV |
2050 | available bandwidth. On the plus side, active-backup mode |
2051 | does not require any configuration of the switch, so it may | |
2052 | have value if the hardware available does not support any of | |
2053 | the load balance modes. | |
1da177e4 LT |
2054 | |
2055 | balance-xor: This mode will limit traffic such that packets destined | |
2056 | for specific peers will always be sent over the same | |
2057 | interface. Since the destination is determined by the MAC | |
00354cfb JV |
2058 | addresses involved, this mode works best in a "local" network |
2059 | configuration (as described above), with destinations all on | |
2060 | the same local network. This mode is likely to be suboptimal | |
2061 | if all your traffic is passed through a single router (i.e., a | |
2062 | "gatewayed" network configuration, as described above). | |
2063 | ||
2064 | As with balance-rr, the switch ports need to be configured for | |
1da177e4 LT |
2065 | "etherchannel" or "trunking." |
2066 | ||
2067 | broadcast: Like active-backup, there is not much advantage to this | |
2068 | mode in this type of network topology. | |
2069 | ||
2070 | 802.3ad: This mode can be a good choice for this type of network | |
2071 | topology. The 802.3ad mode is an IEEE standard, so all peers | |
2072 | that implement 802.3ad should interoperate well. The 802.3ad | |
2073 | protocol includes automatic configuration of the aggregates, | |
2074 | so minimal manual configuration of the switch is needed | |
2075 | (typically only to designate that some set of devices is | |
00354cfb JV |
2076 | available for 802.3ad). The 802.3ad standard also mandates |
2077 | that frames be delivered in order (within certain limits), so | |
2078 | in general single connections will not see misordering of | |
1da177e4 LT |
2079 | packets. The 802.3ad mode does have some drawbacks: the |
2080 | standard mandates that all devices in the aggregate operate at | |
2081 | the same speed and duplex. Also, as with all bonding load | |
2082 | balance modes other than balance-rr, no single connection will | |
2083 | be able to utilize more than a single interface's worth of | |
00354cfb JV |
2084 | bandwidth. |
2085 | ||
2086 | Additionally, the linux bonding 802.3ad implementation | |
2087 | distributes traffic by peer (using an XOR of MAC addresses), | |
2088 | so in a "gatewayed" configuration, all outgoing traffic will | |
2089 | generally use the same device. Incoming traffic may also end | |
2090 | up on a single device, but that is dependent upon the | |
2091 | balancing policy of the peer's 8023.ad implementation. In a | |
2092 | "local" configuration, traffic will be distributed across the | |
2093 | devices in the bond. | |
2094 | ||
2095 | Finally, the 802.3ad mode mandates the use of the MII monitor, | |
2096 | therefore, the ARP monitor is not available in this mode. | |
2097 | ||
2098 | balance-tlb: The balance-tlb mode balances outgoing traffic by peer. | |
2099 | Since the balancing is done according to MAC address, in a | |
2100 | "gatewayed" configuration (as described above), this mode will | |
2101 | send all traffic across a single device. However, in a | |
2102 | "local" network configuration, this mode balances multiple | |
2103 | local network peers across devices in a vaguely intelligent | |
2104 | manner (not a simple XOR as in balance-xor or 802.3ad mode), | |
2105 | so that mathematically unlucky MAC addresses (i.e., ones that | |
2106 | XOR to the same value) will not all "bunch up" on a single | |
2107 | interface. | |
2108 | ||
2109 | Unlike 802.3ad, interfaces may be of differing speeds, and no | |
2110 | special switch configuration is required. On the down side, | |
2111 | in this mode all incoming traffic arrives over a single | |
2112 | interface, this mode requires certain ethtool support in the | |
2113 | network device driver of the slave interfaces, and the ARP | |
2114 | monitor is not available. | |
2115 | ||
2116 | balance-alb: This mode is everything that balance-tlb is, and more. | |
2117 | It has all of the features (and restrictions) of balance-tlb, | |
2118 | and will also balance incoming traffic from local network | |
2119 | peers (as described in the Bonding Module Options section, | |
2120 | above). | |
2121 | ||
2122 | The only additional down side to this mode is that the network | |
2123 | device driver must support changing the hardware address while | |
2124 | the device is open. | |
2125 | ||
6224e01d | 2126 | 12.1.2 MT Link Monitoring for Single Switch Topology |
00354cfb | 2127 | ---------------------------------------------------- |
1da177e4 LT |
2128 | |
2129 | The choice of link monitoring may largely depend upon which | |
2130 | mode you choose to use. The more advanced load balancing modes do not | |
2131 | support the use of the ARP monitor, and are thus restricted to using | |
00354cfb JV |
2132 | the MII monitor (which does not provide as high a level of end to end |
2133 | assurance as the ARP monitor). | |
2134 | ||
6224e01d | 2135 | 12.2 Maximum Throughput in a Multiple Switch Topology |
00354cfb JV |
2136 | ----------------------------------------------------- |
2137 | ||
2138 | Multiple switches may be utilized to optimize for throughput | |
2139 | when they are configured in parallel as part of an isolated network | |
2140 | between two or more systems, for example: | |
2141 | ||
2142 | +-----------+ | |
2143 | | Host A | | |
2144 | +-+---+---+-+ | |
2145 | | | | | |
2146 | +--------+ | +---------+ | |
2147 | | | | | |
2148 | +------+---+ +-----+----+ +-----+----+ | |
2149 | | Switch A | | Switch B | | Switch C | | |
2150 | +------+---+ +-----+----+ +-----+----+ | |
2151 | | | | | |
2152 | +--------+ | +---------+ | |
2153 | | | | | |
2154 | +-+---+---+-+ | |
2155 | | Host B | | |
2156 | +-----------+ | |
2157 | ||
2158 | In this configuration, the switches are isolated from one | |
2159 | another. One reason to employ a topology such as this is for an | |
2160 | isolated network with many hosts (a cluster configured for high | |
2161 | performance, for example), using multiple smaller switches can be more | |
2162 | cost effective than a single larger switch, e.g., on a network with 24 | |
2163 | hosts, three 24 port switches can be significantly less expensive than | |
2164 | a single 72 port switch. | |
2165 | ||
2166 | If access beyond the network is required, an individual host | |
2167 | can be equipped with an additional network device connected to an | |
2168 | external network; this host then additionally acts as a gateway. | |
2169 | ||
6224e01d | 2170 | 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology |
1da177e4 LT |
2171 | ------------------------------------------------------------- |
2172 | ||
00354cfb JV |
2173 | In actual practice, the bonding mode typically employed in |
2174 | configurations of this type is balance-rr. Historically, in this | |
2175 | network configuration, the usual caveats about out of order packet | |
2176 | delivery are mitigated by the use of network adapters that do not do | |
2177 | any kind of packet coalescing (via the use of NAPI, or because the | |
2178 | device itself does not generate interrupts until some number of | |
2179 | packets has arrived). When employed in this fashion, the balance-rr | |
2180 | mode allows individual connections between two hosts to effectively | |
2181 | utilize greater than one interface's bandwidth. | |
1da177e4 | 2182 | |
6224e01d | 2183 | 12.2.2 MT Link Monitoring for Multiple Switch Topology |
00354cfb | 2184 | ------------------------------------------------------ |
1da177e4 | 2185 | |
00354cfb JV |
2186 | Again, in actual practice, the MII monitor is most often used |
2187 | in this configuration, as performance is given preference over | |
2188 | availability. The ARP monitor will function in this topology, but its | |
2189 | advantages over the MII monitor are mitigated by the volume of probes | |
2190 | needed as the number of systems involved grows (remember that each | |
2191 | host in the network is configured with bonding). | |
1da177e4 | 2192 | |
6224e01d | 2193 | 13. Switch Behavior Issues |
00354cfb | 2194 | ========================== |
1da177e4 | 2195 | |
6224e01d | 2196 | 13.1 Link Establishment and Failover Delays |
00354cfb JV |
2197 | ------------------------------------------- |
2198 | ||
2199 | Some switches exhibit undesirable behavior with regard to the | |
2200 | timing of link up and down reporting by the switch. | |
1da177e4 LT |
2201 | |
2202 | First, when a link comes up, some switches may indicate that | |
2203 | the link is up (carrier available), but not pass traffic over the | |
2204 | interface for some period of time. This delay is typically due to | |
2205 | some type of autonegotiation or routing protocol, but may also occur | |
2206 | during switch initialization (e.g., during recovery after a switch | |
2207 | failure). If you find this to be a problem, specify an appropriate | |
2208 | value to the updelay bonding module option to delay the use of the | |
2209 | relevant interface(s). | |
2210 | ||
2211 | Second, some switches may "bounce" the link state one or more | |
2212 | times while a link is changing state. This occurs most commonly while | |
2213 | the switch is initializing. Again, an appropriate updelay value may | |
00354cfb | 2214 | help. |
1da177e4 LT |
2215 | |
2216 | Note that when a bonding interface has no active links, the | |
00354cfb JV |
2217 | driver will immediately reuse the first link that goes up, even if the |
2218 | updelay parameter has been specified (the updelay is ignored in this | |
2219 | case). If there are slave interfaces waiting for the updelay timeout | |
2220 | to expire, the interface that first went into that state will be | |
2221 | immediately reused. This reduces down time of the network if the | |
2222 | value of updelay has been overestimated, and since this occurs only in | |
2223 | cases with no connectivity, there is no additional penalty for | |
2224 | ignoring the updelay. | |
1da177e4 LT |
2225 | |
2226 | In addition to the concerns about switch timings, if your | |
2227 | switches take a long time to go into backup mode, it may be desirable | |
2228 | to not activate a backup interface immediately after a link goes down. | |
2229 | Failover may be delayed via the downdelay bonding module option. | |
2230 | ||
6224e01d | 2231 | 13.2 Duplicated Incoming Packets |
00354cfb JV |
2232 | -------------------------------- |
2233 | ||
9a6c6867 JV |
2234 | NOTE: Starting with version 3.0.2, the bonding driver has logic to |
2235 | suppress duplicate packets, which should largely eliminate this problem. | |
2236 | The following description is kept for reference. | |
2237 | ||
00354cfb JV |
2238 | It is not uncommon to observe a short burst of duplicated |
2239 | traffic when the bonding device is first used, or after it has been | |
2240 | idle for some period of time. This is most easily observed by issuing | |
2241 | a "ping" to some other host on the network, and noticing that the | |
2242 | output from ping flags duplicates (typically one per slave). | |
2243 | ||
2244 | For example, on a bond in active-backup mode with five slaves | |
2245 | all connected to one switch, the output may appear as follows: | |
2246 | ||
2247 | # ping -n 10.0.4.2 | |
2248 | PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data. | |
2249 | 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms | |
2250 | 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) | |
2251 | 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) | |
2252 | 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) | |
2253 | 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) | |
2254 | 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms | |
2255 | 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms | |
2256 | 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms | |
2257 | ||
2258 | This is not due to an error in the bonding driver, rather, it | |
2259 | is a side effect of how many switches update their MAC forwarding | |
2260 | tables. Initially, the switch does not associate the MAC address in | |
2261 | the packet with a particular switch port, and so it may send the | |
2262 | traffic to all ports until its MAC forwarding table is updated. Since | |
2263 | the interfaces attached to the bond may occupy multiple ports on a | |
2264 | single switch, when the switch (temporarily) floods the traffic to all | |
2265 | ports, the bond device receives multiple copies of the same packet | |
2266 | (one per slave device). | |
2267 | ||
2268 | The duplicated packet behavior is switch dependent, some | |
2269 | switches exhibit this, and some do not. On switches that display this | |
2270 | behavior, it can be induced by clearing the MAC forwarding table (on | |
2271 | most Cisco switches, the privileged command "clear mac address-table | |
2272 | dynamic" will accomplish this). | |
2273 | ||
6224e01d | 2274 | 14. Hardware Specific Considerations |
1da177e4 LT |
2275 | ==================================== |
2276 | ||
2277 | This section contains additional information for configuring | |
2278 | bonding on specific hardware platforms, or for interfacing bonding | |
2279 | with particular switches or other devices. | |
2280 | ||
6224e01d | 2281 | 14.1 IBM BladeCenter |
1da177e4 LT |
2282 | -------------------- |
2283 | ||
2284 | This applies to the JS20 and similar systems. | |
2285 | ||
2286 | On the JS20 blades, the bonding driver supports only | |
2287 | balance-rr, active-backup, balance-tlb and balance-alb modes. This is | |
2288 | largely due to the network topology inside the BladeCenter, detailed | |
2289 | below. | |
2290 | ||
2291 | JS20 network adapter information | |
2292 | -------------------------------- | |
2293 | ||
2294 | All JS20s come with two Broadcom Gigabit Ethernet ports | |
00354cfb JV |
2295 | integrated on the planar (that's "motherboard" in IBM-speak). In the |
2296 | BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to | |
2297 | I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2. | |
2298 | An add-on Broadcom daughter card can be installed on a JS20 to provide | |
2299 | two more Gigabit Ethernet ports. These ports, eth2 and eth3, are | |
2300 | wired to I/O Modules 3 and 4, respectively. | |
1da177e4 LT |
2301 | |
2302 | Each I/O Module may contain either a switch or a passthrough | |
2303 | module (which allows ports to be directly connected to an external | |
2304 | switch). Some bonding modes require a specific BladeCenter internal | |
2305 | network topology in order to function; these are detailed below. | |
2306 | ||
2307 | Additional BladeCenter-specific networking information can be | |
2308 | found in two IBM Redbooks (www.ibm.com/redbooks): | |
2309 | ||
2310 | "IBM eServer BladeCenter Networking Options" | |
2311 | "IBM eServer BladeCenter Layer 2-7 Network Switching" | |
2312 | ||
2313 | BladeCenter networking configuration | |
2314 | ------------------------------------ | |
2315 | ||
2316 | Because a BladeCenter can be configured in a very large number | |
2317 | of ways, this discussion will be confined to describing basic | |
2318 | configurations. | |
2319 | ||
00354cfb | 2320 | Normally, Ethernet Switch Modules (ESMs) are used in I/O |
1da177e4 LT |
2321 | modules 1 and 2. In this configuration, the eth0 and eth1 ports of a |
2322 | JS20 will be connected to different internal switches (in the | |
2323 | respective I/O modules). | |
2324 | ||
00354cfb JV |
2325 | A passthrough module (OPM or CPM, optical or copper, |
2326 | passthrough module) connects the I/O module directly to an external | |
2327 | switch. By using PMs in I/O module #1 and #2, the eth0 and eth1 | |
2328 | interfaces of a JS20 can be redirected to the outside world and | |
2329 | connected to a common external switch. | |
2330 | ||
2331 | Depending upon the mix of ESMs and PMs, the network will | |
2332 | appear to bonding as either a single switch topology (all PMs) or as a | |
2333 | multiple switch topology (one or more ESMs, zero or more PMs). It is | |
2334 | also possible to connect ESMs together, resulting in a configuration | |
2335 | much like the example in "High Availability in a Multiple Switch | |
2336 | Topology," above. | |
2337 | ||
2338 | Requirements for specific modes | |
2339 | ------------------------------- | |
2340 | ||
2341 | The balance-rr mode requires the use of passthrough modules | |
2342 | for devices in the bond, all connected to an common external switch. | |
2343 | That switch must be configured for "etherchannel" or "trunking" on the | |
1da177e4 LT |
2344 | appropriate ports, as is usual for balance-rr. |
2345 | ||
2346 | The balance-alb and balance-tlb modes will function with | |
2347 | either switch modules or passthrough modules (or a mix). The only | |
2348 | specific requirement for these modes is that all network interfaces | |
2349 | must be able to reach all destinations for traffic sent over the | |
2350 | bonding device (i.e., the network must converge at some point outside | |
2351 | the BladeCenter). | |
2352 | ||
2353 | The active-backup mode has no additional requirements. | |
2354 | ||
2355 | Link monitoring issues | |
2356 | ---------------------- | |
2357 | ||
2358 | When an Ethernet Switch Module is in place, only the ARP | |
2359 | monitor will reliably detect link loss to an external switch. This is | |
2360 | nothing unusual, but examination of the BladeCenter cabinet would | |
2361 | suggest that the "external" network ports are the ethernet ports for | |
2362 | the system, when it fact there is a switch between these "external" | |
2363 | ports and the devices on the JS20 system itself. The MII monitor is | |
2364 | only able to detect link failures between the ESM and the JS20 system. | |
2365 | ||
2366 | When a passthrough module is in place, the MII monitor does | |
2367 | detect failures to the "external" port, which is then directly | |
2368 | connected to the JS20 system. | |
2369 | ||
2370 | Other concerns | |
2371 | -------------- | |
2372 | ||
00354cfb | 2373 | The Serial Over LAN (SoL) link is established over the primary |
1da177e4 LT |
2374 | ethernet (eth0) only, therefore, any loss of link to eth0 will result |
2375 | in losing your SoL connection. It will not fail over with other | |
00354cfb JV |
2376 | network traffic, as the SoL system is beyond the control of the |
2377 | bonding driver. | |
1da177e4 LT |
2378 | |
2379 | It may be desirable to disable spanning tree on the switch | |
2380 | (either the internal Ethernet Switch Module, or an external switch) to | |
00354cfb | 2381 | avoid fail-over delay issues when using bonding. |
1da177e4 LT |
2382 | |
2383 | ||
6224e01d | 2384 | 15. Frequently Asked Questions |
1da177e4 LT |
2385 | ============================== |
2386 | ||
2387 | 1. Is it SMP safe? | |
2388 | ||
2389 | Yes. The old 2.0.xx channel bonding patch was not SMP safe. | |
2390 | The new driver was designed to be SMP safe from the start. | |
2391 | ||
2392 | 2. What type of cards will work with it? | |
2393 | ||
2394 | Any Ethernet type cards (you can even mix cards - a Intel | |
00354cfb JV |
2395 | EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes, |
2396 | devices need not be of the same speed. | |
1da177e4 | 2397 | |
9a6c6867 JV |
2398 | Starting with version 3.2.1, bonding also supports Infiniband |
2399 | slaves in active-backup mode. | |
2400 | ||
1da177e4 LT |
2401 | 3. How many bonding devices can I have? |
2402 | ||
2403 | There is no limit. | |
2404 | ||
2405 | 4. How many slaves can a bonding device have? | |
2406 | ||
2407 | This is limited only by the number of network interfaces Linux | |
2408 | supports and/or the number of network cards you can place in your | |
2409 | system. | |
2410 | ||
2411 | 5. What happens when a slave link dies? | |
2412 | ||
2413 | If link monitoring is enabled, then the failing device will be | |
2414 | disabled. The active-backup mode will fail over to a backup link, and | |
2415 | other modes will ignore the failed link. The link will continue to be | |
2416 | monitored, and should it recover, it will rejoin the bond (in whatever | |
00354cfb JV |
2417 | manner is appropriate for the mode). See the sections on High |
2418 | Availability and the documentation for each mode for additional | |
2419 | information. | |
1da177e4 LT |
2420 | |
2421 | Link monitoring can be enabled via either the miimon or | |
00354cfb | 2422 | arp_interval parameters (described in the module parameters section, |
1da177e4 LT |
2423 | above). In general, miimon monitors the carrier state as sensed by |
2424 | the underlying network device, and the arp monitor (arp_interval) | |
2425 | monitors connectivity to another host on the local network. | |
2426 | ||
2427 | If no link monitoring is configured, the bonding driver will | |
2428 | be unable to detect link failures, and will assume that all links are | |
2429 | always available. This will likely result in lost packets, and a | |
00354cfb | 2430 | resulting degradation of performance. The precise performance loss |
1da177e4 LT |
2431 | depends upon the bonding mode and network configuration. |
2432 | ||
2433 | 6. Can bonding be used for High Availability? | |
2434 | ||
2435 | Yes. See the section on High Availability for details. | |
2436 | ||
2437 | 7. Which switches/systems does it work with? | |
2438 | ||
2439 | The full answer to this depends upon the desired mode. | |
2440 | ||
2441 | In the basic balance modes (balance-rr and balance-xor), it | |
2442 | works with any system that supports etherchannel (also called | |
2443 | trunking). Most managed switches currently available have such | |
00354cfb | 2444 | support, and many unmanaged switches as well. |
1da177e4 LT |
2445 | |
2446 | The advanced balance modes (balance-tlb and balance-alb) do | |
2447 | not have special switch requirements, but do need device drivers that | |
2448 | support specific features (described in the appropriate section under | |
00354cfb | 2449 | module parameters, above). |
1da177e4 | 2450 | |
6224e01d | 2451 | In 802.3ad mode, it works with systems that support IEEE |
1da177e4 LT |
2452 | 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged |
2453 | switches currently available support 802.3ad. | |
2454 | ||
2455 | The active-backup mode should work with any Layer-II switch. | |
2456 | ||
2457 | 8. Where does a bonding device get its MAC address from? | |
2458 | ||
9a6c6867 JV |
2459 | When using slave devices that have fixed MAC addresses, or when |
2460 | the fail_over_mac option is enabled, the bonding device's MAC address is | |
2461 | the MAC address of the active slave. | |
2462 | ||
2463 | For other configurations, if not explicitly configured (with | |
2464 | ifconfig or ip link), the MAC address of the bonding device is taken from | |
2465 | its first slave device. This MAC address is then passed to all following | |
2466 | slaves and remains persistent (even if the first slave is removed) until | |
2467 | the bonding device is brought down or reconfigured. | |
1da177e4 LT |
2468 | |
2469 | If you wish to change the MAC address, you can set it with | |
00354cfb | 2470 | ifconfig or ip link: |
1da177e4 LT |
2471 | |
2472 | # ifconfig bond0 hw ether 00:11:22:33:44:55 | |
2473 | ||
00354cfb JV |
2474 | # ip link set bond0 address 66:77:88:99:aa:bb |
2475 | ||
1da177e4 LT |
2476 | The MAC address can be also changed by bringing down/up the |
2477 | device and then changing its slaves (or their order): | |
2478 | ||
2479 | # ifconfig bond0 down ; modprobe -r bonding | |
2480 | # ifconfig bond0 .... up | |
2481 | # ifenslave bond0 eth... | |
2482 | ||
2483 | This method will automatically take the address from the next | |
2484 | slave that is added. | |
2485 | ||
2486 | To restore your slaves' MAC addresses, you need to detach them | |
2487 | from the bond (`ifenslave -d bond0 eth0'). The bonding driver will | |
2488 | then restore the MAC addresses that the slaves had before they were | |
2489 | enslaved. | |
2490 | ||
00354cfb | 2491 | 16. Resources and Links |
1da177e4 LT |
2492 | ======================= |
2493 | ||
2494 | The latest version of the bonding driver can be found in the latest | |
2495 | version of the linux kernel, found on http://kernel.org | |
2496 | ||
00354cfb JV |
2497 | The latest version of this document can be found in either the latest |
2498 | kernel source (named Documentation/networking/bonding.txt), or on the | |
2499 | bonding sourceforge site: | |
2500 | ||
2501 | http://www.sourceforge.net/projects/bonding | |
2502 | ||
1da177e4 LT |
2503 | Discussions regarding the bonding driver take place primarily on the |
2504 | bonding-devel mailing list, hosted at sourceforge.net. If you have | |
00354cfb | 2505 | questions or problems, post them to the list. The list address is: |
1da177e4 LT |
2506 | |
2507 | bonding-devel@lists.sourceforge.net | |
2508 | ||
00354cfb JV |
2509 | The administrative interface (to subscribe or unsubscribe) can |
2510 | be found at: | |
1da177e4 | 2511 | |
00354cfb | 2512 | https://lists.sourceforge.net/lists/listinfo/bonding-devel |
1da177e4 LT |
2513 | |
2514 | Donald Becker's Ethernet Drivers and diag programs may be found at : | |
0ea6e611 | 2515 | - http://web.archive.org/web/*/http://www.scyld.com/network/ |
1da177e4 LT |
2516 | |
2517 | You will also find a lot of information regarding Ethernet, NWay, MII, | |
2518 | etc. at www.scyld.com. | |
2519 | ||
2520 | -- END -- |