bonding.txt 112 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785278627872788278927902791279227932794279527962797279827992800280128022803280428052806280728082809281028112812281328142815281628172818281928202821282228232824282528262827
  1. Linux Ethernet Bonding Driver HOWTO
  2. Latest update: 27 April 2011
  3. Initial release : Thomas Davis <tadavis at lbl.gov>
  4. Corrections, HA extensions : 2000/10/03-15 :
  5. - Willy Tarreau <willy at meta-x.org>
  6. - Constantine Gavrilov <const-g at xpert.com>
  7. - Chad N. Tindel <ctindel at ieee dot org>
  8. - Janice Girouard <girouard at us dot ibm dot com>
  9. - Jay Vosburgh <fubar at us dot ibm dot com>
  10. Reorganized and updated Feb 2005 by Jay Vosburgh
  11. Added Sysfs information: 2006/04/24
  12. - Mitch Williams <mitch.a.williams at intel.com>
  13. Introduction
  14. ============
  15. The Linux bonding driver provides a method for aggregating
  16. multiple network interfaces into a single logical "bonded" interface.
  17. The behavior of the bonded interfaces depends upon the mode; generally
  18. speaking, modes provide either hot standby or load balancing services.
  19. Additionally, link integrity monitoring may be performed.
  20. The bonding driver originally came from Donald Becker's
  21. beowulf patches for kernel 2.0. It has changed quite a bit since, and
  22. the original tools from extreme-linux and beowulf sites will not work
  23. with this version of the driver.
  24. For new versions of the driver, updated userspace tools, and
  25. who to ask for help, please follow the links at the end of this file.
  26. Table of Contents
  27. =================
  28. 1. Bonding Driver Installation
  29. 2. Bonding Driver Options
  30. 3. Configuring Bonding Devices
  31. 3.1 Configuration with Sysconfig Support
  32. 3.1.1 Using DHCP with Sysconfig
  33. 3.1.2 Configuring Multiple Bonds with Sysconfig
  34. 3.2 Configuration with Initscripts Support
  35. 3.2.1 Using DHCP with Initscripts
  36. 3.2.2 Configuring Multiple Bonds with Initscripts
  37. 3.3 Configuring Bonding Manually with Ifenslave
  38. 3.3.1 Configuring Multiple Bonds Manually
  39. 3.4 Configuring Bonding Manually via Sysfs
  40. 3.5 Configuration with Interfaces Support
  41. 3.6 Overriding Configuration for Special Cases
  42. 3.7 Configuring LACP for 802.3ad mode in a more secure way
  43. 4. Querying Bonding Configuration
  44. 4.1 Bonding Configuration
  45. 4.2 Network Configuration
  46. 5. Switch Configuration
  47. 6. 802.1q VLAN Support
  48. 7. Link Monitoring
  49. 7.1 ARP Monitor Operation
  50. 7.2 Configuring Multiple ARP Targets
  51. 7.3 MII Monitor Operation
  52. 8. Potential Trouble Sources
  53. 8.1 Adventures in Routing
  54. 8.2 Ethernet Device Renaming
  55. 8.3 Painfully Slow Or No Failed Link Detection By Miimon
  56. 9. SNMP agents
  57. 10. Promiscuous mode
  58. 11. Configuring Bonding for High Availability
  59. 11.1 High Availability in a Single Switch Topology
  60. 11.2 High Availability in a Multiple Switch Topology
  61. 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
  62. 11.2.2 HA Link Monitoring for Multiple Switch Topology
  63. 12. Configuring Bonding for Maximum Throughput
  64. 12.1 Maximum Throughput in a Single Switch Topology
  65. 12.1.1 MT Bonding Mode Selection for Single Switch Topology
  66. 12.1.2 MT Link Monitoring for Single Switch Topology
  67. 12.2 Maximum Throughput in a Multiple Switch Topology
  68. 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
  69. 12.2.2 MT Link Monitoring for Multiple Switch Topology
  70. 13. Switch Behavior Issues
  71. 13.1 Link Establishment and Failover Delays
  72. 13.2 Duplicated Incoming Packets
  73. 14. Hardware Specific Considerations
  74. 14.1 IBM BladeCenter
  75. 15. Frequently Asked Questions
  76. 16. Resources and Links
  77. 1. Bonding Driver Installation
  78. ==============================
  79. Most popular distro kernels ship with the bonding driver
  80. already available as a module. If your distro does not, or you
  81. have need to compile bonding from source (e.g., configuring and
  82. installing a mainline kernel from kernel.org), you'll need to perform
  83. the following steps:
  84. 1.1 Configure and build the kernel with bonding
  85. -----------------------------------------------
  86. The current version of the bonding driver is available in the
  87. drivers/net/bonding subdirectory of the most recent kernel source
  88. (which is available on http://kernel.org). Most users "rolling their
  89. own" will want to use the most recent kernel from kernel.org.
  90. Configure kernel with "make menuconfig" (or "make xconfig" or
  91. "make config"), then select "Bonding driver support" in the "Network
  92. device support" section. It is recommended that you configure the
  93. driver as module since it is currently the only way to pass parameters
  94. to the driver or configure more than one bonding device.
  95. Build and install the new kernel and modules.
  96. 1.2 Bonding Control Utility
  97. -------------------------------------
  98. It is recommended to configure bonding via iproute2 (netlink)
  99. or sysfs, the old ifenslave control utility is obsolete.
  100. 2. Bonding Driver Options
  101. =========================
  102. Options for the bonding driver are supplied as parameters to the
  103. bonding module at load time, or are specified via sysfs.
  104. Module options may be given as command line arguments to the
  105. insmod or modprobe command, but are usually specified in either the
  106. /etc/modrobe.d/*.conf configuration files, or in a distro-specific
  107. configuration file (some of which are detailed in the next section).
  108. Details on bonding support for sysfs is provided in the
  109. "Configuring Bonding Manually via Sysfs" section, below.
  110. The available bonding driver parameters are listed below. If a
  111. parameter is not specified the default value is used. When initially
  112. configuring a bond, it is recommended "tail -f /var/log/messages" be
  113. run in a separate window to watch for bonding driver error messages.
  114. It is critical that either the miimon or arp_interval and
  115. arp_ip_target parameters be specified, otherwise serious network
  116. degradation will occur during link failures. Very few devices do not
  117. support at least miimon, so there is really no reason not to use it.
  118. Options with textual values will accept either the text name
  119. or, for backwards compatibility, the option value. E.g.,
  120. "mode=802.3ad" and "mode=4" set the same mode.
  121. The parameters are as follows:
  122. active_slave
  123. Specifies the new active slave for modes that support it
  124. (active-backup, balance-alb and balance-tlb). Possible values
  125. are the name of any currently enslaved interface, or an empty
  126. string. If a name is given, the slave and its link must be up in order
  127. to be selected as the new active slave. If an empty string is
  128. specified, the current active slave is cleared, and a new active
  129. slave is selected automatically.
  130. Note that this is only available through the sysfs interface. No module
  131. parameter by this name exists.
  132. The normal value of this option is the name of the currently
  133. active slave, or the empty string if there is no active slave or
  134. the current mode does not use an active slave.
  135. ad_actor_sys_prio
  136. In an AD system, this specifies the system priority. The allowed range
  137. is 1 - 65535. If the value is not specified, it takes 65535 as the
  138. default value.
  139. This parameter has effect only in 802.3ad mode and is available through
  140. SysFs interface.
  141. ad_actor_system
  142. In an AD system, this specifies the mac-address for the actor in
  143. protocol packet exchanges (LACPDUs). The value cannot be NULL or
  144. multicast. It is preferred to have the local-admin bit set for this
  145. mac but driver does not enforce it. If the value is not given then
  146. system defaults to using the masters' mac address as actors' system
  147. address.
  148. This parameter has effect only in 802.3ad mode and is available through
  149. SysFs interface.
  150. ad_select
  151. Specifies the 802.3ad aggregation selection logic to use. The
  152. possible values and their effects are:
  153. stable or 0
  154. The active aggregator is chosen by largest aggregate
  155. bandwidth.
  156. Reselection of the active aggregator occurs only when all
  157. slaves of the active aggregator are down or the active
  158. aggregator has no slaves.
  159. This is the default value.
  160. bandwidth or 1
  161. The active aggregator is chosen by largest aggregate
  162. bandwidth. Reselection occurs if:
  163. - A slave is added to or removed from the bond
  164. - Any slave's link state changes
  165. - Any slave's 802.3ad association state changes
  166. - The bond's administrative state changes to up
  167. count or 2
  168. The active aggregator is chosen by the largest number of
  169. ports (slaves). Reselection occurs as described under the
  170. "bandwidth" setting, above.
  171. The bandwidth and count selection policies permit failover of
  172. 802.3ad aggregations when partial failure of the active aggregator
  173. occurs. This keeps the aggregator with the highest availability
  174. (either in bandwidth or in number of ports) active at all times.
  175. This option was added in bonding version 3.4.0.
  176. ad_user_port_key
  177. In an AD system, the port-key has three parts as shown below -
  178. Bits Use
  179. 00 Duplex
  180. 01-05 Speed
  181. 06-15 User-defined
  182. This defines the upper 10 bits of the port key. The values can be
  183. from 0 - 1023. If not given, the system defaults to 0.
  184. This parameter has effect only in 802.3ad mode and is available through
  185. SysFs interface.
  186. all_slaves_active
  187. Specifies that duplicate frames (received on inactive ports) should be
  188. dropped (0) or delivered (1).
  189. Normally, bonding will drop duplicate frames (received on inactive
  190. ports), which is desirable for most users. But there are some times
  191. it is nice to allow duplicate frames to be delivered.
  192. The default value is 0 (drop duplicate frames received on inactive
  193. ports).
  194. arp_interval
  195. Specifies the ARP link monitoring frequency in milliseconds.
  196. The ARP monitor works by periodically checking the slave
  197. devices to determine whether they have sent or received
  198. traffic recently (the precise criteria depends upon the
  199. bonding mode, and the state of the slave). Regular traffic is
  200. generated via ARP probes issued for the addresses specified by
  201. the arp_ip_target option.
  202. This behavior can be modified by the arp_validate option,
  203. below.
  204. If ARP monitoring is used in an etherchannel compatible mode
  205. (modes 0 and 2), the switch should be configured in a mode
  206. that evenly distributes packets across all links. If the
  207. switch is configured to distribute the packets in an XOR
  208. fashion, all replies from the ARP targets will be received on
  209. the same link which could cause the other team members to
  210. fail. ARP monitoring should not be used in conjunction with
  211. miimon. A value of 0 disables ARP monitoring. The default
  212. value is 0.
  213. arp_ip_target
  214. Specifies the IP addresses to use as ARP monitoring peers when
  215. arp_interval is > 0. These are the targets of the ARP request
  216. sent to determine the health of the link to the targets.
  217. Specify these values in ddd.ddd.ddd.ddd format. Multiple IP
  218. addresses must be separated by a comma. At least one IP
  219. address must be given for ARP monitoring to function. The
  220. maximum number of targets that can be specified is 16. The
  221. default value is no IP addresses.
  222. arp_validate
  223. Specifies whether or not ARP probes and replies should be
  224. validated in any mode that supports arp monitoring, or whether
  225. non-ARP traffic should be filtered (disregarded) for link
  226. monitoring purposes.
  227. Possible values are:
  228. none or 0
  229. No validation or filtering is performed.
  230. active or 1
  231. Validation is performed only for the active slave.
  232. backup or 2
  233. Validation is performed only for backup slaves.
  234. all or 3
  235. Validation is performed for all slaves.
  236. filter or 4
  237. Filtering is applied to all slaves. No validation is
  238. performed.
  239. filter_active or 5
  240. Filtering is applied to all slaves, validation is performed
  241. only for the active slave.
  242. filter_backup or 6
  243. Filtering is applied to all slaves, validation is performed
  244. only for backup slaves.
  245. Validation:
  246. Enabling validation causes the ARP monitor to examine the incoming
  247. ARP requests and replies, and only consider a slave to be up if it
  248. is receiving the appropriate ARP traffic.
  249. For an active slave, the validation checks ARP replies to confirm
  250. that they were generated by an arp_ip_target. Since backup slaves
  251. do not typically receive these replies, the validation performed
  252. for backup slaves is on the broadcast ARP request sent out via the
  253. active slave. It is possible that some switch or network
  254. configurations may result in situations wherein the backup slaves
  255. do not receive the ARP requests; in such a situation, validation
  256. of backup slaves must be disabled.
  257. The validation of ARP requests on backup slaves is mainly helping
  258. bonding to decide which slaves are more likely to work in case of
  259. the active slave failure, it doesn't really guarantee that the
  260. backup slave will work if it's selected as the next active slave.
  261. Validation is useful in network configurations in which multiple
  262. bonding hosts are concurrently issuing ARPs to one or more targets
  263. beyond a common switch. Should the link between the switch and
  264. target fail (but not the switch itself), the probe traffic
  265. generated by the multiple bonding instances will fool the standard
  266. ARP monitor into considering the links as still up. Use of
  267. validation can resolve this, as the ARP monitor will only consider
  268. ARP requests and replies associated with its own instance of
  269. bonding.
  270. Filtering:
  271. Enabling filtering causes the ARP monitor to only use incoming ARP
  272. packets for link availability purposes. Arriving packets that are
  273. not ARPs are delivered normally, but do not count when determining
  274. if a slave is available.
  275. Filtering operates by only considering the reception of ARP
  276. packets (any ARP packet, regardless of source or destination) when
  277. determining if a slave has received traffic for link availability
  278. purposes.
  279. Filtering is useful in network configurations in which significant
  280. levels of third party broadcast traffic would fool the standard
  281. ARP monitor into considering the links as still up. Use of
  282. filtering can resolve this, as only ARP traffic is considered for
  283. link availability purposes.
  284. This option was added in bonding version 3.1.0.
  285. arp_all_targets
  286. Specifies the quantity of arp_ip_targets that must be reachable
  287. in order for the ARP monitor to consider a slave as being up.
  288. This option affects only active-backup mode for slaves with
  289. arp_validation enabled.
  290. Possible values are:
  291. any or 0
  292. consider the slave up only when any of the arp_ip_targets
  293. is reachable
  294. all or 1
  295. consider the slave up only when all of the arp_ip_targets
  296. are reachable
  297. downdelay
  298. Specifies the time, in milliseconds, to wait before disabling
  299. a slave after a link failure has been detected. This option
  300. is only valid for the miimon link monitor. The downdelay
  301. value should be a multiple of the miimon value; if not, it
  302. will be rounded down to the nearest multiple. The default
  303. value is 0.
  304. fail_over_mac
  305. Specifies whether active-backup mode should set all slaves to
  306. the same MAC address at enslavement (the traditional
  307. behavior), or, when enabled, perform special handling of the
  308. bond's MAC address in accordance with the selected policy.
  309. Possible values are:
  310. none or 0
  311. This setting disables fail_over_mac, and causes
  312. bonding to set all slaves of an active-backup bond to
  313. the same MAC address at enslavement time. This is the
  314. default.
  315. active or 1
  316. The "active" fail_over_mac policy indicates that the
  317. MAC address of the bond should always be the MAC
  318. address of the currently active slave. The MAC
  319. address of the slaves is not changed; instead, the MAC
  320. address of the bond changes during a failover.
  321. This policy is useful for devices that cannot ever
  322. alter their MAC address, or for devices that refuse
  323. incoming broadcasts with their own source MAC (which
  324. interferes with the ARP monitor).
  325. The down side of this policy is that every device on
  326. the network must be updated via gratuitous ARP,
  327. vs. just updating a switch or set of switches (which
  328. often takes place for any traffic, not just ARP
  329. traffic, if the switch snoops incoming traffic to
  330. update its tables) for the traditional method. If the
  331. gratuitous ARP is lost, communication may be
  332. disrupted.
  333. When this policy is used in conjunction with the mii
  334. monitor, devices which assert link up prior to being
  335. able to actually transmit and receive are particularly
  336. susceptible to loss of the gratuitous ARP, and an
  337. appropriate updelay setting may be required.
  338. follow or 2
  339. The "follow" fail_over_mac policy causes the MAC
  340. address of the bond to be selected normally (normally
  341. the MAC address of the first slave added to the bond).
  342. However, the second and subsequent slaves are not set
  343. to this MAC address while they are in a backup role; a
  344. slave is programmed with the bond's MAC address at
  345. failover time (and the formerly active slave receives
  346. the newly active slave's MAC address).
  347. This policy is useful for multiport devices that
  348. either become confused or incur a performance penalty
  349. when multiple ports are programmed with the same MAC
  350. address.
  351. The default policy is none, unless the first slave cannot
  352. change its MAC address, in which case the active policy is
  353. selected by default.
  354. This option may be modified via sysfs only when no slaves are
  355. present in the bond.
  356. This option was added in bonding version 3.2.0. The "follow"
  357. policy was added in bonding version 3.3.0.
  358. lacp_rate
  359. Option specifying the rate in which we'll ask our link partner
  360. to transmit LACPDU packets in 802.3ad mode. Possible values
  361. are:
  362. slow or 0
  363. Request partner to transmit LACPDUs every 30 seconds
  364. fast or 1
  365. Request partner to transmit LACPDUs every 1 second
  366. The default is slow.
  367. max_bonds
  368. Specifies the number of bonding devices to create for this
  369. instance of the bonding driver. E.g., if max_bonds is 3, and
  370. the bonding driver is not already loaded, then bond0, bond1
  371. and bond2 will be created. The default value is 1. Specifying
  372. a value of 0 will load bonding, but will not create any devices.
  373. miimon
  374. Specifies the MII link monitoring frequency in milliseconds.
  375. This determines how often the link state of each slave is
  376. inspected for link failures. A value of zero disables MII
  377. link monitoring. A value of 100 is a good starting point.
  378. The use_carrier option, below, affects how the link state is
  379. determined. See the High Availability section for additional
  380. information. The default value is 0.
  381. min_links
  382. Specifies the minimum number of links that must be active before
  383. asserting carrier. It is similar to the Cisco EtherChannel min-links
  384. feature. This allows setting the minimum number of member ports that
  385. must be up (link-up state) before marking the bond device as up
  386. (carrier on). This is useful for situations where higher level services
  387. such as clustering want to ensure a minimum number of low bandwidth
  388. links are active before switchover. This option only affect 802.3ad
  389. mode.
  390. The default value is 0. This will cause carrier to be asserted (for
  391. 802.3ad mode) whenever there is an active aggregator, regardless of the
  392. number of available links in that aggregator. Note that, because an
  393. aggregator cannot be active without at least one available link,
  394. setting this option to 0 or to 1 has the exact same effect.
  395. mode
  396. Specifies one of the bonding policies. The default is
  397. balance-rr (round robin). Possible values are:
  398. balance-rr or 0
  399. Round-robin policy: Transmit packets in sequential
  400. order from the first available slave through the
  401. last. This mode provides load balancing and fault
  402. tolerance.
  403. active-backup or 1
  404. Active-backup policy: Only one slave in the bond is
  405. active. A different slave becomes active if, and only
  406. if, the active slave fails. The bond's MAC address is
  407. externally visible on only one port (network adapter)
  408. to avoid confusing the switch.
  409. In bonding version 2.6.2 or later, when a failover
  410. occurs in active-backup mode, bonding will issue one
  411. or more gratuitous ARPs on the newly active slave.
  412. One gratuitous ARP is issued for the bonding master
  413. interface and each VLAN interfaces configured above
  414. it, provided that the interface has at least one IP
  415. address configured. Gratuitous ARPs issued for VLAN
  416. interfaces are tagged with the appropriate VLAN id.
  417. This mode provides fault tolerance. The primary
  418. option, documented below, affects the behavior of this
  419. mode.
  420. balance-xor or 2
  421. XOR policy: Transmit based on the selected transmit
  422. hash policy. The default policy is a simple [(source
  423. MAC address XOR'd with destination MAC address XOR
  424. packet type ID) modulo slave count]. Alternate transmit
  425. policies may be selected via the xmit_hash_policy option,
  426. described below.
  427. This mode provides load balancing and fault tolerance.
  428. broadcast or 3
  429. Broadcast policy: transmits everything on all slave
  430. interfaces. This mode provides fault tolerance.
  431. 802.3ad or 4
  432. IEEE 802.3ad Dynamic link aggregation. Creates
  433. aggregation groups that share the same speed and
  434. duplex settings. Utilizes all slaves in the active
  435. aggregator according to the 802.3ad specification.
  436. Slave selection for outgoing traffic is done according
  437. to the transmit hash policy, which may be changed from
  438. the default simple XOR policy via the xmit_hash_policy
  439. option, documented below. Note that not all transmit
  440. policies may be 802.3ad compliant, particularly in
  441. regards to the packet mis-ordering requirements of
  442. section 43.2.4 of the 802.3ad standard. Differing
  443. peer implementations will have varying tolerances for
  444. noncompliance.
  445. Prerequisites:
  446. 1. Ethtool support in the base drivers for retrieving
  447. the speed and duplex of each slave.
  448. 2. A switch that supports IEEE 802.3ad Dynamic link
  449. aggregation.
  450. Most switches will require some type of configuration
  451. to enable 802.3ad mode.
  452. balance-tlb or 5
  453. Adaptive transmit load balancing: channel bonding that
  454. does not require any special switch support.
  455. In tlb_dynamic_lb=1 mode; the outgoing traffic is
  456. distributed according to the current load (computed
  457. relative to the speed) on each slave.
  458. In tlb_dynamic_lb=0 mode; the load balancing based on
  459. current load is disabled and the load is distributed
  460. only using the hash distribution.
  461. Incoming traffic is received by the current slave.
  462. If the receiving slave fails, another slave takes over
  463. the MAC address of the failed receiving slave.
  464. Prerequisite:
  465. Ethtool support in the base drivers for retrieving the
  466. speed of each slave.
  467. balance-alb or 6
  468. Adaptive load balancing: includes balance-tlb plus
  469. receive load balancing (rlb) for IPV4 traffic, and
  470. does not require any special switch support. The
  471. receive load balancing is achieved by ARP negotiation.
  472. The bonding driver intercepts the ARP Replies sent by
  473. the local system on their way out and overwrites the
  474. source hardware address with the unique hardware
  475. address of one of the slaves in the bond such that
  476. different peers use different hardware addresses for
  477. the server.
  478. Receive traffic from connections created by the server
  479. is also balanced. When the local system sends an ARP
  480. Request the bonding driver copies and saves the peer's
  481. IP information from the ARP packet. When the ARP
  482. Reply arrives from the peer, its hardware address is
  483. retrieved and the bonding driver initiates an ARP
  484. reply to this peer assigning it to one of the slaves
  485. in the bond. A problematic outcome of using ARP
  486. negotiation for balancing is that each time that an
  487. ARP request is broadcast it uses the hardware address
  488. of the bond. Hence, peers learn the hardware address
  489. of the bond and the balancing of receive traffic
  490. collapses to the current slave. This is handled by
  491. sending updates (ARP Replies) to all the peers with
  492. their individually assigned hardware address such that
  493. the traffic is redistributed. Receive traffic is also
  494. redistributed when a new slave is added to the bond
  495. and when an inactive slave is re-activated. The
  496. receive load is distributed sequentially (round robin)
  497. among the group of highest speed slaves in the bond.
  498. When a link is reconnected or a new slave joins the
  499. bond the receive traffic is redistributed among all
  500. active slaves in the bond by initiating ARP Replies
  501. with the selected MAC address to each of the
  502. clients. The updelay parameter (detailed below) must
  503. be set to a value equal or greater than the switch's
  504. forwarding delay so that the ARP Replies sent to the
  505. peers will not be blocked by the switch.
  506. Prerequisites:
  507. 1. Ethtool support in the base drivers for retrieving
  508. the speed of each slave.
  509. 2. Base driver support for setting the hardware
  510. address of a device while it is open. This is
  511. required so that there will always be one slave in the
  512. team using the bond hardware address (the
  513. curr_active_slave) while having a unique hardware
  514. address for each slave in the bond. If the
  515. curr_active_slave fails its hardware address is
  516. swapped with the new curr_active_slave that was
  517. chosen.
  518. num_grat_arp
  519. num_unsol_na
  520. Specify the number of peer notifications (gratuitous ARPs and
  521. unsolicited IPv6 Neighbor Advertisements) to be issued after a
  522. failover event. As soon as the link is up on the new slave
  523. (possibly immediately) a peer notification is sent on the
  524. bonding device and each VLAN sub-device. This is repeated at
  525. each link monitor interval (arp_interval or miimon, whichever
  526. is active) if the number is greater than 1.
  527. The valid range is 0 - 255; the default value is 1. These options
  528. affect only the active-backup mode. These options were added for
  529. bonding versions 3.3.0 and 3.4.0 respectively.
  530. From Linux 3.0 and bonding version 3.7.1, these notifications
  531. are generated by the ipv4 and ipv6 code and the numbers of
  532. repetitions cannot be set independently.
  533. packets_per_slave
  534. Specify the number of packets to transmit through a slave before
  535. moving to the next one. When set to 0 then a slave is chosen at
  536. random.
  537. The valid range is 0 - 65535; the default value is 1. This option
  538. has effect only in balance-rr mode.
  539. primary
  540. A string (eth0, eth2, etc) specifying which slave is the
  541. primary device. The specified device will always be the
  542. active slave while it is available. Only when the primary is
  543. off-line will alternate devices be used. This is useful when
  544. one slave is preferred over another, e.g., when one slave has
  545. higher throughput than another.
  546. The primary option is only valid for active-backup(1),
  547. balance-tlb (5) and balance-alb (6) mode.
  548. primary_reselect
  549. Specifies the reselection policy for the primary slave. This
  550. affects how the primary slave is chosen to become the active slave
  551. when failure of the active slave or recovery of the primary slave
  552. occurs. This option is designed to prevent flip-flopping between
  553. the primary slave and other slaves. Possible values are:
  554. always or 0 (default)
  555. The primary slave becomes the active slave whenever it
  556. comes back up.
  557. better or 1
  558. The primary slave becomes the active slave when it comes
  559. back up, if the speed and duplex of the primary slave is
  560. better than the speed and duplex of the current active
  561. slave.
  562. failure or 2
  563. The primary slave becomes the active slave only if the
  564. current active slave fails and the primary slave is up.
  565. The primary_reselect setting is ignored in two cases:
  566. If no slaves are active, the first slave to recover is
  567. made the active slave.
  568. When initially enslaved, the primary slave is always made
  569. the active slave.
  570. Changing the primary_reselect policy via sysfs will cause an
  571. immediate selection of the best active slave according to the new
  572. policy. This may or may not result in a change of the active
  573. slave, depending upon the circumstances.
  574. This option was added for bonding version 3.6.0.
  575. tlb_dynamic_lb
  576. Specifies if dynamic shuffling of flows is enabled in tlb
  577. mode. The value has no effect on any other modes.
  578. The default behavior of tlb mode is to shuffle active flows across
  579. slaves based on the load in that interval. This gives nice lb
  580. characteristics but can cause packet reordering. If re-ordering is
  581. a concern use this variable to disable flow shuffling and rely on
  582. load balancing provided solely by the hash distribution.
  583. xmit-hash-policy can be used to select the appropriate hashing for
  584. the setup.
  585. The sysfs entry can be used to change the setting per bond device
  586. and the initial value is derived from the module parameter. The
  587. sysfs entry is allowed to be changed only if the bond device is
  588. down.
  589. The default value is "1" that enables flow shuffling while value "0"
  590. disables it. This option was added in bonding driver 3.7.1
  591. updelay
  592. Specifies the time, in milliseconds, to wait before enabling a
  593. slave after a link recovery has been detected. This option is
  594. only valid for the miimon link monitor. The updelay value
  595. should be a multiple of the miimon value; if not, it will be
  596. rounded down to the nearest multiple. The default value is 0.
  597. use_carrier
  598. Specifies whether or not miimon should use MII or ETHTOOL
  599. ioctls vs. netif_carrier_ok() to determine the link
  600. status. The MII or ETHTOOL ioctls are less efficient and
  601. utilize a deprecated calling sequence within the kernel. The
  602. netif_carrier_ok() relies on the device driver to maintain its
  603. state with netif_carrier_on/off; at this writing, most, but
  604. not all, device drivers support this facility.
  605. If bonding insists that the link is up when it should not be,
  606. it may be that your network device driver does not support
  607. netif_carrier_on/off. The default state for netif_carrier is
  608. "carrier on," so if a driver does not support netif_carrier,
  609. it will appear as if the link is always up. In this case,
  610. setting use_carrier to 0 will cause bonding to revert to the
  611. MII / ETHTOOL ioctl method to determine the link state.
  612. A value of 1 enables the use of netif_carrier_ok(), a value of
  613. 0 will use the deprecated MII / ETHTOOL ioctls. The default
  614. value is 1.
  615. xmit_hash_policy
  616. Selects the transmit hash policy to use for slave selection in
  617. balance-xor, 802.3ad, and tlb modes. Possible values are:
  618. layer2
  619. Uses XOR of hardware MAC addresses and packet type ID
  620. field to generate the hash. The formula is
  621. hash = source MAC XOR destination MAC XOR packet type ID
  622. slave number = hash modulo slave count
  623. This algorithm will place all traffic to a particular
  624. network peer on the same slave.
  625. This algorithm is 802.3ad compliant.
  626. layer2+3
  627. This policy uses a combination of layer2 and layer3
  628. protocol information to generate the hash.
  629. Uses XOR of hardware MAC addresses and IP addresses to
  630. generate the hash. The formula is
  631. hash = source MAC XOR destination MAC XOR packet type ID
  632. hash = hash XOR source IP XOR destination IP
  633. hash = hash XOR (hash RSHIFT 16)
  634. hash = hash XOR (hash RSHIFT 8)
  635. And then hash is reduced modulo slave count.
  636. If the protocol is IPv6 then the source and destination
  637. addresses are first hashed using ipv6_addr_hash.
  638. This algorithm will place all traffic to a particular
  639. network peer on the same slave. For non-IP traffic,
  640. the formula is the same as for the layer2 transmit
  641. hash policy.
  642. This policy is intended to provide a more balanced
  643. distribution of traffic than layer2 alone, especially
  644. in environments where a layer3 gateway device is
  645. required to reach most destinations.
  646. This algorithm is 802.3ad compliant.
  647. layer3+4
  648. This policy uses upper layer protocol information,
  649. when available, to generate the hash. This allows for
  650. traffic to a particular network peer to span multiple
  651. slaves, although a single connection will not span
  652. multiple slaves.
  653. The formula for unfragmented TCP and UDP packets is
  654. hash = source port, destination port (as in the header)
  655. hash = hash XOR source IP XOR destination IP
  656. hash = hash XOR (hash RSHIFT 16)
  657. hash = hash XOR (hash RSHIFT 8)
  658. And then hash is reduced modulo slave count.
  659. If the protocol is IPv6 then the source and destination
  660. addresses are first hashed using ipv6_addr_hash.
  661. For fragmented TCP or UDP packets and all other IPv4 and
  662. IPv6 protocol traffic, the source and destination port
  663. information is omitted. For non-IP traffic, the
  664. formula is the same as for the layer2 transmit hash
  665. policy.
  666. This algorithm is not fully 802.3ad compliant. A
  667. single TCP or UDP conversation containing both
  668. fragmented and unfragmented packets will see packets
  669. striped across two interfaces. This may result in out
  670. of order delivery. Most traffic types will not meet
  671. this criteria, as TCP rarely fragments traffic, and
  672. most UDP traffic is not involved in extended
  673. conversations. Other implementations of 802.3ad may
  674. or may not tolerate this noncompliance.
  675. encap2+3
  676. This policy uses the same formula as layer2+3 but it
  677. relies on skb_flow_dissect to obtain the header fields
  678. which might result in the use of inner headers if an
  679. encapsulation protocol is used. For example this will
  680. improve the performance for tunnel users because the
  681. packets will be distributed according to the encapsulated
  682. flows.
  683. encap3+4
  684. This policy uses the same formula as layer3+4 but it
  685. relies on skb_flow_dissect to obtain the header fields
  686. which might result in the use of inner headers if an
  687. encapsulation protocol is used. For example this will
  688. improve the performance for tunnel users because the
  689. packets will be distributed according to the encapsulated
  690. flows.
  691. The default value is layer2. This option was added in bonding
  692. version 2.6.3. In earlier versions of bonding, this parameter
  693. does not exist, and the layer2 policy is the only policy. The
  694. layer2+3 value was added for bonding version 3.2.2.
  695. resend_igmp
  696. Specifies the number of IGMP membership reports to be issued after
  697. a failover event. One membership report is issued immediately after
  698. the failover, subsequent packets are sent in each 200ms interval.
  699. The valid range is 0 - 255; the default value is 1. A value of 0
  700. prevents the IGMP membership report from being issued in response
  701. to the failover event.
  702. This option is useful for bonding modes balance-rr (0), active-backup
  703. (1), balance-tlb (5) and balance-alb (6), in which a failover can
  704. switch the IGMP traffic from one slave to another. Therefore a fresh
  705. IGMP report must be issued to cause the switch to forward the incoming
  706. IGMP traffic over the newly selected slave.
  707. This option was added for bonding version 3.7.0.
  708. lp_interval
  709. Specifies the number of seconds between instances where the bonding
  710. driver sends learning packets to each slaves peer switch.
  711. The valid range is 1 - 0x7fffffff; the default value is 1. This Option
  712. has effect only in balance-tlb and balance-alb modes.
  713. 3. Configuring Bonding Devices
  714. ==============================
  715. You can configure bonding using either your distro's network
  716. initialization scripts, or manually using either iproute2 or the
  717. sysfs interface. Distros generally use one of three packages for the
  718. network initialization scripts: initscripts, sysconfig or interfaces.
  719. Recent versions of these packages have support for bonding, while older
  720. versions do not.
  721. We will first describe the options for configuring bonding for
  722. distros using versions of initscripts, sysconfig and interfaces with full
  723. or partial support for bonding, then provide information on enabling
  724. bonding without support from the network initialization scripts (i.e.,
  725. older versions of initscripts or sysconfig).
  726. If you're unsure whether your distro uses sysconfig,
  727. initscripts or interfaces, or don't know if it's new enough, have no fear.
  728. Determining this is fairly straightforward.
  729. First, look for a file called interfaces in /etc/network directory.
  730. If this file is present in your system, then your system use interfaces. See
  731. Configuration with Interfaces Support.
  732. Else, issue the command:
  733. $ rpm -qf /sbin/ifup
  734. It will respond with a line of text starting with either
  735. "initscripts" or "sysconfig," followed by some numbers. This is the
  736. package that provides your network initialization scripts.
  737. Next, to determine if your installation supports bonding,
  738. issue the command:
  739. $ grep ifenslave /sbin/ifup
  740. If this returns any matches, then your initscripts or
  741. sysconfig has support for bonding.
  742. 3.1 Configuration with Sysconfig Support
  743. ----------------------------------------
  744. This section applies to distros using a version of sysconfig
  745. with bonding support, for example, SuSE Linux Enterprise Server 9.
  746. SuSE SLES 9's networking configuration system does support
  747. bonding, however, at this writing, the YaST system configuration
  748. front end does not provide any means to work with bonding devices.
  749. Bonding devices can be managed by hand, however, as follows.
  750. First, if they have not already been configured, configure the
  751. slave devices. On SLES 9, this is most easily done by running the
  752. yast2 sysconfig configuration utility. The goal is for to create an
  753. ifcfg-id file for each slave device. The simplest way to accomplish
  754. this is to configure the devices for DHCP (this is only to get the
  755. file ifcfg-id file created; see below for some issues with DHCP). The
  756. name of the configuration file for each device will be of the form:
  757. ifcfg-id-xx:xx:xx:xx:xx:xx
  758. Where the "xx" portion will be replaced with the digits from
  759. the device's permanent MAC address.
  760. Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
  761. created, it is necessary to edit the configuration files for the slave
  762. devices (the MAC addresses correspond to those of the slave devices).
  763. Before editing, the file will contain multiple lines, and will look
  764. something like this:
  765. BOOTPROTO='dhcp'
  766. STARTMODE='on'
  767. USERCTL='no'
  768. UNIQUE='XNzu.WeZGOGF+4wE'
  769. _nm_name='bus-pci-0001:61:01.0'
  770. Change the BOOTPROTO and STARTMODE lines to the following:
  771. BOOTPROTO='none'
  772. STARTMODE='off'
  773. Do not alter the UNIQUE or _nm_name lines. Remove any other
  774. lines (USERCTL, etc).
  775. Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
  776. it's time to create the configuration file for the bonding device
  777. itself. This file is named ifcfg-bondX, where X is the number of the
  778. bonding device to create, starting at 0. The first such file is
  779. ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig
  780. network configuration system will correctly start multiple instances
  781. of bonding.
  782. The contents of the ifcfg-bondX file is as follows:
  783. BOOTPROTO="static"
  784. BROADCAST="10.0.2.255"
  785. IPADDR="10.0.2.10"
  786. NETMASK="255.255.0.0"
  787. NETWORK="10.0.2.0"
  788. REMOTE_IPADDR=""
  789. STARTMODE="onboot"
  790. BONDING_MASTER="yes"
  791. BONDING_MODULE_OPTS="mode=active-backup miimon=100"
  792. BONDING_SLAVE0="eth0"
  793. BONDING_SLAVE1="bus-pci-0000:06:08.1"
  794. Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
  795. values with the appropriate values for your network.
  796. The STARTMODE specifies when the device is brought online.
  797. The possible values are:
  798. onboot: The device is started at boot time. If you're not
  799. sure, this is probably what you want.
  800. manual: The device is started only when ifup is called
  801. manually. Bonding devices may be configured this
  802. way if you do not wish them to start automatically
  803. at boot for some reason.
  804. hotplug: The device is started by a hotplug event. This is not
  805. a valid choice for a bonding device.
  806. off or ignore: The device configuration is ignored.
  807. The line BONDING_MASTER='yes' indicates that the device is a
  808. bonding master device. The only useful value is "yes."
  809. The contents of BONDING_MODULE_OPTS are supplied to the
  810. instance of the bonding module for this device. Specify the options
  811. for the bonding mode, link monitoring, and so on here. Do not include
  812. the max_bonds bonding parameter; this will confuse the configuration
  813. system if you have multiple bonding devices.
  814. Finally, supply one BONDING_SLAVEn="slave device" for each
  815. slave. where "n" is an increasing value, one for each slave. The
  816. "slave device" is either an interface name, e.g., "eth0", or a device
  817. specifier for the network device. The interface name is easier to
  818. find, but the ethN names are subject to change at boot time if, e.g.,
  819. a device early in the sequence has failed. The device specifiers
  820. (bus-pci-0000:06:08.1 in the example above) specify the physical
  821. network device, and will not change unless the device's bus location
  822. changes (for example, it is moved from one PCI slot to another). The
  823. example above uses one of each type for demonstration purposes; most
  824. configurations will choose one or the other for all slave devices.
  825. When all configuration files have been modified or created,
  826. networking must be restarted for the configuration changes to take
  827. effect. This can be accomplished via the following:
  828. # /etc/init.d/network restart
  829. Note that the network control script (/sbin/ifdown) will
  830. remove the bonding module as part of the network shutdown processing,
  831. so it is not necessary to remove the module by hand if, e.g., the
  832. module parameters have changed.
  833. Also, at this writing, YaST/YaST2 will not manage bonding
  834. devices (they do not show bonding interfaces on its list of network
  835. devices). It is necessary to edit the configuration file by hand to
  836. change the bonding configuration.
  837. Additional general options and details of the ifcfg file
  838. format can be found in an example ifcfg template file:
  839. /etc/sysconfig/network/ifcfg.template
  840. Note that the template does not document the various BONDING_
  841. settings described above, but does describe many of the other options.
  842. 3.1.1 Using DHCP with Sysconfig
  843. -------------------------------
  844. Under sysconfig, configuring a device with BOOTPROTO='dhcp'
  845. will cause it to query DHCP for its IP address information. At this
  846. writing, this does not function for bonding devices; the scripts
  847. attempt to obtain the device address from DHCP prior to adding any of
  848. the slave devices. Without active slaves, the DHCP requests are not
  849. sent to the network.
  850. 3.1.2 Configuring Multiple Bonds with Sysconfig
  851. -----------------------------------------------
  852. The sysconfig network initialization system is capable of
  853. handling multiple bonding devices. All that is necessary is for each
  854. bonding instance to have an appropriately configured ifcfg-bondX file
  855. (as described above). Do not specify the "max_bonds" parameter to any
  856. instance of bonding, as this will confuse sysconfig. If you require
  857. multiple bonding devices with identical parameters, create multiple
  858. ifcfg-bondX files.
  859. Because the sysconfig scripts supply the bonding module
  860. options in the ifcfg-bondX file, it is not necessary to add them to
  861. the system /etc/modules.d/*.conf configuration files.
  862. 3.2 Configuration with Initscripts Support
  863. ------------------------------------------
  864. This section applies to distros using a recent version of
  865. initscripts with bonding support, for example, Red Hat Enterprise Linux
  866. version 3 or later, Fedora, etc. On these systems, the network
  867. initialization scripts have knowledge of bonding, and can be configured to
  868. control bonding devices. Note that older versions of the initscripts
  869. package have lower levels of support for bonding; this will be noted where
  870. applicable.
  871. These distros will not automatically load the network adapter
  872. driver unless the ethX device is configured with an IP address.
  873. Because of this constraint, users must manually configure a
  874. network-script file for all physical adapters that will be members of
  875. a bondX link. Network script files are located in the directory:
  876. /etc/sysconfig/network-scripts
  877. The file name must be prefixed with "ifcfg-eth" and suffixed
  878. with the adapter's physical adapter number. For example, the script
  879. for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
  880. Place the following text in the file:
  881. DEVICE=eth0
  882. USERCTL=no
  883. ONBOOT=yes
  884. MASTER=bond0
  885. SLAVE=yes
  886. BOOTPROTO=none
  887. The DEVICE= line will be different for every ethX device and
  888. must correspond with the name of the file, i.e., ifcfg-eth1 must have
  889. a device line of DEVICE=eth1. The setting of the MASTER= line will
  890. also depend on the final bonding interface name chosen for your bond.
  891. As with other network devices, these typically start at 0, and go up
  892. one for each device, i.e., the first bonding instance is bond0, the
  893. second is bond1, and so on.
  894. Next, create a bond network script. The file name for this
  895. script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
  896. the number of the bond. For bond0 the file is named "ifcfg-bond0",
  897. for bond1 it is named "ifcfg-bond1", and so on. Within that file,
  898. place the following text:
  899. DEVICE=bond0
  900. IPADDR=192.168.1.1
  901. NETMASK=255.255.255.0
  902. NETWORK=192.168.1.0
  903. BROADCAST=192.168.1.255
  904. ONBOOT=yes
  905. BOOTPROTO=none
  906. USERCTL=no
  907. Be sure to change the networking specific lines (IPADDR,
  908. NETMASK, NETWORK and BROADCAST) to match your network configuration.
  909. For later versions of initscripts, such as that found with Fedora
  910. 7 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
  911. and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
  912. file, e.g. a line of the format:
  913. BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
  914. will configure the bond with the specified options. The options
  915. specified in BONDING_OPTS are identical to the bonding module parameters
  916. except for the arp_ip_target field when using versions of initscripts older
  917. than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2). When
  918. using older versions each target should be included as a separate option and
  919. should be preceded by a '+' to indicate it should be added to the list of
  920. queried targets, e.g.,
  921. arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
  922. is the proper syntax to specify multiple targets. When specifying
  923. options via BONDING_OPTS, it is not necessary to edit /etc/modprobe.d/*.conf.
  924. For even older versions of initscripts that do not support
  925. BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
  926. your distro) to load the bonding module with your desired options when the
  927. bond0 interface is brought up. The following lines in /etc/modprobe.d/*.conf
  928. will load the bonding module, and select its options:
  929. alias bond0 bonding
  930. options bond0 mode=balance-alb miimon=100
  931. Replace the sample parameters with the appropriate set of
  932. options for your configuration.
  933. Finally run "/etc/rc.d/init.d/network restart" as root. This
  934. will restart the networking subsystem and your bond link should be now
  935. up and running.
  936. 3.2.1 Using DHCP with Initscripts
  937. ---------------------------------
  938. Recent versions of initscripts (the versions supplied with Fedora
  939. Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
  940. work) have support for assigning IP information to bonding devices via
  941. DHCP.
  942. To configure bonding for DHCP, configure it as described
  943. above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
  944. and add a line consisting of "TYPE=Bonding". Note that the TYPE value
  945. is case sensitive.
  946. 3.2.2 Configuring Multiple Bonds with Initscripts
  947. -------------------------------------------------
  948. Initscripts packages that are included with Fedora 7 and Red Hat
  949. Enterprise Linux 5 support multiple bonding interfaces by simply
  950. specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
  951. number of the bond. This support requires sysfs support in the kernel,
  952. and a bonding driver of version 3.0.0 or later. Other configurations may
  953. not support this method for specifying multiple bonding interfaces; for
  954. those instances, see the "Configuring Multiple Bonds Manually" section,
  955. below.
  956. 3.3 Configuring Bonding Manually with iproute2
  957. -----------------------------------------------
  958. This section applies to distros whose network initialization
  959. scripts (the sysconfig or initscripts package) do not have specific
  960. knowledge of bonding. One such distro is SuSE Linux Enterprise Server
  961. version 8.
  962. The general method for these systems is to place the bonding
  963. module parameters into a config file in /etc/modprobe.d/ (as
  964. appropriate for the installed distro), then add modprobe and/or
  965. `ip link` commands to the system's global init script. The name of
  966. the global init script differs; for sysconfig, it is
  967. /etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
  968. For example, if you wanted to make a simple bond of two e100
  969. devices (presumed to be eth0 and eth1), and have it persist across
  970. reboots, edit the appropriate file (/etc/init.d/boot.local or
  971. /etc/rc.d/rc.local), and add the following:
  972. modprobe bonding mode=balance-alb miimon=100
  973. modprobe e100
  974. ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
  975. ip link set eth0 master bond0
  976. ip link set eth1 master bond0
  977. Replace the example bonding module parameters and bond0
  978. network configuration (IP address, netmask, etc) with the appropriate
  979. values for your configuration.
  980. Unfortunately, this method will not provide support for the
  981. ifup and ifdown scripts on the bond devices. To reload the bonding
  982. configuration, it is necessary to run the initialization script, e.g.,
  983. # /etc/init.d/boot.local
  984. or
  985. # /etc/rc.d/rc.local
  986. It may be desirable in such a case to create a separate script
  987. which only initializes the bonding configuration, then call that
  988. separate script from within boot.local. This allows for bonding to be
  989. enabled without re-running the entire global init script.
  990. To shut down the bonding devices, it is necessary to first
  991. mark the bonding device itself as being down, then remove the
  992. appropriate device driver modules. For our example above, you can do
  993. the following:
  994. # ifconfig bond0 down
  995. # rmmod bonding
  996. # rmmod e100
  997. Again, for convenience, it may be desirable to create a script
  998. with these commands.
  999. 3.3.1 Configuring Multiple Bonds Manually
  1000. -----------------------------------------
  1001. This section contains information on configuring multiple
  1002. bonding devices with differing options for those systems whose network
  1003. initialization scripts lack support for configuring multiple bonds.
  1004. If you require multiple bonding devices, but all with the same
  1005. options, you may wish to use the "max_bonds" module parameter,
  1006. documented above.
  1007. To create multiple bonding devices with differing options, it is
  1008. preferable to use bonding parameters exported by sysfs, documented in the
  1009. section below.
  1010. For versions of bonding without sysfs support, the only means to
  1011. provide multiple instances of bonding with differing options is to load
  1012. the bonding driver multiple times. Note that current versions of the
  1013. sysconfig network initialization scripts handle this automatically; if
  1014. your distro uses these scripts, no special action is needed. See the
  1015. section Configuring Bonding Devices, above, if you're not sure about your
  1016. network initialization scripts.
  1017. To load multiple instances of the module, it is necessary to
  1018. specify a different name for each instance (the module loading system
  1019. requires that every loaded module, even multiple instances of the same
  1020. module, have a unique name). This is accomplished by supplying multiple
  1021. sets of bonding options in /etc/modprobe.d/*.conf, for example:
  1022. alias bond0 bonding
  1023. options bond0 -o bond0 mode=balance-rr miimon=100
  1024. alias bond1 bonding
  1025. options bond1 -o bond1 mode=balance-alb miimon=50
  1026. will load the bonding module two times. The first instance is
  1027. named "bond0" and creates the bond0 device in balance-rr mode with an
  1028. miimon of 100. The second instance is named "bond1" and creates the
  1029. bond1 device in balance-alb mode with an miimon of 50.
  1030. In some circumstances (typically with older distributions),
  1031. the above does not work, and the second bonding instance never sees
  1032. its options. In that case, the second options line can be substituted
  1033. as follows:
  1034. install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
  1035. mode=balance-alb miimon=50
  1036. This may be repeated any number of times, specifying a new and
  1037. unique name in place of bond1 for each subsequent instance.
  1038. It has been observed that some Red Hat supplied kernels are unable
  1039. to rename modules at load time (the "-o bond1" part). Attempts to pass
  1040. that option to modprobe will produce an "Operation not permitted" error.
  1041. This has been reported on some Fedora Core kernels, and has been seen on
  1042. RHEL 4 as well. On kernels exhibiting this problem, it will be impossible
  1043. to configure multiple bonds with differing parameters (as they are older
  1044. kernels, and also lack sysfs support).
  1045. 3.4 Configuring Bonding Manually via Sysfs
  1046. ------------------------------------------
  1047. Starting with version 3.0.0, Channel Bonding may be configured
  1048. via the sysfs interface. This interface allows dynamic configuration
  1049. of all bonds in the system without unloading the module. It also
  1050. allows for adding and removing bonds at runtime. Ifenslave is no
  1051. longer required, though it is still supported.
  1052. Use of the sysfs interface allows you to use multiple bonds
  1053. with different configurations without having to reload the module.
  1054. It also allows you to use multiple, differently configured bonds when
  1055. bonding is compiled into the kernel.
  1056. You must have the sysfs filesystem mounted to configure
  1057. bonding this way. The examples in this document assume that you
  1058. are using the standard mount point for sysfs, e.g. /sys. If your
  1059. sysfs filesystem is mounted elsewhere, you will need to adjust the
  1060. example paths accordingly.
  1061. Creating and Destroying Bonds
  1062. -----------------------------
  1063. To add a new bond foo:
  1064. # echo +foo > /sys/class/net/bonding_masters
  1065. To remove an existing bond bar:
  1066. # echo -bar > /sys/class/net/bonding_masters
  1067. To show all existing bonds:
  1068. # cat /sys/class/net/bonding_masters
  1069. NOTE: due to 4K size limitation of sysfs files, this list may be
  1070. truncated if you have more than a few hundred bonds. This is unlikely
  1071. to occur under normal operating conditions.
  1072. Adding and Removing Slaves
  1073. --------------------------
  1074. Interfaces may be enslaved to a bond using the file
  1075. /sys/class/net/<bond>/bonding/slaves. The semantics for this file
  1076. are the same as for the bonding_masters file.
  1077. To enslave interface eth0 to bond bond0:
  1078. # ifconfig bond0 up
  1079. # echo +eth0 > /sys/class/net/bond0/bonding/slaves
  1080. To free slave eth0 from bond bond0:
  1081. # echo -eth0 > /sys/class/net/bond0/bonding/slaves
  1082. When an interface is enslaved to a bond, symlinks between the
  1083. two are created in the sysfs filesystem. In this case, you would get
  1084. /sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
  1085. /sys/class/net/eth0/master pointing to /sys/class/net/bond0.
  1086. This means that you can tell quickly whether or not an
  1087. interface is enslaved by looking for the master symlink. Thus:
  1088. # echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
  1089. will free eth0 from whatever bond it is enslaved to, regardless of
  1090. the name of the bond interface.
  1091. Changing a Bond's Configuration
  1092. -------------------------------
  1093. Each bond may be configured individually by manipulating the
  1094. files located in /sys/class/net/<bond name>/bonding
  1095. The names of these files correspond directly with the command-
  1096. line parameters described elsewhere in this file, and, with the
  1097. exception of arp_ip_target, they accept the same values. To see the
  1098. current setting, simply cat the appropriate file.
  1099. A few examples will be given here; for specific usage
  1100. guidelines for each parameter, see the appropriate section in this
  1101. document.
  1102. To configure bond0 for balance-alb mode:
  1103. # ifconfig bond0 down
  1104. # echo 6 > /sys/class/net/bond0/bonding/mode
  1105. - or -
  1106. # echo balance-alb > /sys/class/net/bond0/bonding/mode
  1107. NOTE: The bond interface must be down before the mode can be
  1108. changed.
  1109. To enable MII monitoring on bond0 with a 1 second interval:
  1110. # echo 1000 > /sys/class/net/bond0/bonding/miimon
  1111. NOTE: If ARP monitoring is enabled, it will disabled when MII
  1112. monitoring is enabled, and vice-versa.
  1113. To add ARP targets:
  1114. # echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
  1115. # echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
  1116. NOTE: up to 16 target addresses may be specified.
  1117. To remove an ARP target:
  1118. # echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
  1119. To configure the interval between learning packet transmits:
  1120. # echo 12 > /sys/class/net/bond0/bonding/lp_interval
  1121. NOTE: the lp_inteval is the number of seconds between instances where
  1122. the bonding driver sends learning packets to each slaves peer switch. The
  1123. default interval is 1 second.
  1124. Example Configuration
  1125. ---------------------
  1126. We begin with the same example that is shown in section 3.3,
  1127. executed with sysfs, and without using ifenslave.
  1128. To make a simple bond of two e100 devices (presumed to be eth0
  1129. and eth1), and have it persist across reboots, edit the appropriate
  1130. file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
  1131. following:
  1132. modprobe bonding
  1133. modprobe e100
  1134. echo balance-alb > /sys/class/net/bond0/bonding/mode
  1135. ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
  1136. echo 100 > /sys/class/net/bond0/bonding/miimon
  1137. echo +eth0 > /sys/class/net/bond0/bonding/slaves
  1138. echo +eth1 > /sys/class/net/bond0/bonding/slaves
  1139. To add a second bond, with two e1000 interfaces in
  1140. active-backup mode, using ARP monitoring, add the following lines to
  1141. your init script:
  1142. modprobe e1000
  1143. echo +bond1 > /sys/class/net/bonding_masters
  1144. echo active-backup > /sys/class/net/bond1/bonding/mode
  1145. ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
  1146. echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
  1147. echo 2000 > /sys/class/net/bond1/bonding/arp_interval
  1148. echo +eth2 > /sys/class/net/bond1/bonding/slaves
  1149. echo +eth3 > /sys/class/net/bond1/bonding/slaves
  1150. 3.5 Configuration with Interfaces Support
  1151. -----------------------------------------
  1152. This section applies to distros which use /etc/network/interfaces file
  1153. to describe network interface configuration, most notably Debian and it's
  1154. derivatives.
  1155. The ifup and ifdown commands on Debian don't support bonding out of
  1156. the box. The ifenslave-2.6 package should be installed to provide bonding
  1157. support. Once installed, this package will provide bond-* options to be used
  1158. into /etc/network/interfaces.
  1159. Note that ifenslave-2.6 package will load the bonding module and use
  1160. the ifenslave command when appropriate.
  1161. Example Configurations
  1162. ----------------------
  1163. In /etc/network/interfaces, the following stanza will configure bond0, in
  1164. active-backup mode, with eth0 and eth1 as slaves.
  1165. auto bond0
  1166. iface bond0 inet dhcp
  1167. bond-slaves eth0 eth1
  1168. bond-mode active-backup
  1169. bond-miimon 100
  1170. bond-primary eth0 eth1
  1171. If the above configuration doesn't work, you might have a system using
  1172. upstart for system startup. This is most notably true for recent
  1173. Ubuntu versions. The following stanza in /etc/network/interfaces will
  1174. produce the same result on those systems.
  1175. auto bond0
  1176. iface bond0 inet dhcp
  1177. bond-slaves none
  1178. bond-mode active-backup
  1179. bond-miimon 100
  1180. auto eth0
  1181. iface eth0 inet manual
  1182. bond-master bond0
  1183. bond-primary eth0 eth1
  1184. auto eth1
  1185. iface eth1 inet manual
  1186. bond-master bond0
  1187. bond-primary eth0 eth1
  1188. For a full list of bond-* supported options in /etc/network/interfaces and some
  1189. more advanced examples tailored to you particular distros, see the files in
  1190. /usr/share/doc/ifenslave-2.6.
  1191. 3.6 Overriding Configuration for Special Cases
  1192. ----------------------------------------------
  1193. When using the bonding driver, the physical port which transmits a frame is
  1194. typically selected by the bonding driver, and is not relevant to the user or
  1195. system administrator. The output port is simply selected using the policies of
  1196. the selected bonding mode. On occasion however, it is helpful to direct certain
  1197. classes of traffic to certain physical interfaces on output to implement
  1198. slightly more complex policies. For example, to reach a web server over a
  1199. bonded interface in which eth0 connects to a private network, while eth1
  1200. connects via a public network, it may be desirous to bias the bond to send said
  1201. traffic over eth0 first, using eth1 only as a fall back, while all other traffic
  1202. can safely be sent over either interface. Such configurations may be achieved
  1203. using the traffic control utilities inherent in linux.
  1204. By default the bonding driver is multiqueue aware and 16 queues are created
  1205. when the driver initializes (see Documentation/networking/multiqueue.txt
  1206. for details). If more or less queues are desired the module parameter
  1207. tx_queues can be used to change this value. There is no sysfs parameter
  1208. available as the allocation is done at module init time.
  1209. The output of the file /proc/net/bonding/bondX has changed so the output Queue
  1210. ID is now printed for each slave:
  1211. Bonding Mode: fault-tolerance (active-backup)
  1212. Primary Slave: None
  1213. Currently Active Slave: eth0
  1214. MII Status: up
  1215. MII Polling Interval (ms): 0
  1216. Up Delay (ms): 0
  1217. Down Delay (ms): 0
  1218. Slave Interface: eth0
  1219. MII Status: up
  1220. Link Failure Count: 0
  1221. Permanent HW addr: 00:1a:a0:12:8f:cb
  1222. Slave queue ID: 0
  1223. Slave Interface: eth1
  1224. MII Status: up
  1225. Link Failure Count: 0
  1226. Permanent HW addr: 00:1a:a0:12:8f:cc
  1227. Slave queue ID: 2
  1228. The queue_id for a slave can be set using the command:
  1229. # echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
  1230. Any interface that needs a queue_id set should set it with multiple calls
  1231. like the one above until proper priorities are set for all interfaces. On
  1232. distributions that allow configuration via initscripts, multiple 'queue_id'
  1233. arguments can be added to BONDING_OPTS to set all needed slave queues.
  1234. These queue id's can be used in conjunction with the tc utility to configure
  1235. a multiqueue qdisc and filters to bias certain traffic to transmit on certain
  1236. slave devices. For instance, say we wanted, in the above configuration to
  1237. force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
  1238. device. The following commands would accomplish this:
  1239. # tc qdisc add dev bond0 handle 1 root multiq
  1240. # tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip dst \
  1241. 192.168.1.100 action skbedit queue_mapping 2
  1242. These commands tell the kernel to attach a multiqueue queue discipline to the
  1243. bond0 interface and filter traffic enqueued to it, such that packets with a dst
  1244. ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
  1245. This value is then passed into the driver, causing the normal output path
  1246. selection policy to be overridden, selecting instead qid 2, which maps to eth1.
  1247. Note that qid values begin at 1. Qid 0 is reserved to initiate to the driver
  1248. that normal output policy selection should take place. One benefit to simply
  1249. leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
  1250. driver that is now present. This awareness allows tc filters to be placed on
  1251. slave devices as well as bond devices and the bonding driver will simply act as
  1252. a pass-through for selecting output queues on the slave device rather than
  1253. output port selection.
  1254. This feature first appeared in bonding driver version 3.7.0 and support for
  1255. output slave selection was limited to round-robin and active-backup modes.
  1256. 3.7 Configuring LACP for 802.3ad mode in a more secure way
  1257. ----------------------------------------------------------
  1258. When using 802.3ad bonding mode, the Actor (host) and Partner (switch)
  1259. exchange LACPDUs. These LACPDUs cannot be sniffed, because they are
  1260. destined to link local mac addresses (which switches/bridges are not
  1261. supposed to forward). However, most of the values are easily predictable
  1262. or are simply the machine's MAC address (which is trivially known to all
  1263. other hosts in the same L2). This implies that other machines in the L2
  1264. domain can spoof LACPDU packets from other hosts to the switch and potentially
  1265. cause mayhem by joining (from the point of view of the switch) another
  1266. machine's aggregate, thus receiving a portion of that hosts incoming
  1267. traffic and / or spoofing traffic from that machine themselves (potentially
  1268. even successfully terminating some portion of flows). Though this is not
  1269. a likely scenario, one could avoid this possibility by simply configuring
  1270. few bonding parameters:
  1271. (a) ad_actor_system : You can set a random mac-address that can be used for
  1272. these LACPDU exchanges. The value can not be either NULL or Multicast.
  1273. Also it's preferable to set the local-admin bit. Following shell code
  1274. generates a random mac-address as described above.
  1275. # sys_mac_addr=$(printf '%02x:%02x:%02x:%02x:%02x:%02x' \
  1276. $(( (RANDOM & 0xFE) | 0x02 )) \
  1277. $(( RANDOM & 0xFF )) \
  1278. $(( RANDOM & 0xFF )) \
  1279. $(( RANDOM & 0xFF )) \
  1280. $(( RANDOM & 0xFF )) \
  1281. $(( RANDOM & 0xFF )))
  1282. # echo $sys_mac_addr > /sys/class/net/bond0/bonding/ad_actor_system
  1283. (b) ad_actor_sys_prio : Randomize the system priority. The default value
  1284. is 65535, but system can take the value from 1 - 65535. Following shell
  1285. code generates random priority and sets it.
  1286. # sys_prio=$(( 1 + RANDOM + RANDOM ))
  1287. # echo $sys_prio > /sys/class/net/bond0/bonding/ad_actor_sys_prio
  1288. (c) ad_user_port_key : Use the user portion of the port-key. The default
  1289. keeps this empty. These are the upper 10 bits of the port-key and value
  1290. ranges from 0 - 1023. Following shell code generates these 10 bits and
  1291. sets it.
  1292. # usr_port_key=$(( RANDOM & 0x3FF ))
  1293. # echo $usr_port_key > /sys/class/net/bond0/bonding/ad_user_port_key
  1294. 4 Querying Bonding Configuration
  1295. =================================
  1296. 4.1 Bonding Configuration
  1297. -------------------------
  1298. Each bonding device has a read-only file residing in the
  1299. /proc/net/bonding directory. The file contents include information
  1300. about the bonding configuration, options and state of each slave.
  1301. For example, the contents of /proc/net/bonding/bond0 after the
  1302. driver is loaded with parameters of mode=0 and miimon=1000 is
  1303. generally as follows:
  1304. Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
  1305. Bonding Mode: load balancing (round-robin)
  1306. Currently Active Slave: eth0
  1307. MII Status: up
  1308. MII Polling Interval (ms): 1000
  1309. Up Delay (ms): 0
  1310. Down Delay (ms): 0
  1311. Slave Interface: eth1
  1312. MII Status: up
  1313. Link Failure Count: 1
  1314. Slave Interface: eth0
  1315. MII Status: up
  1316. Link Failure Count: 1
  1317. The precise format and contents will change depending upon the
  1318. bonding configuration, state, and version of the bonding driver.
  1319. 4.2 Network configuration
  1320. -------------------------
  1321. The network configuration can be inspected using the ifconfig
  1322. command. Bonding devices will have the MASTER flag set; Bonding slave
  1323. devices will have the SLAVE flag set. The ifconfig output does not
  1324. contain information on which slaves are associated with which masters.
  1325. In the example below, the bond0 interface is the master
  1326. (MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
  1327. bond0 have the same MAC address (HWaddr) as bond0 for all modes except
  1328. TLB and ALB that require a unique MAC address for each slave.
  1329. # /sbin/ifconfig
  1330. bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
  1331. inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0
  1332. UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1
  1333. RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
  1334. TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
  1335. collisions:0 txqueuelen:0
  1336. eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
  1337. UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
  1338. RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
  1339. TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
  1340. collisions:0 txqueuelen:100
  1341. Interrupt:10 Base address:0x1080
  1342. eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4
  1343. UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1
  1344. RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
  1345. TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
  1346. collisions:0 txqueuelen:100
  1347. Interrupt:9 Base address:0x1400
  1348. 5. Switch Configuration
  1349. =======================
  1350. For this section, "switch" refers to whatever system the
  1351. bonded devices are directly connected to (i.e., where the other end of
  1352. the cable plugs into). This may be an actual dedicated switch device,
  1353. or it may be another regular system (e.g., another computer running
  1354. Linux),
  1355. The active-backup, balance-tlb and balance-alb modes do not
  1356. require any specific configuration of the switch.
  1357. The 802.3ad mode requires that the switch have the appropriate
  1358. ports configured as an 802.3ad aggregation. The precise method used
  1359. to configure this varies from switch to switch, but, for example, a
  1360. Cisco 3550 series switch requires that the appropriate ports first be
  1361. grouped together in a single etherchannel instance, then that
  1362. etherchannel is set to mode "lacp" to enable 802.3ad (instead of
  1363. standard EtherChannel).
  1364. The balance-rr, balance-xor and broadcast modes generally
  1365. require that the switch have the appropriate ports grouped together.
  1366. The nomenclature for such a group differs between switches, it may be
  1367. called an "etherchannel" (as in the Cisco example, above), a "trunk
  1368. group" or some other similar variation. For these modes, each switch
  1369. will also have its own configuration options for the switch's transmit
  1370. policy to the bond. Typical choices include XOR of either the MAC or
  1371. IP addresses. The transmit policy of the two peers does not need to
  1372. match. For these three modes, the bonding mode really selects a
  1373. transmit policy for an EtherChannel group; all three will interoperate
  1374. with another EtherChannel group.
  1375. 6. 802.1q VLAN Support
  1376. ======================
  1377. It is possible to configure VLAN devices over a bond interface
  1378. using the 8021q driver. However, only packets coming from the 8021q
  1379. driver and passing through bonding will be tagged by default. Self
  1380. generated packets, for example, bonding's learning packets or ARP
  1381. packets generated by either ALB mode or the ARP monitor mechanism, are
  1382. tagged internally by bonding itself. As a result, bonding must
  1383. "learn" the VLAN IDs configured above it, and use those IDs to tag
  1384. self generated packets.
  1385. For reasons of simplicity, and to support the use of adapters
  1386. that can do VLAN hardware acceleration offloading, the bonding
  1387. interface declares itself as fully hardware offloading capable, it gets
  1388. the add_vid/kill_vid notifications to gather the necessary
  1389. information, and it propagates those actions to the slaves. In case
  1390. of mixed adapter types, hardware accelerated tagged packets that
  1391. should go through an adapter that is not offloading capable are
  1392. "un-accelerated" by the bonding driver so the VLAN tag sits in the
  1393. regular location.
  1394. VLAN interfaces *must* be added on top of a bonding interface
  1395. only after enslaving at least one slave. The bonding interface has a
  1396. hardware address of 00:00:00:00:00:00 until the first slave is added.
  1397. If the VLAN interface is created prior to the first enslavement, it
  1398. would pick up the all-zeroes hardware address. Once the first slave
  1399. is attached to the bond, the bond device itself will pick up the
  1400. slave's hardware address, which is then available for the VLAN device.
  1401. Also, be aware that a similar problem can occur if all slaves
  1402. are released from a bond that still has one or more VLAN interfaces on
  1403. top of it. When a new slave is added, the bonding interface will
  1404. obtain its hardware address from the first slave, which might not
  1405. match the hardware address of the VLAN interfaces (which was
  1406. ultimately copied from an earlier slave).
  1407. There are two methods to insure that the VLAN device operates
  1408. with the correct hardware address if all slaves are removed from a
  1409. bond interface:
  1410. 1. Remove all VLAN interfaces then recreate them
  1411. 2. Set the bonding interface's hardware address so that it
  1412. matches the hardware address of the VLAN interfaces.
  1413. Note that changing a VLAN interface's HW address would set the
  1414. underlying device -- i.e. the bonding interface -- to promiscuous
  1415. mode, which might not be what you want.
  1416. 7. Link Monitoring
  1417. ==================
  1418. The bonding driver at present supports two schemes for
  1419. monitoring a slave device's link state: the ARP monitor and the MII
  1420. monitor.
  1421. At the present time, due to implementation restrictions in the
  1422. bonding driver itself, it is not possible to enable both ARP and MII
  1423. monitoring simultaneously.
  1424. 7.1 ARP Monitor Operation
  1425. -------------------------
  1426. The ARP monitor operates as its name suggests: it sends ARP
  1427. queries to one or more designated peer systems on the network, and
  1428. uses the response as an indication that the link is operating. This
  1429. gives some assurance that traffic is actually flowing to and from one
  1430. or more peers on the local network.
  1431. The ARP monitor relies on the device driver itself to verify
  1432. that traffic is flowing. In particular, the driver must keep up to
  1433. date the last receive time, dev->last_rx, and transmit start time,
  1434. dev->trans_start. If these are not updated by the driver, then the
  1435. ARP monitor will immediately fail any slaves using that driver, and
  1436. those slaves will stay down. If networking monitoring (tcpdump, etc)
  1437. shows the ARP requests and replies on the network, then it may be that
  1438. your device driver is not updating last_rx and trans_start.
  1439. 7.2 Configuring Multiple ARP Targets
  1440. ------------------------------------
  1441. While ARP monitoring can be done with just one target, it can
  1442. be useful in a High Availability setup to have several targets to
  1443. monitor. In the case of just one target, the target itself may go
  1444. down or have a problem making it unresponsive to ARP requests. Having
  1445. an additional target (or several) increases the reliability of the ARP
  1446. monitoring.
  1447. Multiple ARP targets must be separated by commas as follows:
  1448. # example options for ARP monitoring with three targets
  1449. alias bond0 bonding
  1450. options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
  1451. For just a single target the options would resemble:
  1452. # example options for ARP monitoring with one target
  1453. alias bond0 bonding
  1454. options bond0 arp_interval=60 arp_ip_target=192.168.0.100
  1455. 7.3 MII Monitor Operation
  1456. -------------------------
  1457. The MII monitor monitors only the carrier state of the local
  1458. network interface. It accomplishes this in one of three ways: by
  1459. depending upon the device driver to maintain its carrier state, by
  1460. querying the device's MII registers, or by making an ethtool query to
  1461. the device.
  1462. If the use_carrier module parameter is 1 (the default value),
  1463. then the MII monitor will rely on the driver for carrier state
  1464. information (via the netif_carrier subsystem). As explained in the
  1465. use_carrier parameter information, above, if the MII monitor fails to
  1466. detect carrier loss on the device (e.g., when the cable is physically
  1467. disconnected), it may be that the driver does not support
  1468. netif_carrier.
  1469. If use_carrier is 0, then the MII monitor will first query the
  1470. device's (via ioctl) MII registers and check the link state. If that
  1471. request fails (not just that it returns carrier down), then the MII
  1472. monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
  1473. the same information. If both methods fail (i.e., the driver either
  1474. does not support or had some error in processing both the MII register
  1475. and ethtool requests), then the MII monitor will assume the link is
  1476. up.
  1477. 8. Potential Sources of Trouble
  1478. ===============================
  1479. 8.1 Adventures in Routing
  1480. -------------------------
  1481. When bonding is configured, it is important that the slave
  1482. devices not have routes that supersede routes of the master (or,
  1483. generally, not have routes at all). For example, suppose the bonding
  1484. device bond0 has two slaves, eth0 and eth1, and the routing table is
  1485. as follows:
  1486. Kernel IP routing table
  1487. Destination Gateway Genmask Flags MSS Window irtt Iface
  1488. 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0
  1489. 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1
  1490. 10.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0
  1491. 127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo
  1492. This routing configuration will likely still update the
  1493. receive/transmit times in the driver (needed by the ARP monitor), but
  1494. may bypass the bonding driver (because outgoing traffic to, in this
  1495. case, another host on network 10 would use eth0 or eth1 before bond0).
  1496. The ARP monitor (and ARP itself) may become confused by this
  1497. configuration, because ARP requests (generated by the ARP monitor)
  1498. will be sent on one interface (bond0), but the corresponding reply
  1499. will arrive on a different interface (eth0). This reply looks to ARP
  1500. as an unsolicited ARP reply (because ARP matches replies on an
  1501. interface basis), and is discarded. The MII monitor is not affected
  1502. by the state of the routing table.
  1503. The solution here is simply to insure that slaves do not have
  1504. routes of their own, and if for some reason they must, those routes do
  1505. not supersede routes of their master. This should generally be the
  1506. case, but unusual configurations or errant manual or automatic static
  1507. route additions may cause trouble.
  1508. 8.2 Ethernet Device Renaming
  1509. ----------------------------
  1510. On systems with network configuration scripts that do not
  1511. associate physical devices directly with network interface names (so
  1512. that the same physical device always has the same "ethX" name), it may
  1513. be necessary to add some special logic to config files in
  1514. /etc/modprobe.d/.
  1515. For example, given a modules.conf containing the following:
  1516. alias bond0 bonding
  1517. options bond0 mode=some-mode miimon=50
  1518. alias eth0 tg3
  1519. alias eth1 tg3
  1520. alias eth2 e1000
  1521. alias eth3 e1000
  1522. If neither eth0 and eth1 are slaves to bond0, then when the
  1523. bond0 interface comes up, the devices may end up reordered. This
  1524. happens because bonding is loaded first, then its slave device's
  1525. drivers are loaded next. Since no other drivers have been loaded,
  1526. when the e1000 driver loads, it will receive eth0 and eth1 for its
  1527. devices, but the bonding configuration tries to enslave eth2 and eth3
  1528. (which may later be assigned to the tg3 devices).
  1529. Adding the following:
  1530. add above bonding e1000 tg3
  1531. causes modprobe to load e1000 then tg3, in that order, when
  1532. bonding is loaded. This command is fully documented in the
  1533. modules.conf manual page.
  1534. On systems utilizing modprobe an equivalent problem can occur.
  1535. In this case, the following can be added to config files in
  1536. /etc/modprobe.d/ as:
  1537. softdep bonding pre: tg3 e1000
  1538. This will load tg3 and e1000 modules before loading the bonding one.
  1539. Full documentation on this can be found in the modprobe.d and modprobe
  1540. manual pages.
  1541. 8.3. Painfully Slow Or No Failed Link Detection By Miimon
  1542. ---------------------------------------------------------
  1543. By default, bonding enables the use_carrier option, which
  1544. instructs bonding to trust the driver to maintain carrier state.
  1545. As discussed in the options section, above, some drivers do
  1546. not support the netif_carrier_on/_off link state tracking system.
  1547. With use_carrier enabled, bonding will always see these links as up,
  1548. regardless of their actual state.
  1549. Additionally, other drivers do support netif_carrier, but do
  1550. not maintain it in real time, e.g., only polling the link state at
  1551. some fixed interval. In this case, miimon will detect failures, but
  1552. only after some long period of time has expired. If it appears that
  1553. miimon is very slow in detecting link failures, try specifying
  1554. use_carrier=0 to see if that improves the failure detection time. If
  1555. it does, then it may be that the driver checks the carrier state at a
  1556. fixed interval, but does not cache the MII register values (so the
  1557. use_carrier=0 method of querying the registers directly works). If
  1558. use_carrier=0 does not improve the failover, then the driver may cache
  1559. the registers, or the problem may be elsewhere.
  1560. Also, remember that miimon only checks for the device's
  1561. carrier state. It has no way to determine the state of devices on or
  1562. beyond other ports of a switch, or if a switch is refusing to pass
  1563. traffic while still maintaining carrier on.
  1564. 9. SNMP agents
  1565. ===============
  1566. If running SNMP agents, the bonding driver should be loaded
  1567. before any network drivers participating in a bond. This requirement
  1568. is due to the interface index (ipAdEntIfIndex) being associated to
  1569. the first interface found with a given IP address. That is, there is
  1570. only one ipAdEntIfIndex for each IP address. For example, if eth0 and
  1571. eth1 are slaves of bond0 and the driver for eth0 is loaded before the
  1572. bonding driver, the interface for the IP address will be associated
  1573. with the eth0 interface. This configuration is shown below, the IP
  1574. address 192.168.1.1 has an interface index of 2 which indexes to eth0
  1575. in the ifDescr table (ifDescr.2).
  1576. interfaces.ifTable.ifEntry.ifDescr.1 = lo
  1577. interfaces.ifTable.ifEntry.ifDescr.2 = eth0
  1578. interfaces.ifTable.ifEntry.ifDescr.3 = eth1
  1579. interfaces.ifTable.ifEntry.ifDescr.4 = eth2
  1580. interfaces.ifTable.ifEntry.ifDescr.5 = eth3
  1581. interfaces.ifTable.ifEntry.ifDescr.6 = bond0
  1582. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
  1583. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
  1584. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
  1585. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
  1586. This problem is avoided by loading the bonding driver before
  1587. any network drivers participating in a bond. Below is an example of
  1588. loading the bonding driver first, the IP address 192.168.1.1 is
  1589. correctly associated with ifDescr.2.
  1590. interfaces.ifTable.ifEntry.ifDescr.1 = lo
  1591. interfaces.ifTable.ifEntry.ifDescr.2 = bond0
  1592. interfaces.ifTable.ifEntry.ifDescr.3 = eth0
  1593. interfaces.ifTable.ifEntry.ifDescr.4 = eth1
  1594. interfaces.ifTable.ifEntry.ifDescr.5 = eth2
  1595. interfaces.ifTable.ifEntry.ifDescr.6 = eth3
  1596. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
  1597. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
  1598. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
  1599. ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
  1600. While some distributions may not report the interface name in
  1601. ifDescr, the association between the IP address and IfIndex remains
  1602. and SNMP functions such as Interface_Scan_Next will report that
  1603. association.
  1604. 10. Promiscuous mode
  1605. ====================
  1606. When running network monitoring tools, e.g., tcpdump, it is
  1607. common to enable promiscuous mode on the device, so that all traffic
  1608. is seen (instead of seeing only traffic destined for the local host).
  1609. The bonding driver handles promiscuous mode changes to the bonding
  1610. master device (e.g., bond0), and propagates the setting to the slave
  1611. devices.
  1612. For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
  1613. the promiscuous mode setting is propagated to all slaves.
  1614. For the active-backup, balance-tlb and balance-alb modes, the
  1615. promiscuous mode setting is propagated only to the active slave.
  1616. For balance-tlb mode, the active slave is the slave currently
  1617. receiving inbound traffic.
  1618. For balance-alb mode, the active slave is the slave used as a
  1619. "primary." This slave is used for mode-specific control traffic, for
  1620. sending to peers that are unassigned or if the load is unbalanced.
  1621. For the active-backup, balance-tlb and balance-alb modes, when
  1622. the active slave changes (e.g., due to a link failure), the
  1623. promiscuous setting will be propagated to the new active slave.
  1624. 11. Configuring Bonding for High Availability
  1625. =============================================
  1626. High Availability refers to configurations that provide
  1627. maximum network availability by having redundant or backup devices,
  1628. links or switches between the host and the rest of the world. The
  1629. goal is to provide the maximum availability of network connectivity
  1630. (i.e., the network always works), even though other configurations
  1631. could provide higher throughput.
  1632. 11.1 High Availability in a Single Switch Topology
  1633. --------------------------------------------------
  1634. If two hosts (or a host and a single switch) are directly
  1635. connected via multiple physical links, then there is no availability
  1636. penalty to optimizing for maximum bandwidth. In this case, there is
  1637. only one switch (or peer), so if it fails, there is no alternative
  1638. access to fail over to. Additionally, the bonding load balance modes
  1639. support link monitoring of their members, so if individual links fail,
  1640. the load will be rebalanced across the remaining devices.
  1641. See Section 12, "Configuring Bonding for Maximum Throughput"
  1642. for information on configuring bonding with one peer device.
  1643. 11.2 High Availability in a Multiple Switch Topology
  1644. ----------------------------------------------------
  1645. With multiple switches, the configuration of bonding and the
  1646. network changes dramatically. In multiple switch topologies, there is
  1647. a trade off between network availability and usable bandwidth.
  1648. Below is a sample network, configured to maximize the
  1649. availability of the network:
  1650. | |
  1651. |port3 port3|
  1652. +-----+----+ +-----+----+
  1653. | |port2 ISL port2| |
  1654. | switch A +--------------------------+ switch B |
  1655. | | | |
  1656. +-----+----+ +-----++---+
  1657. |port1 port1|
  1658. | +-------+ |
  1659. +-------------+ host1 +---------------+
  1660. eth0 +-------+ eth1
  1661. In this configuration, there is a link between the two
  1662. switches (ISL, or inter switch link), and multiple ports connecting to
  1663. the outside world ("port3" on each switch). There is no technical
  1664. reason that this could not be extended to a third switch.
  1665. 11.2.1 HA Bonding Mode Selection for Multiple Switch Topology
  1666. -------------------------------------------------------------
  1667. In a topology such as the example above, the active-backup and
  1668. broadcast modes are the only useful bonding modes when optimizing for
  1669. availability; the other modes require all links to terminate on the
  1670. same peer for them to behave rationally.
  1671. active-backup: This is generally the preferred mode, particularly if
  1672. the switches have an ISL and play together well. If the
  1673. network configuration is such that one switch is specifically
  1674. a backup switch (e.g., has lower capacity, higher cost, etc),
  1675. then the primary option can be used to insure that the
  1676. preferred link is always used when it is available.
  1677. broadcast: This mode is really a special purpose mode, and is suitable
  1678. only for very specific needs. For example, if the two
  1679. switches are not connected (no ISL), and the networks beyond
  1680. them are totally independent. In this case, if it is
  1681. necessary for some specific one-way traffic to reach both
  1682. independent networks, then the broadcast mode may be suitable.
  1683. 11.2.2 HA Link Monitoring Selection for Multiple Switch Topology
  1684. ----------------------------------------------------------------
  1685. The choice of link monitoring ultimately depends upon your
  1686. switch. If the switch can reliably fail ports in response to other
  1687. failures, then either the MII or ARP monitors should work. For
  1688. example, in the above example, if the "port3" link fails at the remote
  1689. end, the MII monitor has no direct means to detect this. The ARP
  1690. monitor could be configured with a target at the remote end of port3,
  1691. thus detecting that failure without switch support.
  1692. In general, however, in a multiple switch topology, the ARP
  1693. monitor can provide a higher level of reliability in detecting end to
  1694. end connectivity failures (which may be caused by the failure of any
  1695. individual component to pass traffic for any reason). Additionally,
  1696. the ARP monitor should be configured with multiple targets (at least
  1697. one for each switch in the network). This will insure that,
  1698. regardless of which switch is active, the ARP monitor has a suitable
  1699. target to query.
  1700. Note, also, that of late many switches now support a functionality
  1701. generally referred to as "trunk failover." This is a feature of the
  1702. switch that causes the link state of a particular switch port to be set
  1703. down (or up) when the state of another switch port goes down (or up).
  1704. Its purpose is to propagate link failures from logically "exterior" ports
  1705. to the logically "interior" ports that bonding is able to monitor via
  1706. miimon. Availability and configuration for trunk failover varies by
  1707. switch, but this can be a viable alternative to the ARP monitor when using
  1708. suitable switches.
  1709. 12. Configuring Bonding for Maximum Throughput
  1710. ==============================================
  1711. 12.1 Maximizing Throughput in a Single Switch Topology
  1712. ------------------------------------------------------
  1713. In a single switch configuration, the best method to maximize
  1714. throughput depends upon the application and network environment. The
  1715. various load balancing modes each have strengths and weaknesses in
  1716. different environments, as detailed below.
  1717. For this discussion, we will break down the topologies into
  1718. two categories. Depending upon the destination of most traffic, we
  1719. categorize them into either "gatewayed" or "local" configurations.
  1720. In a gatewayed configuration, the "switch" is acting primarily
  1721. as a router, and the majority of traffic passes through this router to
  1722. other networks. An example would be the following:
  1723. +----------+ +----------+
  1724. | |eth0 port1| | to other networks
  1725. | Host A +---------------------+ router +------------------->
  1726. | +---------------------+ | Hosts B and C are out
  1727. | |eth1 port2| | here somewhere
  1728. +----------+ +----------+
  1729. The router may be a dedicated router device, or another host
  1730. acting as a gateway. For our discussion, the important point is that
  1731. the majority of traffic from Host A will pass through the router to
  1732. some other network before reaching its final destination.
  1733. In a gatewayed network configuration, although Host A may
  1734. communicate with many other systems, all of its traffic will be sent
  1735. and received via one other peer on the local network, the router.
  1736. Note that the case of two systems connected directly via
  1737. multiple physical links is, for purposes of configuring bonding, the
  1738. same as a gatewayed configuration. In that case, it happens that all
  1739. traffic is destined for the "gateway" itself, not some other network
  1740. beyond the gateway.
  1741. In a local configuration, the "switch" is acting primarily as
  1742. a switch, and the majority of traffic passes through this switch to
  1743. reach other stations on the same network. An example would be the
  1744. following:
  1745. +----------+ +----------+ +--------+
  1746. | |eth0 port1| +-------+ Host B |
  1747. | Host A +------------+ switch |port3 +--------+
  1748. | +------------+ | +--------+
  1749. | |eth1 port2| +------------------+ Host C |
  1750. +----------+ +----------+port4 +--------+
  1751. Again, the switch may be a dedicated switch device, or another
  1752. host acting as a gateway. For our discussion, the important point is
  1753. that the majority of traffic from Host A is destined for other hosts
  1754. on the same local network (Hosts B and C in the above example).
  1755. In summary, in a gatewayed configuration, traffic to and from
  1756. the bonded device will be to the same MAC level peer on the network
  1757. (the gateway itself, i.e., the router), regardless of its final
  1758. destination. In a local configuration, traffic flows directly to and
  1759. from the final destinations, thus, each destination (Host B, Host C)
  1760. will be addressed directly by their individual MAC addresses.
  1761. This distinction between a gatewayed and a local network
  1762. configuration is important because many of the load balancing modes
  1763. available use the MAC addresses of the local network source and
  1764. destination to make load balancing decisions. The behavior of each
  1765. mode is described below.
  1766. 12.1.1 MT Bonding Mode Selection for Single Switch Topology
  1767. -----------------------------------------------------------
  1768. This configuration is the easiest to set up and to understand,
  1769. although you will have to decide which bonding mode best suits your
  1770. needs. The trade offs for each mode are detailed below:
  1771. balance-rr: This mode is the only mode that will permit a single
  1772. TCP/IP connection to stripe traffic across multiple
  1773. interfaces. It is therefore the only mode that will allow a
  1774. single TCP/IP stream to utilize more than one interface's
  1775. worth of throughput. This comes at a cost, however: the
  1776. striping generally results in peer systems receiving packets out
  1777. of order, causing TCP/IP's congestion control system to kick
  1778. in, often by retransmitting segments.
  1779. It is possible to adjust TCP/IP's congestion limits by
  1780. altering the net.ipv4.tcp_reordering sysctl parameter. The
  1781. usual default value is 3. But keep in mind TCP stack is able
  1782. to automatically increase this when it detects reorders.
  1783. Note that the fraction of packets that will be delivered out of
  1784. order is highly variable, and is unlikely to be zero. The level
  1785. of reordering depends upon a variety of factors, including the
  1786. networking interfaces, the switch, and the topology of the
  1787. configuration. Speaking in general terms, higher speed network
  1788. cards produce more reordering (due to factors such as packet
  1789. coalescing), and a "many to many" topology will reorder at a
  1790. higher rate than a "many slow to one fast" configuration.
  1791. Many switches do not support any modes that stripe traffic
  1792. (instead choosing a port based upon IP or MAC level addresses);
  1793. for those devices, traffic for a particular connection flowing
  1794. through the switch to a balance-rr bond will not utilize greater
  1795. than one interface's worth of bandwidth.
  1796. If you are utilizing protocols other than TCP/IP, UDP for
  1797. example, and your application can tolerate out of order
  1798. delivery, then this mode can allow for single stream datagram
  1799. performance that scales near linearly as interfaces are added
  1800. to the bond.
  1801. This mode requires the switch to have the appropriate ports
  1802. configured for "etherchannel" or "trunking."
  1803. active-backup: There is not much advantage in this network topology to
  1804. the active-backup mode, as the inactive backup devices are all
  1805. connected to the same peer as the primary. In this case, a
  1806. load balancing mode (with link monitoring) will provide the
  1807. same level of network availability, but with increased
  1808. available bandwidth. On the plus side, active-backup mode
  1809. does not require any configuration of the switch, so it may
  1810. have value if the hardware available does not support any of
  1811. the load balance modes.
  1812. balance-xor: This mode will limit traffic such that packets destined
  1813. for specific peers will always be sent over the same
  1814. interface. Since the destination is determined by the MAC
  1815. addresses involved, this mode works best in a "local" network
  1816. configuration (as described above), with destinations all on
  1817. the same local network. This mode is likely to be suboptimal
  1818. if all your traffic is passed through a single router (i.e., a
  1819. "gatewayed" network configuration, as described above).
  1820. As with balance-rr, the switch ports need to be configured for
  1821. "etherchannel" or "trunking."
  1822. broadcast: Like active-backup, there is not much advantage to this
  1823. mode in this type of network topology.
  1824. 802.3ad: This mode can be a good choice for this type of network
  1825. topology. The 802.3ad mode is an IEEE standard, so all peers
  1826. that implement 802.3ad should interoperate well. The 802.3ad
  1827. protocol includes automatic configuration of the aggregates,
  1828. so minimal manual configuration of the switch is needed
  1829. (typically only to designate that some set of devices is
  1830. available for 802.3ad). The 802.3ad standard also mandates
  1831. that frames be delivered in order (within certain limits), so
  1832. in general single connections will not see misordering of
  1833. packets. The 802.3ad mode does have some drawbacks: the
  1834. standard mandates that all devices in the aggregate operate at
  1835. the same speed and duplex. Also, as with all bonding load
  1836. balance modes other than balance-rr, no single connection will
  1837. be able to utilize more than a single interface's worth of
  1838. bandwidth.
  1839. Additionally, the linux bonding 802.3ad implementation
  1840. distributes traffic by peer (using an XOR of MAC addresses
  1841. and packet type ID), so in a "gatewayed" configuration, all
  1842. outgoing traffic will generally use the same device. Incoming
  1843. traffic may also end up on a single device, but that is
  1844. dependent upon the balancing policy of the peer's 8023.ad
  1845. implementation. In a "local" configuration, traffic will be
  1846. distributed across the devices in the bond.
  1847. Finally, the 802.3ad mode mandates the use of the MII monitor,
  1848. therefore, the ARP monitor is not available in this mode.
  1849. balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
  1850. Since the balancing is done according to MAC address, in a
  1851. "gatewayed" configuration (as described above), this mode will
  1852. send all traffic across a single device. However, in a
  1853. "local" network configuration, this mode balances multiple
  1854. local network peers across devices in a vaguely intelligent
  1855. manner (not a simple XOR as in balance-xor or 802.3ad mode),
  1856. so that mathematically unlucky MAC addresses (i.e., ones that
  1857. XOR to the same value) will not all "bunch up" on a single
  1858. interface.
  1859. Unlike 802.3ad, interfaces may be of differing speeds, and no
  1860. special switch configuration is required. On the down side,
  1861. in this mode all incoming traffic arrives over a single
  1862. interface, this mode requires certain ethtool support in the
  1863. network device driver of the slave interfaces, and the ARP
  1864. monitor is not available.
  1865. balance-alb: This mode is everything that balance-tlb is, and more.
  1866. It has all of the features (and restrictions) of balance-tlb,
  1867. and will also balance incoming traffic from local network
  1868. peers (as described in the Bonding Module Options section,
  1869. above).
  1870. The only additional down side to this mode is that the network
  1871. device driver must support changing the hardware address while
  1872. the device is open.
  1873. 12.1.2 MT Link Monitoring for Single Switch Topology
  1874. ----------------------------------------------------
  1875. The choice of link monitoring may largely depend upon which
  1876. mode you choose to use. The more advanced load balancing modes do not
  1877. support the use of the ARP monitor, and are thus restricted to using
  1878. the MII monitor (which does not provide as high a level of end to end
  1879. assurance as the ARP monitor).
  1880. 12.2 Maximum Throughput in a Multiple Switch Topology
  1881. -----------------------------------------------------
  1882. Multiple switches may be utilized to optimize for throughput
  1883. when they are configured in parallel as part of an isolated network
  1884. between two or more systems, for example:
  1885. +-----------+
  1886. | Host A |
  1887. +-+---+---+-+
  1888. | | |
  1889. +--------+ | +---------+
  1890. | | |
  1891. +------+---+ +-----+----+ +-----+----+
  1892. | Switch A | | Switch B | | Switch C |
  1893. +------+---+ +-----+----+ +-----+----+
  1894. | | |
  1895. +--------+ | +---------+
  1896. | | |
  1897. +-+---+---+-+
  1898. | Host B |
  1899. +-----------+
  1900. In this configuration, the switches are isolated from one
  1901. another. One reason to employ a topology such as this is for an
  1902. isolated network with many hosts (a cluster configured for high
  1903. performance, for example), using multiple smaller switches can be more
  1904. cost effective than a single larger switch, e.g., on a network with 24
  1905. hosts, three 24 port switches can be significantly less expensive than
  1906. a single 72 port switch.
  1907. If access beyond the network is required, an individual host
  1908. can be equipped with an additional network device connected to an
  1909. external network; this host then additionally acts as a gateway.
  1910. 12.2.1 MT Bonding Mode Selection for Multiple Switch Topology
  1911. -------------------------------------------------------------
  1912. In actual practice, the bonding mode typically employed in
  1913. configurations of this type is balance-rr. Historically, in this
  1914. network configuration, the usual caveats about out of order packet
  1915. delivery are mitigated by the use of network adapters that do not do
  1916. any kind of packet coalescing (via the use of NAPI, or because the
  1917. device itself does not generate interrupts until some number of
  1918. packets has arrived). When employed in this fashion, the balance-rr
  1919. mode allows individual connections between two hosts to effectively
  1920. utilize greater than one interface's bandwidth.
  1921. 12.2.2 MT Link Monitoring for Multiple Switch Topology
  1922. ------------------------------------------------------
  1923. Again, in actual practice, the MII monitor is most often used
  1924. in this configuration, as performance is given preference over
  1925. availability. The ARP monitor will function in this topology, but its
  1926. advantages over the MII monitor are mitigated by the volume of probes
  1927. needed as the number of systems involved grows (remember that each
  1928. host in the network is configured with bonding).
  1929. 13. Switch Behavior Issues
  1930. ==========================
  1931. 13.1 Link Establishment and Failover Delays
  1932. -------------------------------------------
  1933. Some switches exhibit undesirable behavior with regard to the
  1934. timing of link up and down reporting by the switch.
  1935. First, when a link comes up, some switches may indicate that
  1936. the link is up (carrier available), but not pass traffic over the
  1937. interface for some period of time. This delay is typically due to
  1938. some type of autonegotiation or routing protocol, but may also occur
  1939. during switch initialization (e.g., during recovery after a switch
  1940. failure). If you find this to be a problem, specify an appropriate
  1941. value to the updelay bonding module option to delay the use of the
  1942. relevant interface(s).
  1943. Second, some switches may "bounce" the link state one or more
  1944. times while a link is changing state. This occurs most commonly while
  1945. the switch is initializing. Again, an appropriate updelay value may
  1946. help.
  1947. Note that when a bonding interface has no active links, the
  1948. driver will immediately reuse the first link that goes up, even if the
  1949. updelay parameter has been specified (the updelay is ignored in this
  1950. case). If there are slave interfaces waiting for the updelay timeout
  1951. to expire, the interface that first went into that state will be
  1952. immediately reused. This reduces down time of the network if the
  1953. value of updelay has been overestimated, and since this occurs only in
  1954. cases with no connectivity, there is no additional penalty for
  1955. ignoring the updelay.
  1956. In addition to the concerns about switch timings, if your
  1957. switches take a long time to go into backup mode, it may be desirable
  1958. to not activate a backup interface immediately after a link goes down.
  1959. Failover may be delayed via the downdelay bonding module option.
  1960. 13.2 Duplicated Incoming Packets
  1961. --------------------------------
  1962. NOTE: Starting with version 3.0.2, the bonding driver has logic to
  1963. suppress duplicate packets, which should largely eliminate this problem.
  1964. The following description is kept for reference.
  1965. It is not uncommon to observe a short burst of duplicated
  1966. traffic when the bonding device is first used, or after it has been
  1967. idle for some period of time. This is most easily observed by issuing
  1968. a "ping" to some other host on the network, and noticing that the
  1969. output from ping flags duplicates (typically one per slave).
  1970. For example, on a bond in active-backup mode with five slaves
  1971. all connected to one switch, the output may appear as follows:
  1972. # ping -n 10.0.4.2
  1973. PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
  1974. 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
  1975. 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
  1976. 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
  1977. 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
  1978. 64 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
  1979. 64 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
  1980. 64 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
  1981. 64 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
  1982. This is not due to an error in the bonding driver, rather, it
  1983. is a side effect of how many switches update their MAC forwarding
  1984. tables. Initially, the switch does not associate the MAC address in
  1985. the packet with a particular switch port, and so it may send the
  1986. traffic to all ports until its MAC forwarding table is updated. Since
  1987. the interfaces attached to the bond may occupy multiple ports on a
  1988. single switch, when the switch (temporarily) floods the traffic to all
  1989. ports, the bond device receives multiple copies of the same packet
  1990. (one per slave device).
  1991. The duplicated packet behavior is switch dependent, some
  1992. switches exhibit this, and some do not. On switches that display this
  1993. behavior, it can be induced by clearing the MAC forwarding table (on
  1994. most Cisco switches, the privileged command "clear mac address-table
  1995. dynamic" will accomplish this).
  1996. 14. Hardware Specific Considerations
  1997. ====================================
  1998. This section contains additional information for configuring
  1999. bonding on specific hardware platforms, or for interfacing bonding
  2000. with particular switches or other devices.
  2001. 14.1 IBM BladeCenter
  2002. --------------------
  2003. This applies to the JS20 and similar systems.
  2004. On the JS20 blades, the bonding driver supports only
  2005. balance-rr, active-backup, balance-tlb and balance-alb modes. This is
  2006. largely due to the network topology inside the BladeCenter, detailed
  2007. below.
  2008. JS20 network adapter information
  2009. --------------------------------
  2010. All JS20s come with two Broadcom Gigabit Ethernet ports
  2011. integrated on the planar (that's "motherboard" in IBM-speak). In the
  2012. BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
  2013. I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
  2014. An add-on Broadcom daughter card can be installed on a JS20 to provide
  2015. two more Gigabit Ethernet ports. These ports, eth2 and eth3, are
  2016. wired to I/O Modules 3 and 4, respectively.
  2017. Each I/O Module may contain either a switch or a passthrough
  2018. module (which allows ports to be directly connected to an external
  2019. switch). Some bonding modes require a specific BladeCenter internal
  2020. network topology in order to function; these are detailed below.
  2021. Additional BladeCenter-specific networking information can be
  2022. found in two IBM Redbooks (www.ibm.com/redbooks):
  2023. "IBM eServer BladeCenter Networking Options"
  2024. "IBM eServer BladeCenter Layer 2-7 Network Switching"
  2025. BladeCenter networking configuration
  2026. ------------------------------------
  2027. Because a BladeCenter can be configured in a very large number
  2028. of ways, this discussion will be confined to describing basic
  2029. configurations.
  2030. Normally, Ethernet Switch Modules (ESMs) are used in I/O
  2031. modules 1 and 2. In this configuration, the eth0 and eth1 ports of a
  2032. JS20 will be connected to different internal switches (in the
  2033. respective I/O modules).
  2034. A passthrough module (OPM or CPM, optical or copper,
  2035. passthrough module) connects the I/O module directly to an external
  2036. switch. By using PMs in I/O module #1 and #2, the eth0 and eth1
  2037. interfaces of a JS20 can be redirected to the outside world and
  2038. connected to a common external switch.
  2039. Depending upon the mix of ESMs and PMs, the network will
  2040. appear to bonding as either a single switch topology (all PMs) or as a
  2041. multiple switch topology (one or more ESMs, zero or more PMs). It is
  2042. also possible to connect ESMs together, resulting in a configuration
  2043. much like the example in "High Availability in a Multiple Switch
  2044. Topology," above.
  2045. Requirements for specific modes
  2046. -------------------------------
  2047. The balance-rr mode requires the use of passthrough modules
  2048. for devices in the bond, all connected to an common external switch.
  2049. That switch must be configured for "etherchannel" or "trunking" on the
  2050. appropriate ports, as is usual for balance-rr.
  2051. The balance-alb and balance-tlb modes will function with
  2052. either switch modules or passthrough modules (or a mix). The only
  2053. specific requirement for these modes is that all network interfaces
  2054. must be able to reach all destinations for traffic sent over the
  2055. bonding device (i.e., the network must converge at some point outside
  2056. the BladeCenter).
  2057. The active-backup mode has no additional requirements.
  2058. Link monitoring issues
  2059. ----------------------
  2060. When an Ethernet Switch Module is in place, only the ARP
  2061. monitor will reliably detect link loss to an external switch. This is
  2062. nothing unusual, but examination of the BladeCenter cabinet would
  2063. suggest that the "external" network ports are the ethernet ports for
  2064. the system, when it fact there is a switch between these "external"
  2065. ports and the devices on the JS20 system itself. The MII monitor is
  2066. only able to detect link failures between the ESM and the JS20 system.
  2067. When a passthrough module is in place, the MII monitor does
  2068. detect failures to the "external" port, which is then directly
  2069. connected to the JS20 system.
  2070. Other concerns
  2071. --------------
  2072. The Serial Over LAN (SoL) link is established over the primary
  2073. ethernet (eth0) only, therefore, any loss of link to eth0 will result
  2074. in losing your SoL connection. It will not fail over with other
  2075. network traffic, as the SoL system is beyond the control of the
  2076. bonding driver.
  2077. It may be desirable to disable spanning tree on the switch
  2078. (either the internal Ethernet Switch Module, or an external switch) to
  2079. avoid fail-over delay issues when using bonding.
  2080. 15. Frequently Asked Questions
  2081. ==============================
  2082. 1. Is it SMP safe?
  2083. Yes. The old 2.0.xx channel bonding patch was not SMP safe.
  2084. The new driver was designed to be SMP safe from the start.
  2085. 2. What type of cards will work with it?
  2086. Any Ethernet type cards (you can even mix cards - a Intel
  2087. EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes,
  2088. devices need not be of the same speed.
  2089. Starting with version 3.2.1, bonding also supports Infiniband
  2090. slaves in active-backup mode.
  2091. 3. How many bonding devices can I have?
  2092. There is no limit.
  2093. 4. How many slaves can a bonding device have?
  2094. This is limited only by the number of network interfaces Linux
  2095. supports and/or the number of network cards you can place in your
  2096. system.
  2097. 5. What happens when a slave link dies?
  2098. If link monitoring is enabled, then the failing device will be
  2099. disabled. The active-backup mode will fail over to a backup link, and
  2100. other modes will ignore the failed link. The link will continue to be
  2101. monitored, and should it recover, it will rejoin the bond (in whatever
  2102. manner is appropriate for the mode). See the sections on High
  2103. Availability and the documentation for each mode for additional
  2104. information.
  2105. Link monitoring can be enabled via either the miimon or
  2106. arp_interval parameters (described in the module parameters section,
  2107. above). In general, miimon monitors the carrier state as sensed by
  2108. the underlying network device, and the arp monitor (arp_interval)
  2109. monitors connectivity to another host on the local network.
  2110. If no link monitoring is configured, the bonding driver will
  2111. be unable to detect link failures, and will assume that all links are
  2112. always available. This will likely result in lost packets, and a
  2113. resulting degradation of performance. The precise performance loss
  2114. depends upon the bonding mode and network configuration.
  2115. 6. Can bonding be used for High Availability?
  2116. Yes. See the section on High Availability for details.
  2117. 7. Which switches/systems does it work with?
  2118. The full answer to this depends upon the desired mode.
  2119. In the basic balance modes (balance-rr and balance-xor), it
  2120. works with any system that supports etherchannel (also called
  2121. trunking). Most managed switches currently available have such
  2122. support, and many unmanaged switches as well.
  2123. The advanced balance modes (balance-tlb and balance-alb) do
  2124. not have special switch requirements, but do need device drivers that
  2125. support specific features (described in the appropriate section under
  2126. module parameters, above).
  2127. In 802.3ad mode, it works with systems that support IEEE
  2128. 802.3ad Dynamic Link Aggregation. Most managed and many unmanaged
  2129. switches currently available support 802.3ad.
  2130. The active-backup mode should work with any Layer-II switch.
  2131. 8. Where does a bonding device get its MAC address from?
  2132. When using slave devices that have fixed MAC addresses, or when
  2133. the fail_over_mac option is enabled, the bonding device's MAC address is
  2134. the MAC address of the active slave.
  2135. For other configurations, if not explicitly configured (with
  2136. ifconfig or ip link), the MAC address of the bonding device is taken from
  2137. its first slave device. This MAC address is then passed to all following
  2138. slaves and remains persistent (even if the first slave is removed) until
  2139. the bonding device is brought down or reconfigured.
  2140. If you wish to change the MAC address, you can set it with
  2141. ifconfig or ip link:
  2142. # ifconfig bond0 hw ether 00:11:22:33:44:55
  2143. # ip link set bond0 address 66:77:88:99:aa:bb
  2144. The MAC address can be also changed by bringing down/up the
  2145. device and then changing its slaves (or their order):
  2146. # ifconfig bond0 down ; modprobe -r bonding
  2147. # ifconfig bond0 .... up
  2148. # ifenslave bond0 eth...
  2149. This method will automatically take the address from the next
  2150. slave that is added.
  2151. To restore your slaves' MAC addresses, you need to detach them
  2152. from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
  2153. then restore the MAC addresses that the slaves had before they were
  2154. enslaved.
  2155. 16. Resources and Links
  2156. =======================
  2157. The latest version of the bonding driver can be found in the latest
  2158. version of the linux kernel, found on http://kernel.org
  2159. The latest version of this document can be found in the latest kernel
  2160. source (named Documentation/networking/bonding.txt).
  2161. Discussions regarding the usage of the bonding driver take place on the
  2162. bonding-devel mailing list, hosted at sourceforge.net. If you have questions or
  2163. problems, post them to the list. The list address is:
  2164. bonding-devel@lists.sourceforge.net
  2165. The administrative interface (to subscribe or unsubscribe) can
  2166. be found at:
  2167. https://lists.sourceforge.net/lists/listinfo/bonding-devel
  2168. Discussions regarding the development of the bonding driver take place
  2169. on the main Linux network mailing list, hosted at vger.kernel.org. The list
  2170. address is:
  2171. netdev@vger.kernel.org
  2172. The administrative interface (to subscribe or unsubscribe) can
  2173. be found at:
  2174. http://vger.kernel.org/vger-lists.html#netdev
  2175. Donald Becker's Ethernet Drivers and diag programs may be found at :
  2176. - http://web.archive.org/web/*/http://www.scyld.com/network/
  2177. You will also find a lot of information regarding Ethernet, NWay, MII,
  2178. etc. at www.scyld.com.
  2179. -- END --