linux/Documentation/networking/bonding.txt
<<
>>
Prefs
   1
   2                Linux Ethernet Bonding Driver HOWTO
   3
   4                Latest update: 27 April 2011
   5
   6Initial release : Thomas Davis <tadavis at lbl.gov>
   7Corrections, HA extensions : 2000/10/03-15 :
   8  - Willy Tarreau <willy at meta-x.org>
   9  - Constantine Gavrilov <const-g at xpert.com>
  10  - Chad N. Tindel <ctindel at ieee dot org>
  11  - Janice Girouard <girouard at us dot ibm dot com>
  12  - Jay Vosburgh <fubar at us dot ibm dot com>
  13
  14Reorganized and updated Feb 2005 by Jay Vosburgh
  15Added Sysfs information: 2006/04/24
  16  - Mitch Williams <mitch.a.williams at intel.com>
  17
  18Introduction
  19============
  20
  21        The Linux bonding driver provides a method for aggregating
  22multiple network interfaces into a single logical "bonded" interface.
  23The behavior of the bonded interfaces depends upon the mode; generally
  24speaking, modes provide either hot standby or load balancing services.
  25Additionally, link integrity monitoring may be performed.
  26        
  27        The bonding driver originally came from Donald Becker's
  28beowulf patches for kernel 2.0. It has changed quite a bit since, and
  29the original tools from extreme-linux and beowulf sites will not work
  30with this version of the driver.
  31
  32        For new versions of the driver, updated userspace tools, and
  33who to ask for help, please follow the links at the end of this file.
  34
  35Table of Contents
  36=================
  37
  381. Bonding Driver Installation
  39
  402. Bonding Driver Options
  41
  423. Configuring Bonding Devices
  433.1     Configuration with Sysconfig Support
  443.1.1           Using DHCP with Sysconfig
  453.1.2           Configuring Multiple Bonds with Sysconfig
  463.2     Configuration with Initscripts Support
  473.2.1           Using DHCP with Initscripts
  483.2.2           Configuring Multiple Bonds with Initscripts
  493.3     Configuring Bonding Manually with Ifenslave
  503.3.1           Configuring Multiple Bonds Manually
  513.4     Configuring Bonding Manually via Sysfs
  523.5     Configuration with Interfaces Support
  533.6     Overriding Configuration for Special Cases
  54
  554. Querying Bonding Configuration
  564.1     Bonding Configuration
  574.2     Network Configuration
  58
  595. Switch Configuration
  60
  616. 802.1q VLAN Support
  62
  637. Link Monitoring
  647.1     ARP Monitor Operation
  657.2     Configuring Multiple ARP Targets
  667.3     MII Monitor Operation
  67
  688. Potential Trouble Sources
  698.1     Adventures in Routing
  708.2     Ethernet Device Renaming
  718.3     Painfully Slow Or No Failed Link Detection By Miimon
  72
  739. SNMP agents
  74
  7510. Promiscuous mode
  76
  7711. Configuring Bonding for High Availability
  7811.1    High Availability in a Single Switch Topology
  7911.2    High Availability in a Multiple Switch Topology
  8011.2.1          HA Bonding Mode Selection for Multiple Switch Topology
  8111.2.2          HA Link Monitoring for Multiple Switch Topology
  82
  8312. Configuring Bonding for Maximum Throughput
  8412.1    Maximum Throughput in a Single Switch Topology
  8512.1.1          MT Bonding Mode Selection for Single Switch Topology
  8612.1.2          MT Link Monitoring for Single Switch Topology
  8712.2    Maximum Throughput in a Multiple Switch Topology
  8812.2.1          MT Bonding Mode Selection for Multiple Switch Topology
  8912.2.2          MT Link Monitoring for Multiple Switch Topology
  90
  9113. Switch Behavior Issues
  9213.1    Link Establishment and Failover Delays
  9313.2    Duplicated Incoming Packets
  94
  9514. Hardware Specific Considerations
  9614.1    IBM BladeCenter
  97
  9815. Frequently Asked Questions
  99
 10016. Resources and Links
 101
 102
 1031. Bonding Driver Installation
 104==============================
 105
 106        Most popular distro kernels ship with the bonding driver
 107already available as a module. If your distro does not, or you
 108have need to compile bonding from source (e.g., configuring and
 109installing a mainline kernel from kernel.org), you'll need to perform
 110the following steps:
 111
 1121.1 Configure and build the kernel with bonding
 113-----------------------------------------------
 114
 115        The current version of the bonding driver is available in the
 116drivers/net/bonding subdirectory of the most recent kernel source
 117(which is available on http://kernel.org).  Most users "rolling their
 118own" will want to use the most recent kernel from kernel.org.
 119
 120        Configure kernel with "make menuconfig" (or "make xconfig" or
 121"make config"), then select "Bonding driver support" in the "Network
 122device support" section.  It is recommended that you configure the
 123driver as module since it is currently the only way to pass parameters
 124to the driver or configure more than one bonding device.
 125
 126        Build and install the new kernel and modules.
 127
 1281.2 Bonding Control Utility
 129-------------------------------------
 130
 131         It is recommended to configure bonding via iproute2 (netlink)
 132or sysfs, the old ifenslave control utility is obsolete.
 133
 1342. Bonding Driver Options
 135=========================
 136
 137        Options for the bonding driver are supplied as parameters to the
 138bonding module at load time, or are specified via sysfs.
 139
 140        Module options may be given as command line arguments to the
 141insmod or modprobe command, but are usually specified in either the
 142/etc/modrobe.d/*.conf configuration files, or in a distro-specific
 143configuration file (some of which are detailed in the next section).
 144
 145        Details on bonding support for sysfs is provided in the
 146"Configuring Bonding Manually via Sysfs" section, below.
 147
 148        The available bonding driver parameters are listed below. If a
 149parameter is not specified the default value is used.  When initially
 150configuring a bond, it is recommended "tail -f /var/log/messages" be
 151run in a separate window to watch for bonding driver error messages.
 152
 153        It is critical that either the miimon or arp_interval and
 154arp_ip_target parameters be specified, otherwise serious network
 155degradation will occur during link failures.  Very few devices do not
 156support at least miimon, so there is really no reason not to use it.
 157
 158        Options with textual values will accept either the text name
 159or, for backwards compatibility, the option value.  E.g.,
 160"mode=802.3ad" and "mode=4" set the same mode.
 161
 162        The parameters are as follows:
 163
 164active_slave
 165
 166        Specifies the new active slave for modes that support it
 167        (active-backup, balance-alb and balance-tlb).  Possible values
 168        are the name of any currently enslaved interface, or an empty
 169        string.  If a name is given, the slave and its link must be up in order
 170        to be selected as the new active slave.  If an empty string is
 171        specified, the current active slave is cleared, and a new active
 172        slave is selected automatically.
 173
 174        Note that this is only available through the sysfs interface. No module
 175        parameter by this name exists.
 176
 177        The normal value of this option is the name of the currently
 178        active slave, or the empty string if there is no active slave or
 179        the current mode does not use an active slave.
 180
 181ad_select
 182
 183        Specifies the 802.3ad aggregation selection logic to use.  The
 184        possible values and their effects are:
 185
 186        stable or 0
 187
 188                The active aggregator is chosen by largest aggregate
 189                bandwidth.
 190
 191                Reselection of the active aggregator occurs only when all
 192                slaves of the active aggregator are down or the active
 193                aggregator has no slaves.
 194
 195                This is the default value.
 196
 197        bandwidth or 1
 198
 199                The active aggregator is chosen by largest aggregate
 200                bandwidth.  Reselection occurs if:
 201
 202                - A slave is added to or removed from the bond
 203
 204                - Any slave's link state changes
 205
 206                - Any slave's 802.3ad association state changes
 207
 208                - The bond's administrative state changes to up
 209
 210        count or 2
 211
 212                The active aggregator is chosen by the largest number of
 213                ports (slaves).  Reselection occurs as described under the
 214                "bandwidth" setting, above.
 215
 216        The bandwidth and count selection policies permit failover of
 217        802.3ad aggregations when partial failure of the active aggregator
 218        occurs.  This keeps the aggregator with the highest availability
 219        (either in bandwidth or in number of ports) active at all times.
 220
 221        This option was added in bonding version 3.4.0.
 222
 223all_slaves_active
 224
 225        Specifies that duplicate frames (received on inactive ports) should be
 226        dropped (0) or delivered (1).
 227
 228        Normally, bonding will drop duplicate frames (received on inactive
 229        ports), which is desirable for most users. But there are some times
 230        it is nice to allow duplicate frames to be delivered.
 231
 232        The default value is 0 (drop duplicate frames received on inactive
 233        ports).
 234
 235arp_interval
 236
 237        Specifies the ARP link monitoring frequency in milliseconds.
 238
 239        The ARP monitor works by periodically checking the slave
 240        devices to determine whether they have sent or received
 241        traffic recently (the precise criteria depends upon the
 242        bonding mode, and the state of the slave).  Regular traffic is
 243        generated via ARP probes issued for the addresses specified by
 244        the arp_ip_target option.
 245
 246        This behavior can be modified by the arp_validate option,
 247        below.
 248
 249        If ARP monitoring is used in an etherchannel compatible mode
 250        (modes 0 and 2), the switch should be configured in a mode
 251        that evenly distributes packets across all links. If the
 252        switch is configured to distribute the packets in an XOR
 253        fashion, all replies from the ARP targets will be received on
 254        the same link which could cause the other team members to
 255        fail.  ARP monitoring should not be used in conjunction with
 256        miimon.  A value of 0 disables ARP monitoring.  The default
 257        value is 0.
 258
 259arp_ip_target
 260
 261        Specifies the IP addresses to use as ARP monitoring peers when
 262        arp_interval is > 0.  These are the targets of the ARP request
 263        sent to determine the health of the link to the targets.
 264        Specify these values in ddd.ddd.ddd.ddd format.  Multiple IP
 265        addresses must be separated by a comma.  At least one IP
 266        address must be given for ARP monitoring to function.  The
 267        maximum number of targets that can be specified is 16.  The
 268        default value is no IP addresses.
 269
 270arp_validate
 271
 272        Specifies whether or not ARP probes and replies should be
 273        validated in any mode that supports arp monitoring, or whether
 274        non-ARP traffic should be filtered (disregarded) for link
 275        monitoring purposes.
 276
 277        Possible values are:
 278
 279        none or 0
 280
 281                No validation or filtering is performed.
 282
 283        active or 1
 284
 285                Validation is performed only for the active slave.
 286
 287        backup or 2
 288
 289                Validation is performed only for backup slaves.
 290
 291        all or 3
 292
 293                Validation is performed for all slaves.
 294
 295        filter or 4
 296
 297                Filtering is applied to all slaves. No validation is
 298                performed.
 299
 300        filter_active or 5
 301
 302                Filtering is applied to all slaves, validation is performed
 303                only for the active slave.
 304
 305        filter_backup or 6
 306
 307                Filtering is applied to all slaves, validation is performed
 308                only for backup slaves.
 309
 310        Validation:
 311
 312        Enabling validation causes the ARP monitor to examine the incoming
 313        ARP requests and replies, and only consider a slave to be up if it
 314        is receiving the appropriate ARP traffic.
 315
 316        For an active slave, the validation checks ARP replies to confirm
 317        that they were generated by an arp_ip_target.  Since backup slaves
 318        do not typically receive these replies, the validation performed
 319        for backup slaves is on the broadcast ARP request sent out via the
 320        active slave.  It is possible that some switch or network
 321        configurations may result in situations wherein the backup slaves
 322        do not receive the ARP requests; in such a situation, validation
 323        of backup slaves must be disabled.
 324
 325        The validation of ARP requests on backup slaves is mainly helping
 326        bonding to decide which slaves are more likely to work in case of
 327        the active slave failure, it doesn't really guarantee that the
 328        backup slave will work if it's selected as the next active slave.
 329
 330        Validation is useful in network configurations in which multiple
 331        bonding hosts are concurrently issuing ARPs to one or more targets
 332        beyond a common switch.  Should the link between the switch and
 333        target fail (but not the switch itself), the probe traffic
 334        generated by the multiple bonding instances will fool the standard
 335        ARP monitor into considering the links as still up.  Use of
 336        validation can resolve this, as the ARP monitor will only consider
 337        ARP requests and replies associated with its own instance of
 338        bonding.
 339
 340        Filtering:
 341
 342        Enabling filtering causes the ARP monitor to only use incoming ARP
 343        packets for link availability purposes.  Arriving packets that are
 344        not ARPs are delivered normally, but do not count when determining
 345        if a slave is available.
 346
 347        Filtering operates by only considering the reception of ARP
 348        packets (any ARP packet, regardless of source or destination) when
 349        determining if a slave has received traffic for link availability
 350        purposes.
 351
 352        Filtering is useful in network configurations in which significant
 353        levels of third party broadcast traffic would fool the standard
 354        ARP monitor into considering the links as still up.  Use of
 355        filtering can resolve this, as only ARP traffic is considered for
 356        link availability purposes.
 357
 358        This option was added in bonding version 3.1.0.
 359
 360arp_all_targets
 361
 362        Specifies the quantity of arp_ip_targets that must be reachable
 363        in order for the ARP monitor to consider a slave as being up.
 364        This option affects only active-backup mode for slaves with
 365        arp_validation enabled.
 366
 367        Possible values are:
 368
 369        any or 0
 370
 371                consider the slave up only when any of the arp_ip_targets
 372                is reachable
 373
 374        all or 1
 375
 376                consider the slave up only when all of the arp_ip_targets
 377                are reachable
 378
 379downdelay
 380
 381        Specifies the time, in milliseconds, to wait before disabling
 382        a slave after a link failure has been detected.  This option
 383        is only valid for the miimon link monitor.  The downdelay
 384        value should be a multiple of the miimon value; if not, it
 385        will be rounded down to the nearest multiple.  The default
 386        value is 0.
 387
 388fail_over_mac
 389
 390        Specifies whether active-backup mode should set all slaves to
 391        the same MAC address at enslavement (the traditional
 392        behavior), or, when enabled, perform special handling of the
 393        bond's MAC address in accordance with the selected policy.
 394
 395        Possible values are:
 396
 397        none or 0
 398
 399                This setting disables fail_over_mac, and causes
 400                bonding to set all slaves of an active-backup bond to
 401                the same MAC address at enslavement time.  This is the
 402                default.
 403
 404        active or 1
 405
 406                The "active" fail_over_mac policy indicates that the
 407                MAC address of the bond should always be the MAC
 408                address of the currently active slave.  The MAC
 409                address of the slaves is not changed; instead, the MAC
 410                address of the bond changes during a failover.
 411
 412                This policy is useful for devices that cannot ever
 413                alter their MAC address, or for devices that refuse
 414                incoming broadcasts with their own source MAC (which
 415                interferes with the ARP monitor).
 416
 417                The down side of this policy is that every device on
 418                the network must be updated via gratuitous ARP,
 419                vs. just updating a switch or set of switches (which
 420                often takes place for any traffic, not just ARP
 421                traffic, if the switch snoops incoming traffic to
 422                update its tables) for the traditional method.  If the
 423                gratuitous ARP is lost, communication may be
 424                disrupted.
 425
 426                When this policy is used in conjunction with the mii
 427                monitor, devices which assert link up prior to being
 428                able to actually transmit and receive are particularly
 429                susceptible to loss of the gratuitous ARP, and an
 430                appropriate updelay setting may be required.
 431
 432        follow or 2
 433
 434                The "follow" fail_over_mac policy causes the MAC
 435                address of the bond to be selected normally (normally
 436                the MAC address of the first slave added to the bond).
 437                However, the second and subsequent slaves are not set
 438                to this MAC address while they are in a backup role; a
 439                slave is programmed with the bond's MAC address at
 440                failover time (and the formerly active slave receives
 441                the newly active slave's MAC address).
 442
 443                This policy is useful for multiport devices that
 444                either become confused or incur a performance penalty
 445                when multiple ports are programmed with the same MAC
 446                address.
 447
 448
 449        The default policy is none, unless the first slave cannot
 450        change its MAC address, in which case the active policy is
 451        selected by default.
 452
 453        This option may be modified via sysfs only when no slaves are
 454        present in the bond.
 455
 456        This option was added in bonding version 3.2.0.  The "follow"
 457        policy was added in bonding version 3.3.0.
 458
 459lacp_rate
 460
 461        Option specifying the rate in which we'll ask our link partner
 462        to transmit LACPDU packets in 802.3ad mode.  Possible values
 463        are:
 464
 465        slow or 0
 466                Request partner to transmit LACPDUs every 30 seconds
 467
 468        fast or 1
 469                Request partner to transmit LACPDUs every 1 second
 470
 471        The default is slow.
 472
 473max_bonds
 474
 475        Specifies the number of bonding devices to create for this
 476        instance of the bonding driver.  E.g., if max_bonds is 3, and
 477        the bonding driver is not already loaded, then bond0, bond1
 478        and bond2 will be created.  The default value is 1.  Specifying
 479        a value of 0 will load bonding, but will not create any devices.
 480
 481miimon
 482
 483        Specifies the MII link monitoring frequency in milliseconds.
 484        This determines how often the link state of each slave is
 485        inspected for link failures.  A value of zero disables MII
 486        link monitoring.  A value of 100 is a good starting point.
 487        The use_carrier option, below, affects how the link state is
 488        determined.  See the High Availability section for additional
 489        information.  The default value is 0.
 490
 491min_links
 492
 493        Specifies the minimum number of links that must be active before
 494        asserting carrier. It is similar to the Cisco EtherChannel min-links
 495        feature. This allows setting the minimum number of member ports that
 496        must be up (link-up state) before marking the bond device as up
 497        (carrier on). This is useful for situations where higher level services
 498        such as clustering want to ensure a minimum number of low bandwidth
 499        links are active before switchover. This option only affect 802.3ad
 500        mode.
 501
 502        The default value is 0. This will cause carrier to be asserted (for
 503        802.3ad mode) whenever there is an active aggregator, regardless of the
 504        number of available links in that aggregator. Note that, because an
 505        aggregator cannot be active without at least one available link,
 506        setting this option to 0 or to 1 has the exact same effect.
 507
 508mode
 509
 510        Specifies one of the bonding policies. The default is
 511        balance-rr (round robin).  Possible values are:
 512
 513        balance-rr or 0
 514
 515                Round-robin policy: Transmit packets in sequential
 516                order from the first available slave through the
 517                last.  This mode provides load balancing and fault
 518                tolerance.
 519
 520        active-backup or 1
 521
 522                Active-backup policy: Only one slave in the bond is
 523                active.  A different slave becomes active if, and only
 524                if, the active slave fails.  The bond's MAC address is
 525                externally visible on only one port (network adapter)
 526                to avoid confusing the switch.
 527
 528                In bonding version 2.6.2 or later, when a failover
 529                occurs in active-backup mode, bonding will issue one
 530                or more gratuitous ARPs on the newly active slave.
 531                One gratuitous ARP is issued for the bonding master
 532                interface and each VLAN interfaces configured above
 533                it, provided that the interface has at least one IP
 534                address configured.  Gratuitous ARPs issued for VLAN
 535                interfaces are tagged with the appropriate VLAN id.
 536
 537                This mode provides fault tolerance.  The primary
 538                option, documented below, affects the behavior of this
 539                mode.
 540
 541        balance-xor or 2
 542
 543                XOR policy: Transmit based on the selected transmit
 544                hash policy.  The default policy is a simple [(source
 545                MAC address XOR'd with destination MAC address XOR
 546                packet type ID) modulo slave count].  Alternate transmit
 547                policies may be selected via the xmit_hash_policy option,
 548                described below.
 549
 550                This mode provides load balancing and fault tolerance.
 551
 552        broadcast or 3
 553
 554                Broadcast policy: transmits everything on all slave
 555                interfaces.  This mode provides fault tolerance.
 556
 557        802.3ad or 4
 558
 559                IEEE 802.3ad Dynamic link aggregation.  Creates
 560                aggregation groups that share the same speed and
 561                duplex settings.  Utilizes all slaves in the active
 562                aggregator according to the 802.3ad specification.
 563
 564                Slave selection for outgoing traffic is done according
 565                to the transmit hash policy, which may be changed from
 566                the default simple XOR policy via the xmit_hash_policy
 567                option, documented below.  Note that not all transmit
 568                policies may be 802.3ad compliant, particularly in
 569                regards to the packet mis-ordering requirements of
 570                section 43.2.4 of the 802.3ad standard.  Differing
 571                peer implementations will have varying tolerances for
 572                noncompliance.
 573
 574                Prerequisites:
 575
 576                1. Ethtool support in the base drivers for retrieving
 577                the speed and duplex of each slave.
 578
 579                2. A switch that supports IEEE 802.3ad Dynamic link
 580                aggregation.
 581
 582                Most switches will require some type of configuration
 583                to enable 802.3ad mode.
 584
 585        balance-tlb or 5
 586
 587                Adaptive transmit load balancing: channel bonding that
 588                does not require any special switch support.
 589
 590                In tlb_dynamic_lb=1 mode; the outgoing traffic is
 591                distributed according to the current load (computed
 592                relative to the speed) on each slave.
 593
 594                In tlb_dynamic_lb=0 mode; the load balancing based on
 595                current load is disabled and the load is distributed
 596                only using the hash distribution.
 597
 598                Incoming traffic is received by the current slave.
 599                If the receiving slave fails, another slave takes over
 600                the MAC address of the failed receiving slave.
 601
 602                Prerequisite:
 603
 604                Ethtool support in the base drivers for retrieving the
 605                speed of each slave.
 606
 607        balance-alb or 6
 608
 609                Adaptive load balancing: includes balance-tlb plus
 610                receive load balancing (rlb) for IPV4 traffic, and
 611                does not require any special switch support.  The
 612                receive load balancing is achieved by ARP negotiation.
 613                The bonding driver intercepts the ARP Replies sent by
 614                the local system on their way out and overwrites the
 615                source hardware address with the unique hardware
 616                address of one of the slaves in the bond such that
 617                different peers use different hardware addresses for
 618                the server.
 619
 620                Receive traffic from connections created by the server
 621                is also balanced.  When the local system sends an ARP
 622                Request the bonding driver copies and saves the peer's
 623                IP information from the ARP packet.  When the ARP
 624                Reply arrives from the peer, its hardware address is
 625                retrieved and the bonding driver initiates an ARP
 626                reply to this peer assigning it to one of the slaves
 627                in the bond.  A problematic outcome of using ARP
 628                negotiation for balancing is that each time that an
 629                ARP request is broadcast it uses the hardware address
 630                of the bond.  Hence, peers learn the hardware address
 631                of the bond and the balancing of receive traffic
 632                collapses to the current slave.  This is handled by
 633                sending updates (ARP Replies) to all the peers with
 634                their individually assigned hardware address such that
 635                the traffic is redistributed.  Receive traffic is also
 636                redistributed when a new slave is added to the bond
 637                and when an inactive slave is re-activated.  The
 638                receive load is distributed sequentially (round robin)
 639                among the group of highest speed slaves in the bond.
 640
 641                When a link is reconnected or a new slave joins the
 642                bond the receive traffic is redistributed among all
 643                active slaves in the bond by initiating ARP Replies
 644                with the selected MAC address to each of the
 645                clients. The updelay parameter (detailed below) must
 646                be set to a value equal or greater than the switch's
 647                forwarding delay so that the ARP Replies sent to the
 648                peers will not be blocked by the switch.
 649
 650                Prerequisites:
 651
 652                1. Ethtool support in the base drivers for retrieving
 653                the speed of each slave.
 654
 655                2. Base driver support for setting the hardware
 656                address of a device while it is open.  This is
 657                required so that there will always be one slave in the
 658                team using the bond hardware address (the
 659                curr_active_slave) while having a unique hardware
 660                address for each slave in the bond.  If the
 661                curr_active_slave fails its hardware address is
 662                swapped with the new curr_active_slave that was
 663                chosen.
 664
 665num_grat_arp
 666num_unsol_na
 667
 668        Specify the number of peer notifications (gratuitous ARPs and
 669        unsolicited IPv6 Neighbor Advertisements) to be issued after a
 670        failover event.  As soon as the link is up on the new slave
 671        (possibly immediately) a peer notification is sent on the
 672        bonding device and each VLAN sub-device.  This is repeated at
 673        each link monitor interval (arp_interval or miimon, whichever
 674        is active) if the number is greater than 1.
 675
 676        The valid range is 0 - 255; the default value is 1.  These options
 677        affect only the active-backup mode.  These options were added for
 678        bonding versions 3.3.0 and 3.4.0 respectively.
 679
 680        From Linux 3.0 and bonding version 3.7.1, these notifications
 681        are generated by the ipv4 and ipv6 code and the numbers of
 682        repetitions cannot be set independently.
 683
 684packets_per_slave
 685
 686        Specify the number of packets to transmit through a slave before
 687        moving to the next one. When set to 0 then a slave is chosen at
 688        random.
 689
 690        The valid range is 0 - 65535; the default value is 1. This option
 691        has effect only in balance-rr mode.
 692
 693primary
 694
 695        A string (eth0, eth2, etc) specifying which slave is the
 696        primary device.  The specified device will always be the
 697        active slave while it is available.  Only when the primary is
 698        off-line will alternate devices be used.  This is useful when
 699        one slave is preferred over another, e.g., when one slave has
 700        higher throughput than another.
 701
 702        The primary option is only valid for active-backup(1),
 703        balance-tlb (5) and balance-alb (6) mode.
 704
 705primary_reselect
 706
 707        Specifies the reselection policy for the primary slave.  This
 708        affects how the primary slave is chosen to become the active slave
 709        when failure of the active slave or recovery of the primary slave
 710        occurs.  This option is designed to prevent flip-flopping between
 711        the primary slave and other slaves.  Possible values are:
 712
 713        always or 0 (default)
 714
 715                The primary slave becomes the active slave whenever it
 716                comes back up.
 717
 718        better or 1
 719
 720                The primary slave becomes the active slave when it comes
 721                back up, if the speed and duplex of the primary slave is
 722                better than the speed and duplex of the current active
 723                slave.
 724
 725        failure or 2
 726
 727                The primary slave becomes the active slave only if the
 728                current active slave fails and the primary slave is up.
 729
 730        The primary_reselect setting is ignored in two cases:
 731
 732                If no slaves are active, the first slave to recover is
 733                made the active slave.
 734
 735                When initially enslaved, the primary slave is always made
 736                the active slave.
 737
 738        Changing the primary_reselect policy via sysfs will cause an
 739        immediate selection of the best active slave according to the new
 740        policy.  This may or may not result in a change of the active
 741        slave, depending upon the circumstances.
 742
 743        This option was added for bonding version 3.6.0.
 744
 745tlb_dynamic_lb
 746
 747        Specifies if dynamic shuffling of flows is enabled in tlb
 748        mode. The value has no effect on any other modes.
 749
 750        The default behavior of tlb mode is to shuffle active flows across
 751        slaves based on the load in that interval. This gives nice lb
 752        characteristics but can cause packet reordering. If re-ordering is
 753        a concern use this variable to disable flow shuffling and rely on
 754        load balancing provided solely by the hash distribution.
 755        xmit-hash-policy can be used to select the appropriate hashing for
 756        the setup.
 757
 758        The sysfs entry can be used to change the setting per bond device
 759        and the initial value is derived from the module parameter. The
 760        sysfs entry is allowed to be changed only if the bond device is
 761        down.
 762
 763        The default value is "1" that enables flow shuffling while value "0"
 764        disables it. This option was added in bonding driver 3.7.1
 765
 766
 767updelay
 768
 769        Specifies the time, in milliseconds, to wait before enabling a
 770        slave after a link recovery has been detected.  This option is
 771        only valid for the miimon link monitor.  The updelay value
 772        should be a multiple of the miimon value; if not, it will be
 773        rounded down to the nearest multiple.  The default value is 0.
 774
 775use_carrier
 776
 777        Specifies whether or not miimon should use MII or ETHTOOL
 778        ioctls vs. netif_carrier_ok() to determine the link
 779        status. The MII or ETHTOOL ioctls are less efficient and
 780        utilize a deprecated calling sequence within the kernel.  The
 781        netif_carrier_ok() relies on the device driver to maintain its
 782        state with netif_carrier_on/off; at this writing, most, but
 783        not all, device drivers support this facility.
 784
 785        If bonding insists that the link is up when it should not be,
 786        it may be that your network device driver does not support
 787        netif_carrier_on/off.  The default state for netif_carrier is
 788        "carrier on," so if a driver does not support netif_carrier,
 789        it will appear as if the link is always up.  In this case,
 790        setting use_carrier to 0 will cause bonding to revert to the
 791        MII / ETHTOOL ioctl method to determine the link state.
 792
 793        A value of 1 enables the use of netif_carrier_ok(), a value of
 794        0 will use the deprecated MII / ETHTOOL ioctls.  The default
 795        value is 1.
 796
 797xmit_hash_policy
 798
 799        Selects the transmit hash policy to use for slave selection in
 800        balance-xor, 802.3ad, and tlb modes.  Possible values are:
 801
 802        layer2
 803
 804                Uses XOR of hardware MAC addresses and packet type ID
 805                field to generate the hash. The formula is
 806
 807                hash = source MAC XOR destination MAC XOR packet type ID
 808                slave number = hash modulo slave count
 809
 810                This algorithm will place all traffic to a particular
 811                network peer on the same slave.
 812
 813                This algorithm is 802.3ad compliant.
 814
 815        layer2+3
 816
 817                This policy uses a combination of layer2 and layer3
 818                protocol information to generate the hash.
 819
 820                Uses XOR of hardware MAC addresses and IP addresses to
 821                generate the hash.  The formula is
 822
 823                hash = source MAC XOR destination MAC XOR packet type ID
 824                hash = hash XOR source IP XOR destination IP
 825                hash = hash XOR (hash RSHIFT 16)
 826                hash = hash XOR (hash RSHIFT 8)
 827                And then hash is reduced modulo slave count.
 828
 829                If the protocol is IPv6 then the source and destination
 830                addresses are first hashed using ipv6_addr_hash.
 831
 832                This algorithm will place all traffic to a particular
 833                network peer on the same slave.  For non-IP traffic,
 834                the formula is the same as for the layer2 transmit
 835                hash policy.
 836
 837                This policy is intended to provide a more balanced
 838                distribution of traffic than layer2 alone, especially
 839                in environments where a layer3 gateway device is
 840                required to reach most destinations.
 841
 842                This algorithm is 802.3ad compliant.
 843
 844        layer3+4
 845
 846                This policy uses upper layer protocol information,
 847                when available, to generate the hash.  This allows for
 848                traffic to a particular network peer to span multiple
 849                slaves, although a single connection will not span
 850                multiple slaves.
 851
 852                The formula for unfragmented TCP and UDP packets is
 853
 854                hash = source port, destination port (as in the header)
 855                hash = hash XOR source IP XOR destination IP
 856                hash = hash XOR (hash RSHIFT 16)
 857                hash = hash XOR (hash RSHIFT 8)
 858                And then hash is reduced modulo slave count.
 859
 860                If the protocol is IPv6 then the source and destination
 861                addresses are first hashed using ipv6_addr_hash.
 862
 863                For fragmented TCP or UDP packets and all other IPv4 and
 864                IPv6 protocol traffic, the source and destination port
 865                information is omitted.  For non-IP traffic, the
 866                formula is the same as for the layer2 transmit hash
 867                policy.
 868
 869                This algorithm is not fully 802.3ad compliant.  A
 870                single TCP or UDP conversation containing both
 871                fragmented and unfragmented packets will see packets
 872                striped across two interfaces.  This may result in out
 873                of order delivery.  Most traffic types will not meet
 874                this criteria, as TCP rarely fragments traffic, and
 875                most UDP traffic is not involved in extended
 876                conversations.  Other implementations of 802.3ad may
 877                or may not tolerate this noncompliance.
 878
 879        encap2+3
 880
 881                This policy uses the same formula as layer2+3 but it
 882                relies on skb_flow_dissect to obtain the header fields
 883                which might result in the use of inner headers if an
 884                encapsulation protocol is used. For example this will
 885                improve the performance for tunnel users because the
 886                packets will be distributed according to the encapsulated
 887                flows.
 888
 889        encap3+4
 890
 891                This policy uses the same formula as layer3+4 but it
 892                relies on skb_flow_dissect to obtain the header fields
 893                which might result in the use of inner headers if an
 894                encapsulation protocol is used. For example this will
 895                improve the performance for tunnel users because the
 896                packets will be distributed according to the encapsulated
 897                flows.
 898
 899        The default value is layer2.  This option was added in bonding
 900        version 2.6.3.  In earlier versions of bonding, this parameter
 901        does not exist, and the layer2 policy is the only policy.  The
 902        layer2+3 value was added for bonding version 3.2.2.
 903
 904resend_igmp
 905
 906        Specifies the number of IGMP membership reports to be issued after
 907        a failover event. One membership report is issued immediately after
 908        the failover, subsequent packets are sent in each 200ms interval.
 909
 910        The valid range is 0 - 255; the default value is 1. A value of 0
 911        prevents the IGMP membership report from being issued in response
 912        to the failover event.
 913
 914        This option is useful for bonding modes balance-rr (0), active-backup
 915        (1), balance-tlb (5) and balance-alb (6), in which a failover can
 916        switch the IGMP traffic from one slave to another.  Therefore a fresh
 917        IGMP report must be issued to cause the switch to forward the incoming
 918        IGMP traffic over the newly selected slave.
 919
 920        This option was added for bonding version 3.7.0.
 921
 922lp_interval
 923
 924        Specifies the number of seconds between instances where the bonding
 925        driver sends learning packets to each slaves peer switch.
 926
 927        The valid range is 1 - 0x7fffffff; the default value is 1. This Option
 928        has effect only in balance-tlb and balance-alb modes.
 929
 9303. Configuring Bonding Devices
 931==============================
 932
 933        You can configure bonding using either your distro's network
 934initialization scripts, or manually using either iproute2 or the
 935sysfs interface.  Distros generally use one of three packages for the
 936network initialization scripts: initscripts, sysconfig or interfaces.
 937Recent versions of these packages have support for bonding, while older
 938versions do not.
 939
 940        We will first describe the options for configuring bonding for
 941distros using versions of initscripts, sysconfig and interfaces with full
 942or partial support for bonding, then provide information on enabling
 943bonding without support from the network initialization scripts (i.e.,
 944older versions of initscripts or sysconfig).
 945
 946        If you're unsure whether your distro uses sysconfig,
 947initscripts or interfaces, or don't know if it's new enough, have no fear.
 948Determining this is fairly straightforward.
 949
 950        First, look for a file called interfaces in /etc/network directory.
 951If this file is present in your system, then your system use interfaces. See
 952Configuration with Interfaces Support.
 953
 954        Else, issue the command:
 955
 956$ rpm -qf /sbin/ifup
 957
 958        It will respond with a line of text starting with either
 959"initscripts" or "sysconfig," followed by some numbers.  This is the
 960package that provides your network initialization scripts.
 961
 962        Next, to determine if your installation supports bonding,
 963issue the command:
 964
 965$ grep ifenslave /sbin/ifup
 966
 967        If this returns any matches, then your initscripts or
 968sysconfig has support for bonding.
 969
 9703.1 Configuration with Sysconfig Support
 971----------------------------------------
 972
 973        This section applies to distros using a version of sysconfig
 974with bonding support, for example, SuSE Linux Enterprise Server 9.
 975
 976        SuSE SLES 9's networking configuration system does support
 977bonding, however, at this writing, the YaST system configuration
 978front end does not provide any means to work with bonding devices.
 979Bonding devices can be managed by hand, however, as follows.
 980
 981        First, if they have not already been configured, configure the
 982slave devices.  On SLES 9, this is most easily done by running the
 983yast2 sysconfig configuration utility.  The goal is for to create an
 984ifcfg-id file for each slave device.  The simplest way to accomplish
 985this is to configure the devices for DHCP (this is only to get the
 986file ifcfg-id file created; see below for some issues with DHCP).  The
 987name of the configuration file for each device will be of the form:
 988
 989ifcfg-id-xx:xx:xx:xx:xx:xx
 990
 991        Where the "xx" portion will be replaced with the digits from
 992the device's permanent MAC address.
 993
 994        Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
 995created, it is necessary to edit the configuration files for the slave
 996devices (the MAC addresses correspond to those of the slave devices).
 997Before editing, the file will contain multiple lines, and will look
 998something like this:
 999
1000BOOTPROTO='dhcp'
1001STARTMODE='on'
1002USERCTL='no'
1003UNIQUE='XNzu.WeZGOGF+4wE'
1004_nm_name='bus-pci-0001:61:01.0'
1005
1006        Change the BOOTPROTO and STARTMODE lines to the following:
1007
1008BOOTPROTO='none'
1009STARTMODE='off'
1010
1011        Do not alter the UNIQUE or _nm_name lines.  Remove any other
1012lines (USERCTL, etc).
1013
1014        Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
1015it's time to create the configuration file for the bonding device
1016itself.  This file is named ifcfg-bondX, where X is the number of the
1017bonding device to create, starting at 0.  The first such file is
1018ifcfg-bond0, the second is ifcfg-bond1, and so on.  The sysconfig
1019network configuration system will correctly start multiple instances
1020of bonding.
1021
1022        The contents of the ifcfg-bondX file is as follows:
1023
1024BOOTPROTO="static"
1025BROADCAST="10.0.2.255"
1026IPADDR="10.0.2.10"
1027NETMASK="255.255.0.0"
1028NETWORK="10.0.2.0"
1029REMOTE_IPADDR=""
1030STARTMODE="onboot"
1031BONDING_MASTER="yes"
1032BONDING_MODULE_OPTS="mode=active-backup miimon=100"
1033BONDING_SLAVE0="eth0"
1034BONDING_SLAVE1="bus-pci-0000:06:08.1"
1035
1036        Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
1037values with the appropriate values for your network.
1038
1039        The STARTMODE specifies when the device is brought online.
1040The possible values are:
1041
1042        onboot:  The device is started at boot time.  If you're not
1043                 sure, this is probably what you want.
1044
1045        manual:  The device is started only when ifup is called
1046                 manually.  Bonding devices may be configured this
1047                 way if you do not wish them to start automatically
1048                 at boot for some reason.
1049
1050        hotplug: The device is started by a hotplug event.  This is not
1051                 a valid choice for a bonding device.
1052
1053        off or ignore: The device configuration is ignored.
1054
1055        The line BONDING_MASTER='yes' indicates that the device is a
1056bonding master device.  The only useful value is "yes."
1057
1058        The contents of BONDING_MODULE_OPTS are supplied to the
1059instance of the bonding module for this device.  Specify the options
1060for the bonding mode, link monitoring, and so on here.  Do not include
1061the max_bonds bonding parameter; this will confuse the configuration
1062system if you have multiple bonding devices.
1063
1064        Finally, supply one BONDING_SLAVEn="slave device" for each
1065slave.  where "n" is an increasing value, one for each slave.  The
1066"slave device" is either an interface name, e.g., "eth0", or a device
1067specifier for the network device.  The interface name is easier to
1068find, but the ethN names are subject to change at boot time if, e.g.,
1069a device early in the sequence has failed.  The device specifiers
1070(bus-pci-0000:06:08.1 in the example above) specify the physical
1071network device, and will not change unless the device's bus location
1072changes (for example, it is moved from one PCI slot to another).  The
1073example above uses one of each type for demonstration purposes; most
1074configurations will choose one or the other for all slave devices.
1075
1076        When all configuration files have been modified or created,
1077networking must be restarted for the configuration changes to take
1078effect.  This can be accomplished via the following:
1079
1080# /etc/init.d/network restart
1081
1082        Note that the network control script (/sbin/ifdown) will
1083remove the bonding module as part of the network shutdown processing,
1084so it is not necessary to remove the module by hand if, e.g., the
1085module parameters have changed.
1086
1087        Also, at this writing, YaST/YaST2 will not manage bonding
1088devices (they do not show bonding interfaces on its list of network
1089devices).  It is necessary to edit the configuration file by hand to
1090change the bonding configuration.
1091
1092        Additional general options and details of the ifcfg file
1093format can be found in an example ifcfg template file:
1094
1095/etc/sysconfig/network/ifcfg.template
1096
1097        Note that the template does not document the various BONDING_
1098settings described above, but does describe many of the other options.
1099
11003.1.1 Using DHCP with Sysconfig
1101-------------------------------
1102
1103        Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1104will cause it to query DHCP for its IP address information.  At this
1105writing, this does not function for bonding devices; the scripts
1106attempt to obtain the device address from DHCP prior to adding any of
1107the slave devices.  Without active slaves, the DHCP requests are not
1108sent to the network.
1109
11103.1.2 Configuring Multiple Bonds with Sysconfig
1111-----------------------------------------------
1112
1113        The sysconfig network initialization system is capable of
1114handling multiple bonding devices.  All that is necessary is for each
1115bonding instance to have an appropriately configured ifcfg-bondX file
1116(as described above).  Do not specify the "max_bonds" parameter to any
1117instance of bonding, as this will confuse sysconfig.  If you require
1118multiple bonding devices with identical parameters, create multiple
1119ifcfg-bondX files.
1120
1121        Because the sysconfig scripts supply the bonding module
1122options in the ifcfg-bondX file, it is not necessary to add them to
1123the system /etc/modules.d/*.conf configuration files.
1124
11253.2 Configuration with Initscripts Support
1126------------------------------------------
1127
1128        This section applies to distros using a recent version of
1129initscripts with bonding support, for example, Red Hat Enterprise Linux
1130version 3 or later, Fedora, etc.  On these systems, the network
1131initialization scripts have knowledge of bonding, and can be configured to
1132control bonding devices.  Note that older versions of the initscripts
1133package have lower levels of support for bonding; this will be noted where
1134applicable.
1135
1136        These distros will not automatically load the network adapter
1137driver unless the ethX device is configured with an IP address.
1138Because of this constraint, users must manually configure a
1139network-script file for all physical adapters that will be members of
1140a bondX link.  Network script files are located in the directory:
1141
1142/etc/sysconfig/network-scripts
1143
1144        The file name must be prefixed with "ifcfg-eth" and suffixed
1145with the adapter's physical adapter number.  For example, the script
1146for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1147Place the following text in the file:
1148
1149DEVICE=eth0
1150USERCTL=no
1151ONBOOT=yes
1152MASTER=bond0
1153SLAVE=yes
1154BOOTPROTO=none
1155
1156        The DEVICE= line will be different for every ethX device and
1157must correspond with the name of the file, i.e., ifcfg-eth1 must have
1158a device line of DEVICE=eth1.  The setting of the MASTER= line will
1159also depend on the final bonding interface name chosen for your bond.
1160As with other network devices, these typically start at 0, and go up
1161one for each device, i.e., the first bonding instance is bond0, the
1162second is bond1, and so on.
1163
1164        Next, create a bond network script.  The file name for this
1165script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1166the number of the bond.  For bond0 the file is named "ifcfg-bond0",
1167for bond1 it is named "ifcfg-bond1", and so on.  Within that file,
1168place the following text:
1169
1170DEVICE=bond0
1171IPADDR=192.168.1.1
1172NETMASK=255.255.255.0
1173NETWORK=192.168.1.0
1174BROADCAST=192.168.1.255
1175ONBOOT=yes
1176BOOTPROTO=none
1177USERCTL=no
1178
1179        Be sure to change the networking specific lines (IPADDR,
1180NETMASK, NETWORK and BROADCAST) to match your network configuration.
1181
1182        For later versions of initscripts, such as that found with Fedora
11837 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1184and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
1185file, e.g. a line of the format:
1186
1187BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
1188
1189        will configure the bond with the specified options.  The options
1190specified in BONDING_OPTS are identical to the bonding module parameters
1191except for the arp_ip_target field when using versions of initscripts older
1192than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2).  When
1193using older versions each target should be included as a separate option and
1194should be preceded by a '+' to indicate it should be added to the list of
1195queried targets, e.g.,
1196
1197        arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1198
1199        is the proper syntax to specify multiple targets.  When specifying
1200options via BONDING_OPTS, it is not necessary to edit /etc/modprobe.d/*.conf.
1201
1202        For even older versions of initscripts that do not support
1203BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1204your distro) to load the bonding module with your desired options when the
1205bond0 interface is brought up.  The following lines in /etc/modprobe.d/*.conf
1206will load the bonding module, and select its options:
1207
1208alias bond0 bonding
1209options bond0 mode=balance-alb miimon=100
1210
1211        Replace the sample parameters with the appropriate set of
1212options for your configuration.
1213
1214        Finally run "/etc/rc.d/init.d/network restart" as root.  This
1215will restart the networking subsystem and your bond link should be now
1216up and running.
1217
12183.2.1 Using DHCP with Initscripts
1219---------------------------------
1220
1221        Recent versions of initscripts (the versions supplied with Fedora
1222Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1223work) have support for assigning IP information to bonding devices via
1224DHCP.
1225
1226        To configure bonding for DHCP, configure it as described
1227above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1228and add a line consisting of "TYPE=Bonding".  Note that the TYPE value
1229is case sensitive.
1230
12313.2.2 Configuring Multiple Bonds with Initscripts
1232-------------------------------------------------
1233
1234        Initscripts packages that are included with Fedora 7 and Red Hat
1235Enterprise Linux 5 support multiple bonding interfaces by simply
1236specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1237number of the bond.  This support requires sysfs support in the kernel,
1238and a bonding driver of version 3.0.0 or later.  Other configurations may
1239not support this method for specifying multiple bonding interfaces; for
1240those instances, see the "Configuring Multiple Bonds Manually" section,
1241below.
1242
12433.3 Configuring Bonding Manually with iproute2
1244-----------------------------------------------
1245
1246        This section applies to distros whose network initialization
1247scripts (the sysconfig or initscripts package) do not have specific
1248knowledge of bonding.  One such distro is SuSE Linux Enterprise Server
1249version 8.
1250
1251        The general method for these systems is to place the bonding
1252module parameters into a config file in /etc/modprobe.d/ (as
1253appropriate for the installed distro), then add modprobe and/or
1254`ip link` commands to the system's global init script.  The name of
1255the global init script differs; for sysconfig, it is
1256/etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1257
1258        For example, if you wanted to make a simple bond of two e100
1259devices (presumed to be eth0 and eth1), and have it persist across
1260reboots, edit the appropriate file (/etc/init.d/boot.local or
1261/etc/rc.d/rc.local), and add the following:
1262
1263modprobe bonding mode=balance-alb miimon=100
1264modprobe e100
1265ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1266ip link set eth0 master bond0
1267ip link set eth1 master bond0
1268
1269        Replace the example bonding module parameters and bond0
1270network configuration (IP address, netmask, etc) with the appropriate
1271values for your configuration.
1272
1273        Unfortunately, this method will not provide support for the
1274ifup and ifdown scripts on the bond devices.  To reload the bonding
1275configuration, it is necessary to run the initialization script, e.g.,
1276
1277# /etc/init.d/boot.local
1278
1279        or
1280
1281# /etc/rc.d/rc.local
1282
1283        It may be desirable in such a case to create a separate script
1284which only initializes the bonding configuration, then call that
1285separate script from within boot.local.  This allows for bonding to be
1286enabled without re-running the entire global init script.
1287
1288        To shut down the bonding devices, it is necessary to first
1289mark the bonding device itself as being down, then remove the
1290appropriate device driver modules.  For our example above, you can do
1291the following:
1292
1293# ifconfig bond0 down
1294# rmmod bonding
1295# rmmod e100
1296
1297        Again, for convenience, it may be desirable to create a script
1298with these commands.
1299
1300
13013.3.1 Configuring Multiple Bonds Manually
1302-----------------------------------------
1303
1304        This section contains information on configuring multiple
1305bonding devices with differing options for those systems whose network
1306initialization scripts lack support for configuring multiple bonds.
1307
1308        If you require multiple bonding devices, but all with the same
1309options, you may wish to use the "max_bonds" module parameter,
1310documented above.
1311
1312        To create multiple bonding devices with differing options, it is
1313preferable to use bonding parameters exported by sysfs, documented in the
1314section below.
1315
1316        For versions of bonding without sysfs support, the only means to
1317provide multiple instances of bonding with differing options is to load
1318the bonding driver multiple times.  Note that current versions of the
1319sysconfig network initialization scripts handle this automatically; if
1320your distro uses these scripts, no special action is needed.  See the
1321section Configuring Bonding Devices, above, if you're not sure about your
1322network initialization scripts.
1323
1324        To load multiple instances of the module, it is necessary to
1325specify a different name for each instance (the module loading system
1326requires that every loaded module, even multiple instances of the same
1327module, have a unique name).  This is accomplished by supplying multiple
1328sets of bonding options in /etc/modprobe.d/*.conf, for example:
1329
1330alias bond0 bonding
1331options bond0 -o bond0 mode=balance-rr miimon=100
1332
1333alias bond1 bonding
1334options bond1 -o bond1 mode=balance-alb miimon=50
1335
1336        will load the bonding module two times.  The first instance is
1337named "bond0" and creates the bond0 device in balance-rr mode with an
1338miimon of 100.  The second instance is named "bond1" and creates the
1339bond1 device in balance-alb mode with an miimon of 50.
1340
1341        In some circumstances (typically with older distributions),
1342the above does not work, and the second bonding instance never sees
1343its options.  In that case, the second options line can be substituted
1344as follows:
1345
1346install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1347        mode=balance-alb miimon=50
1348
1349        This may be repeated any number of times, specifying a new and
1350unique name in place of bond1 for each subsequent instance.
1351
1352        It has been observed that some Red Hat supplied kernels are unable
1353to rename modules at load time (the "-o bond1" part).  Attempts to pass
1354that option to modprobe will produce an "Operation not permitted" error.
1355This has been reported on some Fedora Core kernels, and has been seen on
1356RHEL 4 as well.  On kernels exhibiting this problem, it will be impossible
1357to configure multiple bonds with differing parameters (as they are older
1358kernels, and also lack sysfs support).
1359
13603.4 Configuring Bonding Manually via Sysfs
1361------------------------------------------
1362
1363        Starting with version 3.0.0, Channel Bonding may be configured
1364via the sysfs interface.  This interface allows dynamic configuration
1365of all bonds in the system without unloading the module.  It also
1366allows for adding and removing bonds at runtime.  Ifenslave is no
1367longer required, though it is still supported.
1368
1369        Use of the sysfs interface allows you to use multiple bonds
1370with different configurations without having to reload the module.
1371It also allows you to use multiple, differently configured bonds when
1372bonding is compiled into the kernel.
1373
1374        You must have the sysfs filesystem mounted to configure
1375bonding this way.  The examples in this document assume that you
1376are using the standard mount point for sysfs, e.g. /sys.  If your
1377sysfs filesystem is mounted elsewhere, you will need to adjust the
1378example paths accordingly.
1379
1380Creating and Destroying Bonds
1381-----------------------------
1382To add a new bond foo:
1383# echo +foo > /sys/class/net/bonding_masters
1384
1385To remove an existing bond bar:
1386# echo -bar > /sys/class/net/bonding_masters
1387
1388To show all existing bonds:
1389# cat /sys/class/net/bonding_masters
1390
1391NOTE: due to 4K size limitation of sysfs files, this list may be
1392truncated if you have more than a few hundred bonds.  This is unlikely
1393to occur under normal operating conditions.
1394
1395Adding and Removing Slaves
1396--------------------------
1397        Interfaces may be enslaved to a bond using the file
1398/sys/class/net/<bond>/bonding/slaves.  The semantics for this file
1399are the same as for the bonding_masters file.
1400
1401To enslave interface eth0 to bond bond0:
1402# ifconfig bond0 up
1403# echo +eth0 > /sys/class/net/bond0/bonding/slaves
1404
1405To free slave eth0 from bond bond0:
1406# echo -eth0 > /sys/class/net/bond0/bonding/slaves
1407
1408        When an interface is enslaved to a bond, symlinks between the
1409two are created in the sysfs filesystem.  In this case, you would get
1410/sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1411/sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1412
1413        This means that you can tell quickly whether or not an
1414interface is enslaved by looking for the master symlink.  Thus:
1415# echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1416will free eth0 from whatever bond it is enslaved to, regardless of
1417the name of the bond interface.
1418
1419Changing a Bond's Configuration
1420-------------------------------
1421        Each bond may be configured individually by manipulating the
1422files located in /sys/class/net/<bond name>/bonding
1423
1424        The names of these files correspond directly with the command-
1425line parameters described elsewhere in this file, and, with the
1426exception of arp_ip_target, they accept the same values.  To see the
1427current setting, simply cat the appropriate file.
1428
1429        A few examples will be given here; for specific usage
1430guidelines for each parameter, see the appropriate section in this
1431document.
1432
1433To configure bond0 for balance-alb mode:
1434# ifconfig bond0 down
1435# echo 6 > /sys/class/net/bond0/bonding/mode
1436 - or -
1437# echo balance-alb > /sys/class/net/bond0/bonding/mode
1438        NOTE: The bond interface must be down before the mode can be
1439changed.
1440
1441To enable MII monitoring on bond0 with a 1 second interval:
1442# echo 1000 > /sys/class/net/bond0/bonding/miimon
1443        NOTE: If ARP monitoring is enabled, it will disabled when MII
1444monitoring is enabled, and vice-versa.
1445
1446To add ARP targets:
1447# echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1448# echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1449        NOTE:  up to 16 target addresses may be specified.
1450
1451To remove an ARP target:
1452# echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1453
1454To configure the interval between learning packet transmits:
1455# echo 12 > /sys/class/net/bond0/bonding/lp_interval
1456        NOTE: the lp_inteval is the number of seconds between instances where
1457the bonding driver sends learning packets to each slaves peer switch.  The
1458default interval is 1 second.
1459
1460Example Configuration
1461---------------------
1462        We begin with the same example that is shown in section 3.3,
1463executed with sysfs, and without using ifenslave.
1464
1465        To make a simple bond of two e100 devices (presumed to be eth0
1466and eth1), and have it persist across reboots, edit the appropriate
1467file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1468following:
1469
1470modprobe bonding
1471modprobe e100
1472echo balance-alb > /sys/class/net/bond0/bonding/mode
1473ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1474echo 100 > /sys/class/net/bond0/bonding/miimon
1475echo +eth0 > /sys/class/net/bond0/bonding/slaves
1476echo +eth1 > /sys/class/net/bond0/bonding/slaves
1477
1478        To add a second bond, with two e1000 interfaces in
1479active-backup mode, using ARP monitoring, add the following lines to
1480your init script:
1481
1482modprobe e1000
1483echo +bond1 > /sys/class/net/bonding_masters
1484echo active-backup > /sys/class/net/bond1/bonding/mode
1485ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1486echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1487echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1488echo +eth2 > /sys/class/net/bond1/bonding/slaves
1489echo +eth3 > /sys/class/net/bond1/bonding/slaves
1490
14913.5 Configuration with Interfaces Support
1492-----------------------------------------
1493
1494        This section applies to distros which use /etc/network/interfaces file
1495to describe network interface configuration, most notably Debian and it's
1496derivatives.
1497
1498        The ifup and ifdown commands on Debian don't support bonding out of
1499the box. The ifenslave-2.6 package should be installed to provide bonding
1500support.  Once installed, this package will provide bond-* options to be used
1501into /etc/network/interfaces.
1502
1503        Note that ifenslave-2.6 package will load the bonding module and use
1504the ifenslave command when appropriate.
1505
1506Example Configurations
1507----------------------
1508
1509In /etc/network/interfaces, the following stanza will configure bond0, in
1510active-backup mode, with eth0 and eth1 as slaves.
1511
1512auto bond0
1513iface bond0 inet dhcp
1514        bond-slaves eth0 eth1
1515        bond-mode active-backup
1516        bond-miimon 100
1517        bond-primary eth0 eth1
1518
1519If the above configuration doesn't work, you might have a system using
1520upstart for system startup. This is most notably true for recent
1521Ubuntu versions. The following stanza in /etc/network/interfaces will
1522produce the same result on those systems.
1523
1524auto bond0
1525iface bond0 inet dhcp
1526        bond-slaves none
1527        bond-mode active-backup
1528        bond-miimon 100
1529
1530auto eth0
1531iface eth0 inet manual
1532        bond-master bond0
1533        bond-primary eth0 eth1
1534
1535auto eth1
1536iface eth1 inet manual
1537        bond-master bond0
1538        bond-primary eth0 eth1
1539
1540For a full list of bond-* supported options in /etc/network/interfaces and some
1541more advanced examples tailored to you particular distros, see the files in
1542/usr/share/doc/ifenslave-2.6.
1543
15443.6 Overriding Configuration for Special Cases
1545----------------------------------------------
1546
1547When using the bonding driver, the physical port which transmits a frame is
1548typically selected by the bonding driver, and is not relevant to the user or
1549system administrator.  The output port is simply selected using the policies of
1550the selected bonding mode.  On occasion however, it is helpful to direct certain
1551classes of traffic to certain physical interfaces on output to implement
1552slightly more complex policies.  For example, to reach a web server over a
1553bonded interface in which eth0 connects to a private network, while eth1
1554connects via a public network, it may be desirous to bias the bond to send said
1555traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1556can safely be sent over either interface.  Such configurations may be achieved
1557using the traffic control utilities inherent in linux.
1558
1559By default the bonding driver is multiqueue aware and 16 queues are created
1560when the driver initializes (see Documentation/networking/multiqueue.txt
1561for details).  If more or less queues are desired the module parameter
1562tx_queues can be used to change this value.  There is no sysfs parameter
1563available as the allocation is done at module init time.
1564
1565The output of the file /proc/net/bonding/bondX has changed so the output Queue
1566ID is now printed for each slave:
1567
1568Bonding Mode: fault-tolerance (active-backup)
1569Primary Slave: None
1570Currently Active Slave: eth0
1571MII Status: up
1572MII Polling Interval (ms): 0
1573Up Delay (ms): 0
1574Down Delay (ms): 0
1575
1576Slave Interface: eth0
1577MII Status: up
1578Link Failure Count: 0
1579Permanent HW addr: 00:1a:a0:12:8f:cb
1580Slave queue ID: 0
1581
1582Slave Interface: eth1
1583MII Status: up
1584Link Failure Count: 0
1585Permanent HW addr: 00:1a:a0:12:8f:cc
1586Slave queue ID: 2
1587
1588The queue_id for a slave can be set using the command:
1589
1590# echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1591
1592Any interface that needs a queue_id set should set it with multiple calls
1593like the one above until proper priorities are set for all interfaces.  On
1594distributions that allow configuration via initscripts, multiple 'queue_id'
1595arguments can be added to BONDING_OPTS to set all needed slave queues.
1596
1597These queue id's can be used in conjunction with the tc utility to configure
1598a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1599slave devices.  For instance, say we wanted, in the above configuration to
1600force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1601device. The following commands would accomplish this:
1602
1603# tc qdisc add dev bond0 handle 1 root multiq
1604
1605# tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip dst \
1606        192.168.1.100 action skbedit queue_mapping 2
1607
1608These commands tell the kernel to attach a multiqueue queue discipline to the
1609bond0 interface and filter traffic enqueued to it, such that packets with a dst
1610ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1611This value is then passed into the driver, causing the normal output path
1612selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1613
1614Note that qid values begin at 1.  Qid 0 is reserved to initiate to the driver
1615that normal output policy selection should take place.  One benefit to simply
1616leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1617driver that is now present.  This awareness allows tc filters to be placed on
1618slave devices as well as bond devices and the bonding driver will simply act as
1619a pass-through for selecting output queues on the slave device rather than 
1620output port selection.
1621
1622This feature first appeared in bonding driver version 3.7.0 and support for
1623output slave selection was limited to round-robin and active-backup modes.
1624
16254 Querying Bonding Configuration
1626=================================
1627
16284.1 Bonding Configuration
1629-------------------------
1630
1631        Each bonding device has a read-only file residing in the
1632/proc/net/bonding directory.  The file contents include information
1633about the bonding configuration, options and state of each slave.
1634
1635        For example, the contents of /proc/net/bonding/bond0 after the
1636driver is loaded with parameters of mode=0 and miimon=1000 is
1637generally as follows:
1638
1639        Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1640        Bonding Mode: load balancing (round-robin)
1641        Currently Active Slave: eth0
1642        MII Status: up
1643        MII Polling Interval (ms): 1000
1644        Up Delay (ms): 0
1645        Down Delay (ms): 0
1646
1647        Slave Interface: eth1
1648        MII Status: up
1649        Link Failure Count: 1
1650
1651        Slave Interface: eth0
1652        MII Status: up
1653        Link Failure Count: 1
1654
1655        The precise format and contents will change depending upon the
1656bonding configuration, state, and version of the bonding driver.
1657
16584.2 Network configuration
1659-------------------------
1660
1661        The network configuration can be inspected using the ifconfig
1662command.  Bonding devices will have the MASTER flag set; Bonding slave
1663devices will have the SLAVE flag set.  The ifconfig output does not
1664contain information on which slaves are associated with which masters.
1665
1666        In the example below, the bond0 interface is the master
1667(MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1668bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1669TLB and ALB that require a unique MAC address for each slave.
1670
1671# /sbin/ifconfig
1672bond0     Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1673          inet addr:XXX.XXX.XXX.YYY  Bcast:XXX.XXX.XXX.255  Mask:255.255.252.0
1674          UP BROADCAST RUNNING MASTER MULTICAST  MTU:1500  Metric:1
1675          RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1676          TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1677          collisions:0 txqueuelen:0
1678
1679eth0      Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1680          UP BROADCAST RUNNING SLAVE MULTICAST  MTU:1500  Metric:1
1681          RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1682          TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1683          collisions:0 txqueuelen:100
1684          Interrupt:10 Base address:0x1080
1685
1686eth1      Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1687          UP BROADCAST RUNNING SLAVE MULTICAST  MTU:1500  Metric:1
1688          RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1689          TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1690          collisions:0 txqueuelen:100
1691          Interrupt:9 Base address:0x1400
1692
16935. Switch Configuration
1694=======================
1695
1696        For this section, "switch" refers to whatever system the
1697bonded devices are directly connected to (i.e., where the other end of
1698the cable plugs into).  This may be an actual dedicated switch device,
1699or it may be another regular system (e.g., another computer running
1700Linux),
1701
1702        The active-backup, balance-tlb and balance-alb modes do not
1703require any specific configuration of the switch.
1704
1705        The 802.3ad mode requires that the switch have the appropriate
1706ports configured as an 802.3ad aggregation.  The precise method used
1707to configure this varies from switch to switch, but, for example, a
1708Cisco 3550 series switch requires that the appropriate ports first be
1709grouped together in a single etherchannel instance, then that
1710etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1711standard EtherChannel).
1712
1713        The balance-rr, balance-xor and broadcast modes generally
1714require that the switch have the appropriate ports grouped together.
1715The nomenclature for such a group differs between switches, it may be
1716called an "etherchannel" (as in the Cisco example, above), a "trunk
1717group" or some other similar variation.  For these modes, each switch
1718will also have its own configuration options for the switch's transmit
1719policy to the bond.  Typical choices include XOR of either the MAC or
1720IP addresses.  The transmit policy of the two peers does not need to
1721match.  For these three modes, the bonding mode really selects a
1722transmit policy for an EtherChannel group; all three will interoperate
1723with another EtherChannel group.
1724
1725
17266. 802.1q VLAN Support
1727======================
1728
1729        It is possible to configure VLAN devices over a bond interface
1730using the 8021q driver.  However, only packets coming from the 8021q
1731driver and passing through bonding will be tagged by default.  Self
1732generated packets, for example, bonding's learning packets or ARP
1733packets generated by either ALB mode or the ARP monitor mechanism, are
1734tagged internally by bonding itself.  As a result, bonding must
1735"learn" the VLAN IDs configured above it, and use those IDs to tag
1736self generated packets.
1737
1738        For reasons of simplicity, and to support the use of adapters
1739that can do VLAN hardware acceleration offloading, the bonding
1740interface declares itself as fully hardware offloading capable, it gets
1741the add_vid/kill_vid notifications to gather the necessary
1742information, and it propagates those actions to the slaves.  In case
1743of mixed adapter types, hardware accelerated tagged packets that
1744should go through an adapter that is not offloading capable are
1745"un-accelerated" by the bonding driver so the VLAN tag sits in the
1746regular location.
1747
1748        VLAN interfaces *must* be added on top of a bonding interface
1749only after enslaving at least one slave.  The bonding interface has a
1750hardware address of 00:00:00:00:00:00 until the first slave is added.
1751If the VLAN interface is created prior to the first enslavement, it
1752would pick up the all-zeroes hardware address.  Once the first slave
1753is attached to the bond, the bond device itself will pick up the
1754slave's hardware address, which is then available for the VLAN device.
1755
1756        Also, be aware that a similar problem can occur if all slaves
1757are released from a bond that still has one or more VLAN interfaces on
1758top of it.  When a new slave is added, the bonding interface will
1759obtain its hardware address from the first slave, which might not
1760match the hardware address of the VLAN interfaces (which was
1761ultimately copied from an earlier slave).
1762
1763        There are two methods to insure that the VLAN device operates
1764with the correct hardware address if all slaves are removed from a
1765bond interface:
1766
1767        1. Remove all VLAN interfaces then recreate them
1768
1769        2. Set the bonding interface's hardware address so that it
1770matches the hardware address of the VLAN interfaces.
1771
1772        Note that changing a VLAN interface's HW address would set the
1773underlying device -- i.e. the bonding interface -- to promiscuous
1774mode, which might not be what you want.
1775
1776
17777. Link Monitoring
1778==================
1779
1780        The bonding driver at present supports two schemes for
1781monitoring a slave device's link state: the ARP monitor and the MII
1782monitor.
1783
1784        At the present time, due to implementation restrictions in the
1785bonding driver itself, it is not possible to enable both ARP and MII
1786monitoring simultaneously.
1787
17887.1 ARP Monitor Operation
1789-------------------------
1790
1791        The ARP monitor operates as its name suggests: it sends ARP
1792queries to one or more designated peer systems on the network, and
1793uses the response as an indication that the link is operating.  This
1794gives some assurance that traffic is actually flowing to and from one
1795or more peers on the local network.
1796
1797        The ARP monitor relies on the device driver itself to verify
1798that traffic is flowing.  In particular, the driver must keep up to
1799date the last receive time, dev->last_rx, and transmit start time,
1800dev->trans_start.  If these are not updated by the driver, then the
1801ARP monitor will immediately fail any slaves using that driver, and
1802those slaves will stay down.  If networking monitoring (tcpdump, etc)
1803shows the ARP requests and replies on the network, then it may be that
1804your device driver is not updating last_rx and trans_start.
1805
18067.2 Configuring Multiple ARP Targets
1807------------------------------------
1808
1809        While ARP monitoring can be done with just one target, it can
1810be useful in a High Availability setup to have several targets to
1811monitor.  In the case of just one target, the target itself may go
1812down or have a problem making it unresponsive to ARP requests.  Having
1813an additional target (or several) increases the reliability of the ARP
1814monitoring.
1815
1816        Multiple ARP targets must be separated by commas as follows:
1817
1818# example options for ARP monitoring with three targets
1819alias bond0 bonding
1820options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1821
1822        For just a single target the options would resemble:
1823
1824# example options for ARP monitoring with one target
1825alias bond0 bonding
1826options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1827
1828
18297.3 MII Monitor Operation
1830-------------------------
1831
1832        The MII monitor monitors only the carrier state of the local
1833network interface.  It accomplishes this in one of three ways: by
1834depending upon the device driver to maintain its carrier state, by
1835querying the device's MII registers, or by making an ethtool query to
1836the device.
1837
1838        If the use_carrier module parameter is 1 (the default value),
1839then the MII monitor will rely on the driver for carrier state
1840information (via the netif_carrier subsystem).  As explained in the
1841use_carrier parameter information, above, if the MII monitor fails to
1842detect carrier loss on the device (e.g., when the cable is physically
1843disconnected), it may be that the driver does not support
1844netif_carrier.
1845
1846        If use_carrier is 0, then the MII monitor will first query the
1847device's (via ioctl) MII registers and check the link state.  If that
1848request fails (not just that it returns carrier down), then the MII
1849monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1850the same information.  If both methods fail (i.e., the driver either
1851does not support or had some error in processing both the MII register
1852and ethtool requests), then the MII monitor will assume the link is
1853up.
1854
18558. Potential Sources of Trouble
1856===============================
1857
18588.1 Adventures in Routing
1859-------------------------
1860
1861        When bonding is configured, it is important that the slave
1862devices not have routes that supersede routes of the master (or,
1863generally, not have routes at all).  For example, suppose the bonding
1864device bond0 has two slaves, eth0 and eth1, and the routing table is
1865as follows:
1866
1867Kernel IP routing table
1868Destination     Gateway         Genmask         Flags   MSS Window  irtt Iface
186910.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 eth0
187010.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 eth1
187110.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 bond0
1872127.0.0.0       0.0.0.0         255.0.0.0       U        40 0          0 lo
1873
1874        This routing configuration will likely still update the
1875receive/transmit times in the driver (needed by the ARP monitor), but
1876may bypass the bonding driver (because outgoing traffic to, in this
1877case, another host on network 10 would use eth0 or eth1 before bond0).
1878
1879        The ARP monitor (and ARP itself) may become confused by this
1880configuration, because ARP requests (generated by the ARP monitor)
1881will be sent on one interface (bond0), but the corresponding reply
1882will arrive on a different interface (eth0).  This reply looks to ARP
1883as an unsolicited ARP reply (because ARP matches replies on an
1884interface basis), and is discarded.  The MII monitor is not affected
1885by the state of the routing table.
1886
1887        The solution here is simply to insure that slaves do not have
1888routes of their own, and if for some reason they must, those routes do
1889not supersede routes of their master.  This should generally be the
1890case, but unusual configurations or errant manual or automatic static
1891route additions may cause trouble.
1892
18938.2 Ethernet Device Renaming
1894----------------------------
1895
1896        On systems with network configuration scripts that do not
1897associate physical devices directly with network interface names (so
1898that the same physical device always has the same "ethX" name), it may
1899be necessary to add some special logic to config files in
1900/etc/modprobe.d/.
1901
1902        For example, given a modules.conf containing the following:
1903
1904alias bond0 bonding
1905options bond0 mode=some-mode miimon=50
1906alias eth0 tg3
1907alias eth1 tg3
1908alias eth2 e1000
1909alias eth3 e1000
1910
1911        If neither eth0 and eth1 are slaves to bond0, then when the
1912bond0 interface comes up, the devices may end up reordered.  This
1913happens because bonding is loaded first, then its slave device's
1914drivers are loaded next.  Since no other drivers have been loaded,
1915when the e1000 driver loads, it will receive eth0 and eth1 for its
1916devices, but the bonding configuration tries to enslave eth2 and eth3
1917(which may later be assigned to the tg3 devices).
1918
1919        Adding the following:
1920
1921add above bonding e1000 tg3
1922
1923        causes modprobe to load e1000 then tg3, in that order, when
1924bonding is loaded.  This command is fully documented in the
1925modules.conf manual page.
1926
1927        On systems utilizing modprobe an equivalent problem can occur.
1928In this case, the following can be added to config files in
1929/etc/modprobe.d/ as:
1930
1931softdep bonding pre: tg3 e1000
1932
1933        This will load tg3 and e1000 modules before loading the bonding one.
1934Full documentation on this can be found in the modprobe.d and modprobe
1935manual pages.
1936
19378.3. Painfully Slow Or No Failed Link Detection By Miimon
1938---------------------------------------------------------
1939
1940        By default, bonding enables the use_carrier option, which
1941instructs bonding to trust the driver to maintain carrier state.
1942
1943        As discussed in the options section, above, some drivers do
1944not support the netif_carrier_on/_off link state tracking system.
1945With use_carrier enabled, bonding will always see these links as up,
1946regardless of their actual state.
1947
1948        Additionally, other drivers do support netif_carrier, but do
1949not maintain it in real time, e.g., only polling the link state at
1950some fixed interval.  In this case, miimon will detect failures, but
1951only after some long period of time has expired.  If it appears that
1952miimon is very slow in detecting link failures, try specifying
1953use_carrier=0 to see if that improves the failure detection time.  If
1954it does, then it may be that the driver checks the carrier state at a
1955fixed interval, but does not cache the MII register values (so the
1956use_carrier=0 method of querying the registers directly works).  If
1957use_carrier=0 does not improve the failover, then the driver may cache
1958the registers, or the problem may be elsewhere.
1959
1960        Also, remember that miimon only checks for the device's
1961carrier state.  It has no way to determine the state of devices on or
1962beyond other ports of a switch, or if a switch is refusing to pass
1963traffic while still maintaining carrier on.
1964
19659. SNMP agents
1966===============
1967
1968        If running SNMP agents, the bonding driver should be loaded
1969before any network drivers participating in a bond.  This requirement
1970is due to the interface index (ipAdEntIfIndex) being associated to
1971the first interface found with a given IP address.  That is, there is
1972only one ipAdEntIfIndex for each IP address.  For example, if eth0 and
1973eth1 are slaves of bond0 and the driver for eth0 is loaded before the
1974bonding driver, the interface for the IP address will be associated
1975with the eth0 interface.  This configuration is shown below, the IP
1976address 192.168.1.1 has an interface index of 2 which indexes to eth0
1977in the ifDescr table (ifDescr.2).
1978
1979     interfaces.ifTable.ifEntry.ifDescr.1 = lo
1980     interfaces.ifTable.ifEntry.ifDescr.2 = eth0
1981     interfaces.ifTable.ifEntry.ifDescr.3 = eth1
1982     interfaces.ifTable.ifEntry.ifDescr.4 = eth2
1983     interfaces.ifTable.ifEntry.ifDescr.5 = eth3
1984     interfaces.ifTable.ifEntry.ifDescr.6 = bond0
1985     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
1986     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1987     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
1988     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1989
1990        This problem is avoided by loading the bonding driver before
1991any network drivers participating in a bond.  Below is an example of
1992loading the bonding driver first, the IP address 192.168.1.1 is
1993correctly associated with ifDescr.2.
1994
1995     interfaces.ifTable.ifEntry.ifDescr.1 = lo
1996     interfaces.ifTable.ifEntry.ifDescr.2 = bond0
1997     interfaces.ifTable.ifEntry.ifDescr.3 = eth0
1998     interfaces.ifTable.ifEntry.ifDescr.4 = eth1
1999     interfaces.ifTable.ifEntry.ifDescr.5 = eth2
2000     interfaces.ifTable.ifEntry.ifDescr.6 = eth3
2001     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
2002     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
2003     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
2004     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
2005
2006        While some distributions may not report the interface name in
2007ifDescr, the association between the IP address and IfIndex remains
2008and SNMP functions such as Interface_Scan_Next will report that
2009association.
2010
201110. Promiscuous mode
2012====================
2013
2014        When running network monitoring tools, e.g., tcpdump, it is
2015common to enable promiscuous mode on the device, so that all traffic
2016is seen (instead of seeing only traffic destined for the local host).
2017The bonding driver handles promiscuous mode changes to the bonding
2018master device (e.g., bond0), and propagates the setting to the slave
2019devices.
2020
2021        For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
2022the promiscuous mode setting is propagated to all slaves.
2023
2024        For the active-backup, balance-tlb and balance-alb modes, the
2025promiscuous mode setting is propagated only to the active slave.
2026
2027        For balance-tlb mode, the active slave is the slave currently
2028receiving inbound traffic.
2029
2030        For balance-alb mode, the active slave is the slave used as a
2031"primary."  This slave is used for mode-specific control traffic, for
2032sending to peers that are unassigned or if the load is unbalanced.
2033
2034        For the active-backup, balance-tlb and balance-alb modes, when
2035the active slave changes (e.g., due to a link failure), the
2036promiscuous setting will be propagated to the new active slave.
2037
203811. Configuring Bonding for High Availability
2039=============================================
2040
2041        High Availability refers to configurations that provide
2042maximum network availability by having redundant or backup devices,
2043links or switches between the host and the rest of the world.  The
2044goal is to provide the maximum availability of network connectivity
2045(i.e., the network always works), even though other configurations
2046could provide higher throughput.
2047
204811.1 High Availability in a Single Switch Topology
2049--------------------------------------------------
2050
2051        If two hosts (or a host and a single switch) are directly
2052connected via multiple physical links, then there is no availability
2053penalty to optimizing for maximum bandwidth.  In this case, there is
2054only one switch (or peer), so if it fails, there is no alternative
2055access to fail over to.  Additionally, the bonding load balance modes
2056support link monitoring of their members, so if individual links fail,
2057the load will be rebalanced across the remaining devices.
2058
2059        See Section 12, "Configuring Bonding for Maximum Throughput"
2060for information on configuring bonding with one peer device.
2061
206211.2 High Availability in a Multiple Switch Topology
2063----------------------------------------------------
2064
2065        With multiple switches, the configuration of bonding and the
2066network changes dramatically.  In multiple switch topologies, there is
2067a trade off between network availability and usable bandwidth.
2068
2069        Below is a sample network, configured to maximize the
2070availability of the network:
2071
2072                |                                     |
2073                |port3                           port3|
2074          +-----+----+                          +-----+----+
2075          |          |port2       ISL      port2|          |
2076          | switch A +--------------------------+ switch B |
2077          |          |                          |          |
2078          +-----+----+                          +-----++---+
2079                |port1                           port1|
2080                |             +-------+               |
2081                +-------------+ host1 +---------------+
2082                         eth0 +-------+ eth1
2083
2084        In this configuration, there is a link between the two
2085switches (ISL, or inter switch link), and multiple ports connecting to
2086the outside world ("port3" on each switch).  There is no technical
2087reason that this could not be extended to a third switch.
2088
208911.2.1 HA Bonding Mode Selection for Multiple Switch Topology
2090-------------------------------------------------------------
2091
2092        In a topology such as the example above, the active-backup and
2093broadcast modes are the only useful bonding modes when optimizing for
2094availability; the other modes require all links to terminate on the
2095same peer for them to behave rationally.
2096
2097active-backup: This is generally the preferred mode, particularly if
2098        the switches have an ISL and play together well.  If the
2099        network configuration is such that one switch is specifically
2100        a backup switch (e.g., has lower capacity, higher cost, etc),
2101        then the primary option can be used to insure that the
2102        preferred link is always used when it is available.
2103
2104broadcast: This mode is really a special purpose mode, and is suitable
2105        only for very specific needs.  For example, if the two
2106        switches are not connected (no ISL), and the networks beyond
2107        them are totally independent.  In this case, if it is
2108        necessary for some specific one-way traffic to reach both
2109        independent networks, then the broadcast mode may be suitable.
2110
211111.2.2 HA Link Monitoring Selection for Multiple Switch Topology
2112----------------------------------------------------------------
2113
2114        The choice of link monitoring ultimately depends upon your
2115switch.  If the switch can reliably fail ports in response to other
2116failures, then either the MII or ARP monitors should work.  For
2117example, in the above example, if the "port3" link fails at the remote
2118end, the MII monitor has no direct means to detect this.  The ARP
2119monitor could be configured with a target at the remote end of port3,
2120thus detecting that failure without switch support.
2121
2122        In general, however, in a multiple switch topology, the ARP
2123monitor can provide a higher level of reliability in detecting end to
2124end connectivity failures (which may be caused by the failure of any
2125individual component to pass traffic for any reason).  Additionally,
2126the ARP monitor should be configured with multiple targets (at least
2127one for each switch in the network).  This will insure that,
2128regardless of which switch is active, the ARP monitor has a suitable
2129target to query.
2130
2131        Note, also, that of late many switches now support a functionality
2132generally referred to as "trunk failover."  This is a feature of the
2133switch that causes the link state of a particular switch port to be set
2134down (or up) when the state of another switch port goes down (or up).
2135Its purpose is to propagate link failures from logically "exterior" ports
2136to the logically "interior" ports that bonding is able to monitor via
2137miimon.  Availability and configuration for trunk failover varies by
2138switch, but this can be a viable alternative to the ARP monitor when using
2139suitable switches.
2140
214112. Configuring Bonding for Maximum Throughput
2142==============================================
2143
214412.1 Maximizing Throughput in a Single Switch Topology
2145------------------------------------------------------
2146
2147        In a single switch configuration, the best method to maximize
2148throughput depends upon the application and network environment.  The
2149various load balancing modes each have strengths and weaknesses in
2150different environments, as detailed below.
2151
2152        For this discussion, we will break down the topologies into
2153two categories.  Depending upon the destination of most traffic, we
2154categorize them into either "gatewayed" or "local" configurations.
2155
2156        In a gatewayed configuration, the "switch" is acting primarily
2157as a router, and the majority of traffic passes through this router to
2158other networks.  An example would be the following:
2159
2160
2161     +----------+                     +----------+
2162     |          |eth0            port1|          | to other networks
2163     | Host A   +---------------------+ router   +------------------->
2164     |          +---------------------+          | Hosts B and C are out
2165     |          |eth1            port2|          | here somewhere
2166     +----------+                     +----------+
2167
2168        The router may be a dedicated router device, or another host
2169acting as a gateway.  For our discussion, the important point is that
2170the majority of traffic from Host A will pass through the router to
2171some other network before reaching its final destination.
2172
2173        In a gatewayed network configuration, although Host A may
2174communicate with many other systems, all of its traffic will be sent
2175and received via one other peer on the local network, the router.
2176
2177        Note that the case of two systems connected directly via
2178multiple physical links is, for purposes of configuring bonding, the
2179same as a gatewayed configuration.  In that case, it happens that all
2180traffic is destined for the "gateway" itself, not some other network
2181beyond the gateway.
2182
2183        In a local configuration, the "switch" is acting primarily as
2184a switch, and the majority of traffic passes through this switch to
2185reach other stations on the same network.  An example would be the
2186following:
2187
2188    +----------+            +----------+       +--------+
2189    |          |eth0   port1|          +-------+ Host B |
2190    |  Host A  +------------+  switch  |port3  +--------+
2191    |          +------------+          |                  +--------+
2192    |          |eth1   port2|          +------------------+ Host C |
2193    +----------+            +----------+port4             +--------+
2194
2195
2196        Again, the switch may be a dedicated switch device, or another
2197host acting as a gateway.  For our discussion, the important point is
2198that the majority of traffic from Host A is destined for other hosts
2199on the same local network (Hosts B and C in the above example).
2200
2201        In summary, in a gatewayed configuration, traffic to and from
2202the bonded device will be to the same MAC level peer on the network
2203(the gateway itself, i.e., the router), regardless of its final
2204destination.  In a local configuration, traffic flows directly to and
2205from the final destinations, thus, each destination (Host B, Host C)
2206will be addressed directly by their individual MAC addresses.
2207
2208        This distinction between a gatewayed and a local network
2209configuration is important because many of the load balancing modes
2210available use the MAC addresses of the local network source and
2211destination to make load balancing decisions.  The behavior of each
2212mode is described below.
2213
2214
221512.1.1 MT Bonding Mode Selection for Single Switch Topology
2216-----------------------------------------------------------
2217
2218        This configuration is the easiest to set up and to understand,
2219although you will have to decide which bonding mode best suits your
2220needs.  The trade offs for each mode are detailed below:
2221
2222balance-rr: This mode is the only mode that will permit a single
2223        TCP/IP connection to stripe traffic across multiple
2224        interfaces. It is therefore the only mode that will allow a
2225        single TCP/IP stream to utilize more than one interface's
2226        worth of throughput.  This comes at a cost, however: the
2227        striping generally results in peer systems receiving packets out
2228        of order, causing TCP/IP's congestion control system to kick
2229        in, often by retransmitting segments.
2230
2231        It is possible to adjust TCP/IP's congestion limits by
2232        altering the net.ipv4.tcp_reordering sysctl parameter.  The
2233        usual default value is 3. But keep in mind TCP stack is able
2234        to automatically increase this when it detects reorders.
2235
2236        Note that the fraction of packets that will be delivered out of
2237        order is highly variable, and is unlikely to be zero.  The level
2238        of reordering depends upon a variety of factors, including the
2239        networking interfaces, the switch, and the topology of the
2240        configuration.  Speaking in general terms, higher speed network
2241        cards produce more reordering (due to factors such as packet
2242        coalescing), and a "many to many" topology will reorder at a
2243        higher rate than a "many slow to one fast" configuration.
2244
2245        Many switches do not support any modes that stripe traffic
2246        (instead choosing a port based upon IP or MAC level addresses);
2247        for those devices, traffic for a particular connection flowing
2248        through the switch to a balance-rr bond will not utilize greater
2249        than one interface's worth of bandwidth.
2250
2251        If you are utilizing protocols other than TCP/IP, UDP for
2252        example, and your application can tolerate out of order
2253        delivery, then this mode can allow for single stream datagram
2254        performance that scales near linearly as interfaces are added
2255        to the bond.
2256
2257        This mode requires the switch to have the appropriate ports
2258        configured for "etherchannel" or "trunking."
2259
2260active-backup: There is not much advantage in this network topology to
2261        the active-backup mode, as the inactive backup devices are all
2262        connected to the same peer as the primary.  In this case, a
2263        load balancing mode (with link monitoring) will provide the
2264        same level of network availability, but with increased
2265        available bandwidth.  On the plus side, active-backup mode
2266        does not require any configuration of the switch, so it may
2267        have value if the hardware available does not support any of
2268        the load balance modes.
2269
2270balance-xor: This mode will limit traffic such that packets destined
2271        for specific peers will always be sent over the same
2272        interface.  Since the destination is determined by the MAC
2273        addresses involved, this mode works best in a "local" network
2274        configuration (as described above), with destinations all on
2275        the same local network.  This mode is likely to be suboptimal
2276        if all your traffic is passed through a single router (i.e., a
2277        "gatewayed" network configuration, as described above).
2278
2279        As with balance-rr, the switch ports need to be configured for
2280        "etherchannel" or "trunking."
2281
2282broadcast: Like active-backup, there is not much advantage to this
2283        mode in this type of network topology.
2284
2285802.3ad: This mode can be a good choice for this type of network
2286        topology.  The 802.3ad mode is an IEEE standard, so all peers
2287        that implement 802.3ad should interoperate well.  The 802.3ad
2288        protocol includes automatic configuration of the aggregates,
2289        so minimal manual configuration of the switch is needed
2290        (typically only to designate that some set of devices is
2291        available for 802.3ad).  The 802.3ad standard also mandates
2292        that frames be delivered in order (within certain limits), so
2293        in general single connections will not see misordering of
2294        packets.  The 802.3ad mode does have some drawbacks: the
2295        standard mandates that all devices in the aggregate operate at
2296        the same speed and duplex.  Also, as with all bonding load
2297        balance modes other than balance-rr, no single connection will
2298        be able to utilize more than a single interface's worth of
2299        bandwidth.  
2300
2301        Additionally, the linux bonding 802.3ad implementation
2302        distributes traffic by peer (using an XOR of MAC addresses
2303        and packet type ID), so in a "gatewayed" configuration, all
2304        outgoing traffic will generally use the same device.  Incoming
2305        traffic may also end up on a single device, but that is
2306        dependent upon the balancing policy of the peer's 8023.ad
2307        implementation.  In a "local" configuration, traffic will be
2308        distributed across the devices in the bond.
2309
2310        Finally, the 802.3ad mode mandates the use of the MII monitor,
2311        therefore, the ARP monitor is not available in this mode.
2312
2313balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
2314        Since the balancing is done according to MAC address, in a
2315        "gatewayed" configuration (as described above), this mode will
2316        send all traffic across a single device.  However, in a
2317        "local" network configuration, this mode balances multiple
2318        local network peers across devices in a vaguely intelligent
2319        manner (not a simple XOR as in balance-xor or 802.3ad mode),
2320        so that mathematically unlucky MAC addresses (i.e., ones that
2321        XOR to the same value) will not all "bunch up" on a single
2322        interface.
2323
2324        Unlike 802.3ad, interfaces may be of differing speeds, and no
2325        special switch configuration is required.  On the down side,
2326        in this mode all incoming traffic arrives over a single
2327        interface, this mode requires certain ethtool support in the
2328        network device driver of the slave interfaces, and the ARP
2329        monitor is not available.
2330
2331balance-alb: This mode is everything that balance-tlb is, and more.
2332        It has all of the features (and restrictions) of balance-tlb,
2333        and will also balance incoming traffic from local network
2334        peers (as described in the Bonding Module Options section,
2335        above).
2336
2337        The only additional down side to this mode is that the network
2338        device driver must support changing the hardware address while
2339        the device is open.
2340
234112.1.2 MT Link Monitoring for Single Switch Topology
2342----------------------------------------------------
2343
2344        The choice of link monitoring may largely depend upon which
2345mode you choose to use.  The more advanced load balancing modes do not
2346support the use of the ARP monitor, and are thus restricted to using
2347the MII monitor (which does not provide as high a level of end to end
2348assurance as the ARP monitor).
2349
235012.2 Maximum Throughput in a Multiple Switch Topology
2351-----------------------------------------------------
2352
2353        Multiple switches may be utilized to optimize for throughput
2354when they are configured in parallel as part of an isolated network
2355between two or more systems, for example:
2356
2357                       +-----------+
2358                       |  Host A   | 
2359                       +-+---+---+-+
2360                         |   |   |
2361                +--------+   |   +---------+
2362                |            |             |
2363         +------+---+  +-----+----+  +-----+----+
2364         | Switch A |  | Switch B |  | Switch C |
2365         +------+---+  +-----+----+  +-----+----+
2366                |            |             |
2367                +--------+   |   +---------+
2368                         |   |   |
2369                       +-+---+---+-+
2370                       |  Host B   | 
2371                       +-----------+
2372
2373        In this configuration, the switches are isolated from one
2374another.  One reason to employ a topology such as this is for an
2375isolated network with many hosts (a cluster configured for high
2376performance, for example), using multiple smaller switches can be more
2377cost effective than a single larger switch, e.g., on a network with 24
2378hosts, three 24 port switches can be significantly less expensive than
2379a single 72 port switch.
2380
2381        If access beyond the network is required, an individual host
2382can be equipped with an additional network device connected to an
2383external network; this host then additionally acts as a gateway.
2384
238512.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2386-------------------------------------------------------------
2387
2388        In actual practice, the bonding mode typically employed in
2389configurations of this type is balance-rr.  Historically, in this
2390network configuration, the usual caveats about out of order packet
2391delivery are mitigated by the use of network adapters that do not do
2392any kind of packet coalescing (via the use of NAPI, or because the
2393device itself does not generate interrupts until some number of
2394packets has arrived).  When employed in this fashion, the balance-rr
2395mode allows individual connections between two hosts to effectively
2396utilize greater than one interface's bandwidth.
2397
239812.2.2 MT Link Monitoring for Multiple Switch Topology
2399------------------------------------------------------
2400
2401        Again, in actual practice, the MII monitor is most often used
2402in this configuration, as performance is given preference over
2403availability.  The ARP monitor will function in this topology, but its
2404advantages over the MII monitor are mitigated by the volume of probes
2405needed as the number of systems involved grows (remember that each
2406host in the network is configured with bonding).
2407
240813. Switch Behavior Issues
2409==========================
2410
241113.1 Link Establishment and Failover Delays
2412-------------------------------------------
2413
2414        Some switches exhibit undesirable behavior with regard to the
2415timing of link up and down reporting by the switch.
2416
2417        First, when a link comes up, some switches may indicate that
2418the link is up (carrier available), but not pass traffic over the
2419interface for some period of time.  This delay is typically due to
2420some type of autonegotiation or routing protocol, but may also occur
2421during switch initialization (e.g., during recovery after a switch
2422failure).  If you find this to be a problem, specify an appropriate
2423value to the updelay bonding module option to delay the use of the
2424relevant interface(s).
2425
2426        Second, some switches may "bounce" the link state one or more
2427times while a link is changing state.  This occurs most commonly while
2428the switch is initializing.  Again, an appropriate updelay value may
2429help.
2430
2431        Note that when a bonding interface has no active links, the
2432driver will immediately reuse the first link that goes up, even if the
2433updelay parameter has been specified (the updelay is ignored in this
2434case).  If there are slave interfaces waiting for the updelay timeout
2435to expire, the interface that first went into that state will be
2436immediately reused.  This reduces down time of the network if the
2437value of updelay has been overestimated, and since this occurs only in
2438cases with no connectivity, there is no additional penalty for
2439ignoring the updelay.
2440
2441        In addition to the concerns about switch timings, if your
2442switches take a long time to go into backup mode, it may be desirable
2443to not activate a backup interface immediately after a link goes down.
2444Failover may be delayed via the downdelay bonding module option.
2445
244613.2 Duplicated Incoming Packets
2447--------------------------------
2448
2449        NOTE: Starting with version 3.0.2, the bonding driver has logic to
2450suppress duplicate packets, which should largely eliminate this problem.
2451The following description is kept for reference.
2452
2453        It is not uncommon to observe a short burst of duplicated
2454traffic when the bonding device is first used, or after it has been
2455idle for some period of time.  This is most easily observed by issuing
2456a "ping" to some other host on the network, and noticing that the
2457output from ping flags duplicates (typically one per slave).
2458
2459        For example, on a bond in active-backup mode with five slaves
2460all connected to one switch, the output may appear as follows:
2461
2462# ping -n 10.0.4.2
2463PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
246464 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
246564 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
246664 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
246764 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
246864 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
246964 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
247064 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
247164 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2472
2473        This is not due to an error in the bonding driver, rather, it
2474is a side effect of how many switches update their MAC forwarding
2475tables.  Initially, the switch does not associate the MAC address in
2476the packet with a particular switch port, and so it may send the
2477traffic to all ports until its MAC forwarding table is updated.  Since
2478the interfaces attached to the bond may occupy multiple ports on a
2479single switch, when the switch (temporarily) floods the traffic to all
2480ports, the bond device receives multiple copies of the same packet
2481(one per slave device).
2482
2483        The duplicated packet behavior is switch dependent, some
2484switches exhibit this, and some do not.  On switches that display this
2485behavior, it can be induced by clearing the MAC forwarding table (on
2486most Cisco switches, the privileged command "clear mac address-table
2487dynamic" will accomplish this).
2488
248914. Hardware Specific Considerations
2490====================================
2491
2492        This section contains additional information for configuring
2493bonding on specific hardware platforms, or for interfacing bonding
2494with particular switches or other devices.
2495
249614.1 IBM BladeCenter
2497--------------------
2498
2499        This applies to the JS20 and similar systems.
2500
2501        On the JS20 blades, the bonding driver supports only
2502balance-rr, active-backup, balance-tlb and balance-alb modes.  This is
2503largely due to the network topology inside the BladeCenter, detailed
2504below.
2505
2506JS20 network adapter information
2507--------------------------------
2508
2509        All JS20s come with two Broadcom Gigabit Ethernet ports
2510integrated on the planar (that's "motherboard" in IBM-speak).  In the
2511BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2512I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2513An add-on Broadcom daughter card can be installed on a JS20 to provide
2514two more Gigabit Ethernet ports.  These ports, eth2 and eth3, are
2515wired to I/O Modules 3 and 4, respectively.
2516
2517        Each I/O Module may contain either a switch or a passthrough
2518module (which allows ports to be directly connected to an external
2519switch).  Some bonding modes require a specific BladeCenter internal
2520network topology in order to function; these are detailed below.
2521
2522        Additional BladeCenter-specific networking information can be
2523found in two IBM Redbooks (www.ibm.com/redbooks):
2524
2525"IBM eServer BladeCenter Networking Options"
2526"IBM eServer BladeCenter Layer 2-7 Network Switching"
2527
2528BladeCenter networking configuration
2529------------------------------------
2530
2531        Because a BladeCenter can be configured in a very large number
2532of ways, this discussion will be confined to describing basic
2533configurations.
2534
2535        Normally, Ethernet Switch Modules (ESMs) are used in I/O
2536modules 1 and 2.  In this configuration, the eth0 and eth1 ports of a
2537JS20 will be connected to different internal switches (in the
2538respective I/O modules).
2539
2540        A passthrough module (OPM or CPM, optical or copper,
2541passthrough module) connects the I/O module directly to an external
2542switch.  By using PMs in I/O module #1 and #2, the eth0 and eth1
2543interfaces of a JS20 can be redirected to the outside world and
2544connected to a common external switch.
2545
2546        Depending upon the mix of ESMs and PMs, the network will
2547appear to bonding as either a single switch topology (all PMs) or as a
2548multiple switch topology (one or more ESMs, zero or more PMs).  It is
2549also possible to connect ESMs together, resulting in a configuration
2550much like the example in "High Availability in a Multiple Switch
2551Topology," above.
2552
2553Requirements for specific modes
2554-------------------------------
2555
2556        The balance-rr mode requires the use of passthrough modules
2557for devices in the bond, all connected to an common external switch.
2558That switch must be configured for "etherchannel" or "trunking" on the
2559appropriate ports, as is usual for balance-rr.
2560
2561        The balance-alb and balance-tlb modes will function with
2562either switch modules or passthrough modules (or a mix).  The only
2563specific requirement for these modes is that all network interfaces
2564must be able to reach all destinations for traffic sent over the
2565bonding device (i.e., the network must converge at some point outside
2566the BladeCenter).
2567
2568        The active-backup mode has no additional requirements.
2569
2570Link monitoring issues
2571----------------------
2572
2573        When an Ethernet Switch Module is in place, only the ARP
2574monitor will reliably detect link loss to an external switch.  This is
2575nothing unusual, but examination of the BladeCenter cabinet would
2576suggest that the "external" network ports are the ethernet ports for
2577the system, when it fact there is a switch between these "external"
2578ports and the devices on the JS20 system itself.  The MII monitor is
2579only able to detect link failures between the ESM and the JS20 system.
2580
2581        When a passthrough module is in place, the MII monitor does
2582detect failures to the "external" port, which is then directly
2583connected to the JS20 system.
2584
2585Other concerns
2586--------------
2587
2588        The Serial Over LAN (SoL) link is established over the primary
2589ethernet (eth0) only, therefore, any loss of link to eth0 will result
2590in losing your SoL connection.  It will not fail over with other
2591network traffic, as the SoL system is beyond the control of the
2592bonding driver.
2593
2594        It may be desirable to disable spanning tree on the switch
2595(either the internal Ethernet Switch Module, or an external switch) to
2596avoid fail-over delay issues when using bonding.
2597
2598        
259915. Frequently Asked Questions
2600==============================
2601
26021.  Is it SMP safe?
2603
2604        Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2605The new driver was designed to be SMP safe from the start.
2606
26072.  What type of cards will work with it?
2608
2609        Any Ethernet type cards (you can even mix cards - a Intel
2610EtherExpress PRO/100 and a 3com 3c905b, for example).  For most modes,
2611devices need not be of the same speed.
2612
2613        Starting with version 3.2.1, bonding also supports Infiniband
2614slaves in active-backup mode.
2615
26163.  How many bonding devices can I have?
2617
2618        There is no limit.
2619
26204.  How many slaves can a bonding device have?
2621
2622        This is limited only by the number of network interfaces Linux
2623supports and/or the number of network cards you can place in your
2624system.
2625
26265.  What happens when a slave link dies?
2627
2628        If link monitoring is enabled, then the failing device will be
2629disabled.  The active-backup mode will fail over to a backup link, and
2630other modes will ignore the failed link.  The link will continue to be
2631monitored, and should it recover, it will rejoin the bond (in whatever
2632manner is appropriate for the mode). See the sections on High
2633Availability and the documentation for each mode for additional
2634information.
2635        
2636        Link monitoring can be enabled via either the miimon or
2637arp_interval parameters (described in the module parameters section,
2638above).  In general, miimon monitors the carrier state as sensed by
2639the underlying network device, and the arp monitor (arp_interval)
2640monitors connectivity to another host on the local network.
2641
2642        If no link monitoring is configured, the bonding driver will
2643be unable to detect link failures, and will assume that all links are
2644always available.  This will likely result in lost packets, and a
2645resulting degradation of performance.  The precise performance loss
2646depends upon the bonding mode and network configuration.
2647
26486.  Can bonding be used for High Availability?
2649
2650        Yes.  See the section on High Availability for details.
2651
26527.  Which switches/systems does it work with?
2653
2654        The full answer to this depends upon the desired mode.
2655
2656        In the basic balance modes (balance-rr and balance-xor), it
2657works with any system that supports etherchannel (also called
2658trunking).  Most managed switches currently available have such
2659support, and many unmanaged switches as well.
2660
2661        The advanced balance modes (balance-tlb and balance-alb) do
2662not have special switch requirements, but do need device drivers that
2663support specific features (described in the appropriate section under
2664module parameters, above).
2665
2666        In 802.3ad mode, it works with systems that support IEEE
2667802.3ad Dynamic Link Aggregation.  Most managed and many unmanaged
2668switches currently available support 802.3ad.
2669
2670        The active-backup mode should work with any Layer-II switch.
2671
26728.  Where does a bonding device get its MAC address from?
2673
2674        When using slave devices that have fixed MAC addresses, or when
2675the fail_over_mac option is enabled, the bonding device's MAC address is
2676the MAC address of the active slave.
2677
2678        For other configurations, if not explicitly configured (with
2679ifconfig or ip link), the MAC address of the bonding device is taken from
2680its first slave device.  This MAC address is then passed to all following
2681slaves and remains persistent (even if the first slave is removed) until
2682the bonding device is brought down or reconfigured.
2683
2684        If you wish to change the MAC address, you can set it with
2685ifconfig or ip link:
2686
2687# ifconfig bond0 hw ether 00:11:22:33:44:55
2688
2689# ip link set bond0 address 66:77:88:99:aa:bb
2690
2691        The MAC address can be also changed by bringing down/up the
2692device and then changing its slaves (or their order):
2693
2694# ifconfig bond0 down ; modprobe -r bonding
2695# ifconfig bond0 .... up
2696# ifenslave bond0 eth...
2697
2698        This method will automatically take the address from the next
2699slave that is added.
2700
2701        To restore your slaves' MAC addresses, you need to detach them
2702from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2703then restore the MAC addresses that the slaves had before they were
2704enslaved.
2705
270616. Resources and Links
2707=======================
2708
2709        The latest version of the bonding driver can be found in the latest
2710version of the linux kernel, found on http://kernel.org
2711
2712        The latest version of this document can be found in the latest kernel
2713source (named Documentation/networking/bonding.txt).
2714
2715        Discussions regarding the usage of the bonding driver take place on the
2716bonding-devel mailing list, hosted at sourceforge.net. If you have questions or
2717problems, post them to the list.  The list address is:
2718
2719bonding-devel@lists.sourceforge.net
2720
2721        The administrative interface (to subscribe or unsubscribe) can
2722be found at:
2723
2724https://lists.sourceforge.net/lists/listinfo/bonding-devel
2725
2726        Discussions regarding the development of the bonding driver take place
2727on the main Linux network mailing list, hosted at vger.kernel.org. The list
2728address is:
2729
2730netdev@vger.kernel.org
2731
2732        The administrative interface (to subscribe or unsubscribe) can
2733be found at:
2734
2735http://vger.kernel.org/vger-lists.html#netdev
2736
2737Donald Becker's Ethernet Drivers and diag programs may be found at :
2738 - http://web.archive.org/web/*/http://www.scyld.com/network/ 
2739
2740You will also find a lot of information regarding Ethernet, NWay, MII,
2741etc. at www.scyld.com.
2742
2743-- END --
2744
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.