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 and the ifenslave user level control
 108program installed and ready for use. If your distro does not, or you
 109have need to compile bonding from source (e.g., configuring and
 110installing a mainline kernel from kernel.org), you'll need to perform
 111the following steps:
 112
 1131.1 Configure and build the kernel with bonding
 114-----------------------------------------------
 115
 116        The current version of the bonding driver is available in the
 117drivers/net/bonding subdirectory of the most recent kernel source
 118(which is available on http://kernel.org).  Most users "rolling their
 119own" will want to use the most recent kernel from kernel.org.
 120
 121        Configure kernel with "make menuconfig" (or "make xconfig" or
 122"make config"), then select "Bonding driver support" in the "Network
 123device support" section.  It is recommended that you configure the
 124driver as module since it is currently the only way to pass parameters
 125to the driver or configure more than one bonding device.
 126
 127        Build and install the new kernel and modules, then continue
 128below to install ifenslave.
 129
 1301.2 Install ifenslave Control Utility
 131-------------------------------------
 132
 133        The ifenslave user level control program is included in the
 134kernel source tree, in the file Documentation/networking/ifenslave.c.
 135It is generally recommended that you use the ifenslave that
 136corresponds to the kernel that you are using (either from the same
 137source tree or supplied with the distro), however, ifenslave
 138executables from older kernels should function (but features newer
 139than the ifenslave release are not supported).  Running an ifenslave
 140that is newer than the kernel is not supported, and may or may not
 141work.
 142
 143        To install ifenslave, do the following:
 144
 145# gcc -Wall -O -I/usr/src/linux/include ifenslave.c -o ifenslave
 146# cp ifenslave /sbin/ifenslave
 147
 148        If your kernel source is not in "/usr/src/linux," then replace
 149"/usr/src/linux/include" in the above with the location of your kernel
 150source include directory.
 151
 152        You may wish to back up any existing /sbin/ifenslave, or, for
 153testing or informal use, tag the ifenslave to the kernel version
 154(e.g., name the ifenslave executable /sbin/ifenslave-2.6.10).
 155
 156IMPORTANT NOTE:
 157
 158        If you omit the "-I" or specify an incorrect directory, you
 159may end up with an ifenslave that is incompatible with the kernel
 160you're trying to build it for.  Some distros (e.g., Red Hat from 7.1
 161onwards) do not have /usr/include/linux symbolically linked to the
 162default kernel source include directory.
 163
 164SECOND IMPORTANT NOTE:
 165        If you plan to configure bonding using sysfs or using the
 166/etc/network/interfaces file, you do not need to use ifenslave.
 167
 1682. Bonding Driver Options
 169=========================
 170
 171        Options for the bonding driver are supplied as parameters to the
 172bonding module at load time, or are specified via sysfs.
 173
 174        Module options may be given as command line arguments to the
 175insmod or modprobe command, but are usually specified in either the
 176/etc/modrobe.d/*.conf configuration files, or in a distro-specific
 177configuration file (some of which are detailed in the next section).
 178
 179        Details on bonding support for sysfs is provided in the
 180"Configuring Bonding Manually via Sysfs" section, below.
 181
 182        The available bonding driver parameters are listed below. If a
 183parameter is not specified the default value is used.  When initially
 184configuring a bond, it is recommended "tail -f /var/log/messages" be
 185run in a separate window to watch for bonding driver error messages.
 186
 187        It is critical that either the miimon or arp_interval and
 188arp_ip_target parameters be specified, otherwise serious network
 189degradation will occur during link failures.  Very few devices do not
 190support at least miimon, so there is really no reason not to use it.
 191
 192        Options with textual values will accept either the text name
 193or, for backwards compatibility, the option value.  E.g.,
 194"mode=802.3ad" and "mode=4" set the same mode.
 195
 196        The parameters are as follows:
 197
 198active_slave
 199
 200        Specifies the new active slave for modes that support it
 201        (active-backup, balance-alb and balance-tlb).  Possible values
 202        are the name of any currently enslaved interface, or an empty
 203        string.  If a name is given, the slave and its link must be up in order
 204        to be selected as the new active slave.  If an empty string is
 205        specified, the current active slave is cleared, and a new active
 206        slave is selected automatically.
 207
 208        Note that this is only available through the sysfs interface. No module
 209        parameter by this name exists.
 210
 211        The normal value of this option is the name of the currently
 212        active slave, or the empty string if there is no active slave or
 213        the current mode does not use an active slave.
 214
 215ad_select
 216
 217        Specifies the 802.3ad aggregation selection logic to use.  The
 218        possible values and their effects are:
 219
 220        stable or 0
 221
 222                The active aggregator is chosen by largest aggregate
 223                bandwidth.
 224
 225                Reselection of the active aggregator occurs only when all
 226                slaves of the active aggregator are down or the active
 227                aggregator has no slaves.
 228
 229                This is the default value.
 230
 231        bandwidth or 1
 232
 233                The active aggregator is chosen by largest aggregate
 234                bandwidth.  Reselection occurs if:
 235
 236                - A slave is added to or removed from the bond
 237
 238                - Any slave's link state changes
 239
 240                - Any slave's 802.3ad association state changes
 241
 242                - The bond's administrative state changes to up
 243
 244        count or 2
 245
 246                The active aggregator is chosen by the largest number of
 247                ports (slaves).  Reselection occurs as described under the
 248                "bandwidth" setting, above.
 249
 250        The bandwidth and count selection policies permit failover of
 251        802.3ad aggregations when partial failure of the active aggregator
 252        occurs.  This keeps the aggregator with the highest availability
 253        (either in bandwidth or in number of ports) active at all times.
 254
 255        This option was added in bonding version 3.4.0.
 256
 257all_slaves_active
 258
 259        Specifies that duplicate frames (received on inactive ports) should be
 260        dropped (0) or delivered (1).
 261
 262        Normally, bonding will drop duplicate frames (received on inactive
 263        ports), which is desirable for most users. But there are some times
 264        it is nice to allow duplicate frames to be delivered.
 265
 266        The default value is 0 (drop duplicate frames received on inactive
 267        ports).
 268
 269arp_interval
 270
 271        Specifies the ARP link monitoring frequency in milliseconds.
 272
 273        The ARP monitor works by periodically checking the slave
 274        devices to determine whether they have sent or received
 275        traffic recently (the precise criteria depends upon the
 276        bonding mode, and the state of the slave).  Regular traffic is
 277        generated via ARP probes issued for the addresses specified by
 278        the arp_ip_target option.
 279
 280        This behavior can be modified by the arp_validate option,
 281        below.
 282
 283        If ARP monitoring is used in an etherchannel compatible mode
 284        (modes 0 and 2), the switch should be configured in a mode
 285        that evenly distributes packets across all links. If the
 286        switch is configured to distribute the packets in an XOR
 287        fashion, all replies from the ARP targets will be received on
 288        the same link which could cause the other team members to
 289        fail.  ARP monitoring should not be used in conjunction with
 290        miimon.  A value of 0 disables ARP monitoring.  The default
 291        value is 0.
 292
 293arp_ip_target
 294
 295        Specifies the IP addresses to use as ARP monitoring peers when
 296        arp_interval is > 0.  These are the targets of the ARP request
 297        sent to determine the health of the link to the targets.
 298        Specify these values in ddd.ddd.ddd.ddd format.  Multiple IP
 299        addresses must be separated by a comma.  At least one IP
 300        address must be given for ARP monitoring to function.  The
 301        maximum number of targets that can be specified is 16.  The
 302        default value is no IP addresses.
 303
 304arp_validate
 305
 306        Specifies whether or not ARP probes and replies should be
 307        validated in the active-backup mode.  This causes the ARP
 308        monitor to examine the incoming ARP requests and replies, and
 309        only consider a slave to be up if it is receiving the
 310        appropriate ARP traffic.
 311
 312        Possible values are:
 313
 314        none or 0
 315
 316                No validation is performed.  This is the default.
 317
 318        active or 1
 319
 320                Validation is performed only for the active slave.
 321
 322        backup or 2
 323
 324                Validation is performed only for backup slaves.
 325
 326        all or 3
 327
 328                Validation is performed for all slaves.
 329
 330        For the active slave, the validation checks ARP replies to
 331        confirm that they were generated by an arp_ip_target.  Since
 332        backup slaves do not typically receive these replies, the
 333        validation performed for backup slaves is on the ARP request
 334        sent out via the active slave.  It is possible that some
 335        switch or network configurations may result in situations
 336        wherein the backup slaves do not receive the ARP requests; in
 337        such a situation, validation of backup slaves must be
 338        disabled.
 339
 340        This option is useful in network configurations in which
 341        multiple bonding hosts are concurrently issuing ARPs to one or
 342        more targets beyond a common switch.  Should the link between
 343        the switch and target fail (but not the switch itself), the
 344        probe traffic generated by the multiple bonding instances will
 345        fool the standard ARP monitor into considering the links as
 346        still up.  Use of the arp_validate option can resolve this, as
 347        the ARP monitor will only consider ARP requests and replies
 348        associated with its own instance of bonding.
 349
 350        This option was added in bonding version 3.1.0.
 351
 352downdelay
 353
 354        Specifies the time, in milliseconds, to wait before disabling
 355        a slave after a link failure has been detected.  This option
 356        is only valid for the miimon link monitor.  The downdelay
 357        value should be a multiple of the miimon value; if not, it
 358        will be rounded down to the nearest multiple.  The default
 359        value is 0.
 360
 361fail_over_mac
 362
 363        Specifies whether active-backup mode should set all slaves to
 364        the same MAC address at enslavement (the traditional
 365        behavior), or, when enabled, perform special handling of the
 366        bond's MAC address in accordance with the selected policy.
 367
 368        Possible values are:
 369
 370        none or 0
 371
 372                This setting disables fail_over_mac, and causes
 373                bonding to set all slaves of an active-backup bond to
 374                the same MAC address at enslavement time.  This is the
 375                default.
 376
 377        active or 1
 378
 379                The "active" fail_over_mac policy indicates that the
 380                MAC address of the bond should always be the MAC
 381                address of the currently active slave.  The MAC
 382                address of the slaves is not changed; instead, the MAC
 383                address of the bond changes during a failover.
 384
 385                This policy is useful for devices that cannot ever
 386                alter their MAC address, or for devices that refuse
 387                incoming broadcasts with their own source MAC (which
 388                interferes with the ARP monitor).
 389
 390                The down side of this policy is that every device on
 391                the network must be updated via gratuitous ARP,
 392                vs. just updating a switch or set of switches (which
 393                often takes place for any traffic, not just ARP
 394                traffic, if the switch snoops incoming traffic to
 395                update its tables) for the traditional method.  If the
 396                gratuitous ARP is lost, communication may be
 397                disrupted.
 398
 399                When this policy is used in conjunction with the mii
 400                monitor, devices which assert link up prior to being
 401                able to actually transmit and receive are particularly
 402                susceptible to loss of the gratuitous ARP, and an
 403                appropriate updelay setting may be required.
 404
 405        follow or 2
 406
 407                The "follow" fail_over_mac policy causes the MAC
 408                address of the bond to be selected normally (normally
 409                the MAC address of the first slave added to the bond).
 410                However, the second and subsequent slaves are not set
 411                to this MAC address while they are in a backup role; a
 412                slave is programmed with the bond's MAC address at
 413                failover time (and the formerly active slave receives
 414                the newly active slave's MAC address).
 415
 416                This policy is useful for multiport devices that
 417                either become confused or incur a performance penalty
 418                when multiple ports are programmed with the same MAC
 419                address.
 420
 421
 422        The default policy is none, unless the first slave cannot
 423        change its MAC address, in which case the active policy is
 424        selected by default.
 425
 426        This option may be modified via sysfs only when no slaves are
 427        present in the bond.
 428
 429        This option was added in bonding version 3.2.0.  The "follow"
 430        policy was added in bonding version 3.3.0.
 431
 432lacp_rate
 433
 434        Option specifying the rate in which we'll ask our link partner
 435        to transmit LACPDU packets in 802.3ad mode.  Possible values
 436        are:
 437
 438        slow or 0
 439                Request partner to transmit LACPDUs every 30 seconds
 440
 441        fast or 1
 442                Request partner to transmit LACPDUs every 1 second
 443
 444        The default is slow.
 445
 446max_bonds
 447
 448        Specifies the number of bonding devices to create for this
 449        instance of the bonding driver.  E.g., if max_bonds is 3, and
 450        the bonding driver is not already loaded, then bond0, bond1
 451        and bond2 will be created.  The default value is 1.  Specifying
 452        a value of 0 will load bonding, but will not create any devices.
 453
 454miimon
 455
 456        Specifies the MII link monitoring frequency in milliseconds.
 457        This determines how often the link state of each slave is
 458        inspected for link failures.  A value of zero disables MII
 459        link monitoring.  A value of 100 is a good starting point.
 460        The use_carrier option, below, affects how the link state is
 461        determined.  See the High Availability section for additional
 462        information.  The default value is 0.
 463
 464min_links
 465
 466        Specifies the minimum number of links that must be active before
 467        asserting carrier. It is similar to the Cisco EtherChannel min-links
 468        feature. This allows setting the minimum number of member ports that
 469        must be up (link-up state) before marking the bond device as up
 470        (carrier on). This is useful for situations where higher level services
 471        such as clustering want to ensure a minimum number of low bandwidth
 472        links are active before switchover. This option only affect 802.3ad
 473        mode.
 474
 475        The default value is 0. This will cause carrier to be asserted (for
 476        802.3ad mode) whenever there is an active aggregator, regardless of the
 477        number of available links in that aggregator. Note that, because an
 478        aggregator cannot be active without at least one available link,
 479        setting this option to 0 or to 1 has the exact same effect.
 480
 481mode
 482
 483        Specifies one of the bonding policies. The default is
 484        balance-rr (round robin).  Possible values are:
 485
 486        balance-rr or 0
 487
 488                Round-robin policy: Transmit packets in sequential
 489                order from the first available slave through the
 490                last.  This mode provides load balancing and fault
 491                tolerance.
 492
 493        active-backup or 1
 494
 495                Active-backup policy: Only one slave in the bond is
 496                active.  A different slave becomes active if, and only
 497                if, the active slave fails.  The bond's MAC address is
 498                externally visible on only one port (network adapter)
 499                to avoid confusing the switch.
 500
 501                In bonding version 2.6.2 or later, when a failover
 502                occurs in active-backup mode, bonding will issue one
 503                or more gratuitous ARPs on the newly active slave.
 504                One gratuitous ARP is issued for the bonding master
 505                interface and each VLAN interfaces configured above
 506                it, provided that the interface has at least one IP
 507                address configured.  Gratuitous ARPs issued for VLAN
 508                interfaces are tagged with the appropriate VLAN id.
 509
 510                This mode provides fault tolerance.  The primary
 511                option, documented below, affects the behavior of this
 512                mode.
 513
 514        balance-xor or 2
 515
 516                XOR policy: Transmit based on the selected transmit
 517                hash policy.  The default policy is a simple [(source
 518                MAC address XOR'd with destination MAC address) modulo
 519                slave count].  Alternate transmit policies may be
 520                selected via the xmit_hash_policy option, described
 521                below.
 522
 523                This mode provides load balancing and fault tolerance.
 524
 525        broadcast or 3
 526
 527                Broadcast policy: transmits everything on all slave
 528                interfaces.  This mode provides fault tolerance.
 529
 530        802.3ad or 4
 531
 532                IEEE 802.3ad Dynamic link aggregation.  Creates
 533                aggregation groups that share the same speed and
 534                duplex settings.  Utilizes all slaves in the active
 535                aggregator according to the 802.3ad specification.
 536
 537                Slave selection for outgoing traffic is done according
 538                to the transmit hash policy, which may be changed from
 539                the default simple XOR policy via the xmit_hash_policy
 540                option, documented below.  Note that not all transmit
 541                policies may be 802.3ad compliant, particularly in
 542                regards to the packet mis-ordering requirements of
 543                section 43.2.4 of the 802.3ad standard.  Differing
 544                peer implementations will have varying tolerances for
 545                noncompliance.
 546
 547                Prerequisites:
 548
 549                1. Ethtool support in the base drivers for retrieving
 550                the speed and duplex of each slave.
 551
 552                2. A switch that supports IEEE 802.3ad Dynamic link
 553                aggregation.
 554
 555                Most switches will require some type of configuration
 556                to enable 802.3ad mode.
 557
 558        balance-tlb or 5
 559
 560                Adaptive transmit load balancing: channel bonding that
 561                does not require any special switch support.  The
 562                outgoing traffic is distributed according to the
 563                current load (computed relative to the speed) on each
 564                slave.  Incoming traffic is received by the current
 565                slave.  If the receiving slave fails, another slave
 566                takes over the MAC address of the failed receiving
 567                slave.
 568
 569                Prerequisite:
 570
 571                Ethtool support in the base drivers for retrieving the
 572                speed of each slave.
 573
 574        balance-alb or 6
 575
 576                Adaptive load balancing: includes balance-tlb plus
 577                receive load balancing (rlb) for IPV4 traffic, and
 578                does not require any special switch support.  The
 579                receive load balancing is achieved by ARP negotiation.
 580                The bonding driver intercepts the ARP Replies sent by
 581                the local system on their way out and overwrites the
 582                source hardware address with the unique hardware
 583                address of one of the slaves in the bond such that
 584                different peers use different hardware addresses for
 585                the server.
 586
 587                Receive traffic from connections created by the server
 588                is also balanced.  When the local system sends an ARP
 589                Request the bonding driver copies and saves the peer's
 590                IP information from the ARP packet.  When the ARP
 591                Reply arrives from the peer, its hardware address is
 592                retrieved and the bonding driver initiates an ARP
 593                reply to this peer assigning it to one of the slaves
 594                in the bond.  A problematic outcome of using ARP
 595                negotiation for balancing is that each time that an
 596                ARP request is broadcast it uses the hardware address
 597                of the bond.  Hence, peers learn the hardware address
 598                of the bond and the balancing of receive traffic
 599                collapses to the current slave.  This is handled by
 600                sending updates (ARP Replies) to all the peers with
 601                their individually assigned hardware address such that
 602                the traffic is redistributed.  Receive traffic is also
 603                redistributed when a new slave is added to the bond
 604                and when an inactive slave is re-activated.  The
 605                receive load is distributed sequentially (round robin)
 606                among the group of highest speed slaves in the bond.
 607
 608                When a link is reconnected or a new slave joins the
 609                bond the receive traffic is redistributed among all
 610                active slaves in the bond by initiating ARP Replies
 611                with the selected MAC address to each of the
 612                clients. The updelay parameter (detailed below) must
 613                be set to a value equal or greater than the switch's
 614                forwarding delay so that the ARP Replies sent to the
 615                peers will not be blocked by the switch.
 616
 617                Prerequisites:
 618
 619                1. Ethtool support in the base drivers for retrieving
 620                the speed of each slave.
 621
 622                2. Base driver support for setting the hardware
 623                address of a device while it is open.  This is
 624                required so that there will always be one slave in the
 625                team using the bond hardware address (the
 626                curr_active_slave) while having a unique hardware
 627                address for each slave in the bond.  If the
 628                curr_active_slave fails its hardware address is
 629                swapped with the new curr_active_slave that was
 630                chosen.
 631
 632num_grat_arp
 633num_unsol_na
 634
 635        Specify the number of peer notifications (gratuitous ARPs and
 636        unsolicited IPv6 Neighbor Advertisements) to be issued after a
 637        failover event.  As soon as the link is up on the new slave
 638        (possibly immediately) a peer notification is sent on the
 639        bonding device and each VLAN sub-device.  This is repeated at
 640        each link monitor interval (arp_interval or miimon, whichever
 641        is active) if the number is greater than 1.
 642
 643        The valid range is 0 - 255; the default value is 1.  These options
 644        affect only the active-backup mode.  These options were added for
 645        bonding versions 3.3.0 and 3.4.0 respectively.
 646
 647        From Linux 3.0 and bonding version 3.7.1, these notifications
 648        are generated by the ipv4 and ipv6 code and the numbers of
 649        repetitions cannot be set independently.
 650
 651primary
 652
 653        A string (eth0, eth2, etc) specifying which slave is the
 654        primary device.  The specified device will always be the
 655        active slave while it is available.  Only when the primary is
 656        off-line will alternate devices be used.  This is useful when
 657        one slave is preferred over another, e.g., when one slave has
 658        higher throughput than another.
 659
 660        The primary option is only valid for active-backup mode.
 661
 662primary_reselect
 663
 664        Specifies the reselection policy for the primary slave.  This
 665        affects how the primary slave is chosen to become the active slave
 666        when failure of the active slave or recovery of the primary slave
 667        occurs.  This option is designed to prevent flip-flopping between
 668        the primary slave and other slaves.  Possible values are:
 669
 670        always or 0 (default)
 671
 672                The primary slave becomes the active slave whenever it
 673                comes back up.
 674
 675        better or 1
 676
 677                The primary slave becomes the active slave when it comes
 678                back up, if the speed and duplex of the primary slave is
 679                better than the speed and duplex of the current active
 680                slave.
 681
 682        failure or 2
 683
 684                The primary slave becomes the active slave only if the
 685                current active slave fails and the primary slave is up.
 686
 687        The primary_reselect setting is ignored in two cases:
 688
 689                If no slaves are active, the first slave to recover is
 690                made the active slave.
 691
 692                When initially enslaved, the primary slave is always made
 693                the active slave.
 694
 695        Changing the primary_reselect policy via sysfs will cause an
 696        immediate selection of the best active slave according to the new
 697        policy.  This may or may not result in a change of the active
 698        slave, depending upon the circumstances.
 699
 700        This option was added for bonding version 3.6.0.
 701
 702updelay
 703
 704        Specifies the time, in milliseconds, to wait before enabling a
 705        slave after a link recovery has been detected.  This option is
 706        only valid for the miimon link monitor.  The updelay value
 707        should be a multiple of the miimon value; if not, it will be
 708        rounded down to the nearest multiple.  The default value is 0.
 709
 710use_carrier
 711
 712        Specifies whether or not miimon should use MII or ETHTOOL
 713        ioctls vs. netif_carrier_ok() to determine the link
 714        status. The MII or ETHTOOL ioctls are less efficient and
 715        utilize a deprecated calling sequence within the kernel.  The
 716        netif_carrier_ok() relies on the device driver to maintain its
 717        state with netif_carrier_on/off; at this writing, most, but
 718        not all, device drivers support this facility.
 719
 720        If bonding insists that the link is up when it should not be,
 721        it may be that your network device driver does not support
 722        netif_carrier_on/off.  The default state for netif_carrier is
 723        "carrier on," so if a driver does not support netif_carrier,
 724        it will appear as if the link is always up.  In this case,
 725        setting use_carrier to 0 will cause bonding to revert to the
 726        MII / ETHTOOL ioctl method to determine the link state.
 727
 728        A value of 1 enables the use of netif_carrier_ok(), a value of
 729        0 will use the deprecated MII / ETHTOOL ioctls.  The default
 730        value is 1.
 731
 732xmit_hash_policy
 733
 734        Selects the transmit hash policy to use for slave selection in
 735        balance-xor and 802.3ad modes.  Possible values are:
 736
 737        layer2
 738
 739                Uses XOR of hardware MAC addresses to generate the
 740                hash.  The formula is
 741
 742                (source MAC XOR destination MAC) modulo slave count
 743
 744                This algorithm will place all traffic to a particular
 745                network peer on the same slave.
 746
 747                This algorithm is 802.3ad compliant.
 748
 749        layer2+3
 750
 751                This policy uses a combination of layer2 and layer3
 752                protocol information to generate the hash.
 753
 754                Uses XOR of hardware MAC addresses and IP addresses to
 755                generate the hash.  The IPv4 formula is
 756
 757                (((source IP XOR dest IP) AND 0xffff) XOR
 758                        ( source MAC XOR destination MAC ))
 759                                modulo slave count
 760
 761                The IPv6 formula is
 762
 763                hash = (source ip quad 2 XOR dest IP quad 2) XOR
 764                       (source ip quad 3 XOR dest IP quad 3) XOR
 765                       (source ip quad 4 XOR dest IP quad 4)
 766
 767                (((hash >> 24) XOR (hash >> 16) XOR (hash >> 8) XOR hash)
 768                        XOR (source MAC XOR destination MAC))
 769                                modulo slave count
 770
 771                This algorithm will place all traffic to a particular
 772                network peer on the same slave.  For non-IP traffic,
 773                the formula is the same as for the layer2 transmit
 774                hash policy.
 775
 776                This policy is intended to provide a more balanced
 777                distribution of traffic than layer2 alone, especially
 778                in environments where a layer3 gateway device is
 779                required to reach most destinations.
 780
 781                This algorithm is 802.3ad compliant.
 782
 783        layer3+4
 784
 785                This policy uses upper layer protocol information,
 786                when available, to generate the hash.  This allows for
 787                traffic to a particular network peer to span multiple
 788                slaves, although a single connection will not span
 789                multiple slaves.
 790
 791                The formula for unfragmented IPv4 TCP and UDP packets is
 792
 793                ((source port XOR dest port) XOR
 794                         ((source IP XOR dest IP) AND 0xffff)
 795                                modulo slave count
 796
 797                The formula for unfragmented IPv6 TCP and UDP packets is
 798
 799                hash = (source port XOR dest port) XOR
 800                       ((source ip quad 2 XOR dest IP quad 2) XOR
 801                        (source ip quad 3 XOR dest IP quad 3) XOR
 802                        (source ip quad 4 XOR dest IP quad 4))
 803
 804                ((hash >> 24) XOR (hash >> 16) XOR (hash >> 8) XOR hash)
 805                        modulo slave count
 806
 807                For fragmented TCP or UDP packets and all other IPv4 and
 808                IPv6 protocol traffic, the source and destination port
 809                information is omitted.  For non-IP traffic, the
 810                formula is the same as for the layer2 transmit hash
 811                policy.
 812
 813                The IPv4 policy is intended to mimic the behavior of
 814                certain switches, notably Cisco switches with PFC2 as
 815                well as some Foundry and IBM products.
 816
 817                This algorithm is not fully 802.3ad compliant.  A
 818                single TCP or UDP conversation containing both
 819                fragmented and unfragmented packets will see packets
 820                striped across two interfaces.  This may result in out
 821                of order delivery.  Most traffic types will not meet
 822                this criteria, as TCP rarely fragments traffic, and
 823                most UDP traffic is not involved in extended
 824                conversations.  Other implementations of 802.3ad may
 825                or may not tolerate this noncompliance.
 826
 827        The default value is layer2.  This option was added in bonding
 828        version 2.6.3.  In earlier versions of bonding, this parameter
 829        does not exist, and the layer2 policy is the only policy.  The
 830        layer2+3 value was added for bonding version 3.2.2.
 831
 832resend_igmp
 833
 834        Specifies the number of IGMP membership reports to be issued after
 835        a failover event. One membership report is issued immediately after
 836        the failover, subsequent packets are sent in each 200ms interval.
 837
 838        The valid range is 0 - 255; the default value is 1. A value of 0
 839        prevents the IGMP membership report from being issued in response
 840        to the failover event.
 841
 842        This option is useful for bonding modes balance-rr (0), active-backup
 843        (1), balance-tlb (5) and balance-alb (6), in which a failover can
 844        switch the IGMP traffic from one slave to another.  Therefore a fresh
 845        IGMP report must be issued to cause the switch to forward the incoming
 846        IGMP traffic over the newly selected slave.
 847
 848        This option was added for bonding version 3.7.0.
 849
 8503. Configuring Bonding Devices
 851==============================
 852
 853        You can configure bonding using either your distro's network
 854initialization scripts, or manually using either ifenslave or the
 855sysfs interface.  Distros generally use one of three packages for the
 856network initialization scripts: initscripts, sysconfig or interfaces.
 857Recent versions of these packages have support for bonding, while older
 858versions do not.
 859
 860        We will first describe the options for configuring bonding for
 861distros using versions of initscripts, sysconfig and interfaces with full
 862or partial support for bonding, then provide information on enabling
 863bonding without support from the network initialization scripts (i.e.,
 864older versions of initscripts or sysconfig).
 865
 866        If you're unsure whether your distro uses sysconfig,
 867initscripts or interfaces, or don't know if it's new enough, have no fear.
 868Determining this is fairly straightforward.
 869
 870        First, look for a file called interfaces in /etc/network directory.
 871If this file is present in your system, then your system use interfaces. See
 872Configuration with Interfaces Support.
 873
 874        Else, issue the command:
 875
 876$ rpm -qf /sbin/ifup
 877
 878        It will respond with a line of text starting with either
 879"initscripts" or "sysconfig," followed by some numbers.  This is the
 880package that provides your network initialization scripts.
 881
 882        Next, to determine if your installation supports bonding,
 883issue the command:
 884
 885$ grep ifenslave /sbin/ifup
 886
 887        If this returns any matches, then your initscripts or
 888sysconfig has support for bonding.
 889
 8903.1 Configuration with Sysconfig Support
 891----------------------------------------
 892
 893        This section applies to distros using a version of sysconfig
 894with bonding support, for example, SuSE Linux Enterprise Server 9.
 895
 896        SuSE SLES 9's networking configuration system does support
 897bonding, however, at this writing, the YaST system configuration
 898front end does not provide any means to work with bonding devices.
 899Bonding devices can be managed by hand, however, as follows.
 900
 901        First, if they have not already been configured, configure the
 902slave devices.  On SLES 9, this is most easily done by running the
 903yast2 sysconfig configuration utility.  The goal is for to create an
 904ifcfg-id file for each slave device.  The simplest way to accomplish
 905this is to configure the devices for DHCP (this is only to get the
 906file ifcfg-id file created; see below for some issues with DHCP).  The
 907name of the configuration file for each device will be of the form:
 908
 909ifcfg-id-xx:xx:xx:xx:xx:xx
 910
 911        Where the "xx" portion will be replaced with the digits from
 912the device's permanent MAC address.
 913
 914        Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been
 915created, it is necessary to edit the configuration files for the slave
 916devices (the MAC addresses correspond to those of the slave devices).
 917Before editing, the file will contain multiple lines, and will look
 918something like this:
 919
 920BOOTPROTO='dhcp'
 921STARTMODE='on'
 922USERCTL='no'
 923UNIQUE='XNzu.WeZGOGF+4wE'
 924_nm_name='bus-pci-0001:61:01.0'
 925
 926        Change the BOOTPROTO and STARTMODE lines to the following:
 927
 928BOOTPROTO='none'
 929STARTMODE='off'
 930
 931        Do not alter the UNIQUE or _nm_name lines.  Remove any other
 932lines (USERCTL, etc).
 933
 934        Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified,
 935it's time to create the configuration file for the bonding device
 936itself.  This file is named ifcfg-bondX, where X is the number of the
 937bonding device to create, starting at 0.  The first such file is
 938ifcfg-bond0, the second is ifcfg-bond1, and so on.  The sysconfig
 939network configuration system will correctly start multiple instances
 940of bonding.
 941
 942        The contents of the ifcfg-bondX file is as follows:
 943
 944BOOTPROTO="static"
 945BROADCAST="10.0.2.255"
 946IPADDR="10.0.2.10"
 947NETMASK="255.255.0.0"
 948NETWORK="10.0.2.0"
 949REMOTE_IPADDR=""
 950STARTMODE="onboot"
 951BONDING_MASTER="yes"
 952BONDING_MODULE_OPTS="mode=active-backup miimon=100"
 953BONDING_SLAVE0="eth0"
 954BONDING_SLAVE1="bus-pci-0000:06:08.1"
 955
 956        Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK
 957values with the appropriate values for your network.
 958
 959        The STARTMODE specifies when the device is brought online.
 960The possible values are:
 961
 962        onboot:  The device is started at boot time.  If you're not
 963                 sure, this is probably what you want.
 964
 965        manual:  The device is started only when ifup is called
 966                 manually.  Bonding devices may be configured this
 967                 way if you do not wish them to start automatically
 968                 at boot for some reason.
 969
 970        hotplug: The device is started by a hotplug event.  This is not
 971                 a valid choice for a bonding device.
 972
 973        off or ignore: The device configuration is ignored.
 974
 975        The line BONDING_MASTER='yes' indicates that the device is a
 976bonding master device.  The only useful value is "yes."
 977
 978        The contents of BONDING_MODULE_OPTS are supplied to the
 979instance of the bonding module for this device.  Specify the options
 980for the bonding mode, link monitoring, and so on here.  Do not include
 981the max_bonds bonding parameter; this will confuse the configuration
 982system if you have multiple bonding devices.
 983
 984        Finally, supply one BONDING_SLAVEn="slave device" for each
 985slave.  where "n" is an increasing value, one for each slave.  The
 986"slave device" is either an interface name, e.g., "eth0", or a device
 987specifier for the network device.  The interface name is easier to
 988find, but the ethN names are subject to change at boot time if, e.g.,
 989a device early in the sequence has failed.  The device specifiers
 990(bus-pci-0000:06:08.1 in the example above) specify the physical
 991network device, and will not change unless the device's bus location
 992changes (for example, it is moved from one PCI slot to another).  The
 993example above uses one of each type for demonstration purposes; most
 994configurations will choose one or the other for all slave devices.
 995
 996        When all configuration files have been modified or created,
 997networking must be restarted for the configuration changes to take
 998effect.  This can be accomplished via the following:
 999
1000# /etc/init.d/network restart
1001
1002        Note that the network control script (/sbin/ifdown) will
1003remove the bonding module as part of the network shutdown processing,
1004so it is not necessary to remove the module by hand if, e.g., the
1005module parameters have changed.
1006
1007        Also, at this writing, YaST/YaST2 will not manage bonding
1008devices (they do not show bonding interfaces on its list of network
1009devices).  It is necessary to edit the configuration file by hand to
1010change the bonding configuration.
1011
1012        Additional general options and details of the ifcfg file
1013format can be found in an example ifcfg template file:
1014
1015/etc/sysconfig/network/ifcfg.template
1016
1017        Note that the template does not document the various BONDING_
1018settings described above, but does describe many of the other options.
1019
10203.1.1 Using DHCP with Sysconfig
1021-------------------------------
1022
1023        Under sysconfig, configuring a device with BOOTPROTO='dhcp'
1024will cause it to query DHCP for its IP address information.  At this
1025writing, this does not function for bonding devices; the scripts
1026attempt to obtain the device address from DHCP prior to adding any of
1027the slave devices.  Without active slaves, the DHCP requests are not
1028sent to the network.
1029
10303.1.2 Configuring Multiple Bonds with Sysconfig
1031-----------------------------------------------
1032
1033        The sysconfig network initialization system is capable of
1034handling multiple bonding devices.  All that is necessary is for each
1035bonding instance to have an appropriately configured ifcfg-bondX file
1036(as described above).  Do not specify the "max_bonds" parameter to any
1037instance of bonding, as this will confuse sysconfig.  If you require
1038multiple bonding devices with identical parameters, create multiple
1039ifcfg-bondX files.
1040
1041        Because the sysconfig scripts supply the bonding module
1042options in the ifcfg-bondX file, it is not necessary to add them to
1043the system /etc/modules.d/*.conf configuration files.
1044
10453.2 Configuration with Initscripts Support
1046------------------------------------------
1047
1048        This section applies to distros using a recent version of
1049initscripts with bonding support, for example, Red Hat Enterprise Linux
1050version 3 or later, Fedora, etc.  On these systems, the network
1051initialization scripts have knowledge of bonding, and can be configured to
1052control bonding devices.  Note that older versions of the initscripts
1053package have lower levels of support for bonding; this will be noted where
1054applicable.
1055
1056        These distros will not automatically load the network adapter
1057driver unless the ethX device is configured with an IP address.
1058Because of this constraint, users must manually configure a
1059network-script file for all physical adapters that will be members of
1060a bondX link.  Network script files are located in the directory:
1061
1062/etc/sysconfig/network-scripts
1063
1064        The file name must be prefixed with "ifcfg-eth" and suffixed
1065with the adapter's physical adapter number.  For example, the script
1066for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0.
1067Place the following text in the file:
1068
1069DEVICE=eth0
1070USERCTL=no
1071ONBOOT=yes
1072MASTER=bond0
1073SLAVE=yes
1074BOOTPROTO=none
1075
1076        The DEVICE= line will be different for every ethX device and
1077must correspond with the name of the file, i.e., ifcfg-eth1 must have
1078a device line of DEVICE=eth1.  The setting of the MASTER= line will
1079also depend on the final bonding interface name chosen for your bond.
1080As with other network devices, these typically start at 0, and go up
1081one for each device, i.e., the first bonding instance is bond0, the
1082second is bond1, and so on.
1083
1084        Next, create a bond network script.  The file name for this
1085script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is
1086the number of the bond.  For bond0 the file is named "ifcfg-bond0",
1087for bond1 it is named "ifcfg-bond1", and so on.  Within that file,
1088place the following text:
1089
1090DEVICE=bond0
1091IPADDR=192.168.1.1
1092NETMASK=255.255.255.0
1093NETWORK=192.168.1.0
1094BROADCAST=192.168.1.255
1095ONBOOT=yes
1096BOOTPROTO=none
1097USERCTL=no
1098
1099        Be sure to change the networking specific lines (IPADDR,
1100NETMASK, NETWORK and BROADCAST) to match your network configuration.
1101
1102        For later versions of initscripts, such as that found with Fedora
11037 (or later) and Red Hat Enterprise Linux version 5 (or later), it is possible,
1104and, indeed, preferable, to specify the bonding options in the ifcfg-bond0
1105file, e.g. a line of the format:
1106
1107BONDING_OPTS="mode=active-backup arp_interval=60 arp_ip_target=192.168.1.254"
1108
1109        will configure the bond with the specified options.  The options
1110specified in BONDING_OPTS are identical to the bonding module parameters
1111except for the arp_ip_target field when using versions of initscripts older
1112than and 8.57 (Fedora 8) and 8.45.19 (Red Hat Enterprise Linux 5.2).  When
1113using older versions each target should be included as a separate option and
1114should be preceded by a '+' to indicate it should be added to the list of
1115queried targets, e.g.,
1116
1117        arp_ip_target=+192.168.1.1 arp_ip_target=+192.168.1.2
1118
1119        is the proper syntax to specify multiple targets.  When specifying
1120options via BONDING_OPTS, it is not necessary to edit /etc/modprobe.d/*.conf.
1121
1122        For even older versions of initscripts that do not support
1123BONDING_OPTS, it is necessary to edit /etc/modprobe.d/*.conf, depending upon
1124your distro) to load the bonding module with your desired options when the
1125bond0 interface is brought up.  The following lines in /etc/modprobe.d/*.conf
1126will load the bonding module, and select its options:
1127
1128alias bond0 bonding
1129options bond0 mode=balance-alb miimon=100
1130
1131        Replace the sample parameters with the appropriate set of
1132options for your configuration.
1133
1134        Finally run "/etc/rc.d/init.d/network restart" as root.  This
1135will restart the networking subsystem and your bond link should be now
1136up and running.
1137
11383.2.1 Using DHCP with Initscripts
1139---------------------------------
1140
1141        Recent versions of initscripts (the versions supplied with Fedora
1142Core 3 and Red Hat Enterprise Linux 4, or later versions, are reported to
1143work) have support for assigning IP information to bonding devices via
1144DHCP.
1145
1146        To configure bonding for DHCP, configure it as described
1147above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp"
1148and add a line consisting of "TYPE=Bonding".  Note that the TYPE value
1149is case sensitive.
1150
11513.2.2 Configuring Multiple Bonds with Initscripts
1152-------------------------------------------------
1153
1154        Initscripts packages that are included with Fedora 7 and Red Hat
1155Enterprise Linux 5 support multiple bonding interfaces by simply
1156specifying the appropriate BONDING_OPTS= in ifcfg-bondX where X is the
1157number of the bond.  This support requires sysfs support in the kernel,
1158and a bonding driver of version 3.0.0 or later.  Other configurations may
1159not support this method for specifying multiple bonding interfaces; for
1160those instances, see the "Configuring Multiple Bonds Manually" section,
1161below.
1162
11633.3 Configuring Bonding Manually with Ifenslave
1164-----------------------------------------------
1165
1166        This section applies to distros whose network initialization
1167scripts (the sysconfig or initscripts package) do not have specific
1168knowledge of bonding.  One such distro is SuSE Linux Enterprise Server
1169version 8.
1170
1171        The general method for these systems is to place the bonding
1172module parameters into a config file in /etc/modprobe.d/ (as
1173appropriate for the installed distro), then add modprobe and/or
1174ifenslave commands to the system's global init script.  The name of
1175the global init script differs; for sysconfig, it is
1176/etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local.
1177
1178        For example, if you wanted to make a simple bond of two e100
1179devices (presumed to be eth0 and eth1), and have it persist across
1180reboots, edit the appropriate file (/etc/init.d/boot.local or
1181/etc/rc.d/rc.local), and add the following:
1182
1183modprobe bonding mode=balance-alb miimon=100
1184modprobe e100
1185ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1186ifenslave bond0 eth0
1187ifenslave bond0 eth1
1188
1189        Replace the example bonding module parameters and bond0
1190network configuration (IP address, netmask, etc) with the appropriate
1191values for your configuration.
1192
1193        Unfortunately, this method will not provide support for the
1194ifup and ifdown scripts on the bond devices.  To reload the bonding
1195configuration, it is necessary to run the initialization script, e.g.,
1196
1197# /etc/init.d/boot.local
1198
1199        or
1200
1201# /etc/rc.d/rc.local
1202
1203        It may be desirable in such a case to create a separate script
1204which only initializes the bonding configuration, then call that
1205separate script from within boot.local.  This allows for bonding to be
1206enabled without re-running the entire global init script.
1207
1208        To shut down the bonding devices, it is necessary to first
1209mark the bonding device itself as being down, then remove the
1210appropriate device driver modules.  For our example above, you can do
1211the following:
1212
1213# ifconfig bond0 down
1214# rmmod bonding
1215# rmmod e100
1216
1217        Again, for convenience, it may be desirable to create a script
1218with these commands.
1219
1220
12213.3.1 Configuring Multiple Bonds Manually
1222-----------------------------------------
1223
1224        This section contains information on configuring multiple
1225bonding devices with differing options for those systems whose network
1226initialization scripts lack support for configuring multiple bonds.
1227
1228        If you require multiple bonding devices, but all with the same
1229options, you may wish to use the "max_bonds" module parameter,
1230documented above.
1231
1232        To create multiple bonding devices with differing options, it is
1233preferable to use bonding parameters exported by sysfs, documented in the
1234section below.
1235
1236        For versions of bonding without sysfs support, the only means to
1237provide multiple instances of bonding with differing options is to load
1238the bonding driver multiple times.  Note that current versions of the
1239sysconfig network initialization scripts handle this automatically; if
1240your distro uses these scripts, no special action is needed.  See the
1241section Configuring Bonding Devices, above, if you're not sure about your
1242network initialization scripts.
1243
1244        To load multiple instances of the module, it is necessary to
1245specify a different name for each instance (the module loading system
1246requires that every loaded module, even multiple instances of the same
1247module, have a unique name).  This is accomplished by supplying multiple
1248sets of bonding options in /etc/modprobe.d/*.conf, for example:
1249
1250alias bond0 bonding
1251options bond0 -o bond0 mode=balance-rr miimon=100
1252
1253alias bond1 bonding
1254options bond1 -o bond1 mode=balance-alb miimon=50
1255
1256        will load the bonding module two times.  The first instance is
1257named "bond0" and creates the bond0 device in balance-rr mode with an
1258miimon of 100.  The second instance is named "bond1" and creates the
1259bond1 device in balance-alb mode with an miimon of 50.
1260
1261        In some circumstances (typically with older distributions),
1262the above does not work, and the second bonding instance never sees
1263its options.  In that case, the second options line can be substituted
1264as follows:
1265
1266install bond1 /sbin/modprobe --ignore-install bonding -o bond1 \
1267        mode=balance-alb miimon=50
1268
1269        This may be repeated any number of times, specifying a new and
1270unique name in place of bond1 for each subsequent instance.
1271
1272        It has been observed that some Red Hat supplied kernels are unable
1273to rename modules at load time (the "-o bond1" part).  Attempts to pass
1274that option to modprobe will produce an "Operation not permitted" error.
1275This has been reported on some Fedora Core kernels, and has been seen on
1276RHEL 4 as well.  On kernels exhibiting this problem, it will be impossible
1277to configure multiple bonds with differing parameters (as they are older
1278kernels, and also lack sysfs support).
1279
12803.4 Configuring Bonding Manually via Sysfs
1281------------------------------------------
1282
1283        Starting with version 3.0.0, Channel Bonding may be configured
1284via the sysfs interface.  This interface allows dynamic configuration
1285of all bonds in the system without unloading the module.  It also
1286allows for adding and removing bonds at runtime.  Ifenslave is no
1287longer required, though it is still supported.
1288
1289        Use of the sysfs interface allows you to use multiple bonds
1290with different configurations without having to reload the module.
1291It also allows you to use multiple, differently configured bonds when
1292bonding is compiled into the kernel.
1293
1294        You must have the sysfs filesystem mounted to configure
1295bonding this way.  The examples in this document assume that you
1296are using the standard mount point for sysfs, e.g. /sys.  If your
1297sysfs filesystem is mounted elsewhere, you will need to adjust the
1298example paths accordingly.
1299
1300Creating and Destroying Bonds
1301-----------------------------
1302To add a new bond foo:
1303# echo +foo > /sys/class/net/bonding_masters
1304
1305To remove an existing bond bar:
1306# echo -bar > /sys/class/net/bonding_masters
1307
1308To show all existing bonds:
1309# cat /sys/class/net/bonding_masters
1310
1311NOTE: due to 4K size limitation of sysfs files, this list may be
1312truncated if you have more than a few hundred bonds.  This is unlikely
1313to occur under normal operating conditions.
1314
1315Adding and Removing Slaves
1316--------------------------
1317        Interfaces may be enslaved to a bond using the file
1318/sys/class/net/<bond>/bonding/slaves.  The semantics for this file
1319are the same as for the bonding_masters file.
1320
1321To enslave interface eth0 to bond bond0:
1322# ifconfig bond0 up
1323# echo +eth0 > /sys/class/net/bond0/bonding/slaves
1324
1325To free slave eth0 from bond bond0:
1326# echo -eth0 > /sys/class/net/bond0/bonding/slaves
1327
1328        When an interface is enslaved to a bond, symlinks between the
1329two are created in the sysfs filesystem.  In this case, you would get
1330/sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and
1331/sys/class/net/eth0/master pointing to /sys/class/net/bond0.
1332
1333        This means that you can tell quickly whether or not an
1334interface is enslaved by looking for the master symlink.  Thus:
1335# echo -eth0 > /sys/class/net/eth0/master/bonding/slaves
1336will free eth0 from whatever bond it is enslaved to, regardless of
1337the name of the bond interface.
1338
1339Changing a Bond's Configuration
1340-------------------------------
1341        Each bond may be configured individually by manipulating the
1342files located in /sys/class/net/<bond name>/bonding
1343
1344        The names of these files correspond directly with the command-
1345line parameters described elsewhere in this file, and, with the
1346exception of arp_ip_target, they accept the same values.  To see the
1347current setting, simply cat the appropriate file.
1348
1349        A few examples will be given here; for specific usage
1350guidelines for each parameter, see the appropriate section in this
1351document.
1352
1353To configure bond0 for balance-alb mode:
1354# ifconfig bond0 down
1355# echo 6 > /sys/class/net/bond0/bonding/mode
1356 - or -
1357# echo balance-alb > /sys/class/net/bond0/bonding/mode
1358        NOTE: The bond interface must be down before the mode can be
1359changed.
1360
1361To enable MII monitoring on bond0 with a 1 second interval:
1362# echo 1000 > /sys/class/net/bond0/bonding/miimon
1363        NOTE: If ARP monitoring is enabled, it will disabled when MII
1364monitoring is enabled, and vice-versa.
1365
1366To add ARP targets:
1367# echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1368# echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target
1369        NOTE:  up to 16 target addresses may be specified.
1370
1371To remove an ARP target:
1372# echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target
1373
1374Example Configuration
1375---------------------
1376        We begin with the same example that is shown in section 3.3,
1377executed with sysfs, and without using ifenslave.
1378
1379        To make a simple bond of two e100 devices (presumed to be eth0
1380and eth1), and have it persist across reboots, edit the appropriate
1381file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the
1382following:
1383
1384modprobe bonding
1385modprobe e100
1386echo balance-alb > /sys/class/net/bond0/bonding/mode
1387ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up
1388echo 100 > /sys/class/net/bond0/bonding/miimon
1389echo +eth0 > /sys/class/net/bond0/bonding/slaves
1390echo +eth1 > /sys/class/net/bond0/bonding/slaves
1391
1392        To add a second bond, with two e1000 interfaces in
1393active-backup mode, using ARP monitoring, add the following lines to
1394your init script:
1395
1396modprobe e1000
1397echo +bond1 > /sys/class/net/bonding_masters
1398echo active-backup > /sys/class/net/bond1/bonding/mode
1399ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up
1400echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target
1401echo 2000 > /sys/class/net/bond1/bonding/arp_interval
1402echo +eth2 > /sys/class/net/bond1/bonding/slaves
1403echo +eth3 > /sys/class/net/bond1/bonding/slaves
1404
14053.5 Configuration with Interfaces Support
1406-----------------------------------------
1407
1408        This section applies to distros which use /etc/network/interfaces file
1409to describe network interface configuration, most notably Debian and it's
1410derivatives.
1411
1412        The ifup and ifdown commands on Debian don't support bonding out of
1413the box. The ifenslave-2.6 package should be installed to provide bonding
1414support.  Once installed, this package will provide bond-* options to be used
1415into /etc/network/interfaces.
1416
1417        Note that ifenslave-2.6 package will load the bonding module and use
1418the ifenslave command when appropriate.
1419
1420Example Configurations
1421----------------------
1422
1423In /etc/network/interfaces, the following stanza will configure bond0, in
1424active-backup mode, with eth0 and eth1 as slaves.
1425
1426auto bond0
1427iface bond0 inet dhcp
1428        bond-slaves eth0 eth1
1429        bond-mode active-backup
1430        bond-miimon 100
1431        bond-primary eth0 eth1
1432
1433If the above configuration doesn't work, you might have a system using
1434upstart for system startup. This is most notably true for recent
1435Ubuntu versions. The following stanza in /etc/network/interfaces will
1436produce the same result on those systems.
1437
1438auto bond0
1439iface bond0 inet dhcp
1440        bond-slaves none
1441        bond-mode active-backup
1442        bond-miimon 100
1443
1444auto eth0
1445iface eth0 inet manual
1446        bond-master bond0
1447        bond-primary eth0 eth1
1448
1449auto eth1
1450iface eth1 inet manual
1451        bond-master bond0
1452        bond-primary eth0 eth1
1453
1454For a full list of bond-* supported options in /etc/network/interfaces and some
1455more advanced examples tailored to you particular distros, see the files in
1456/usr/share/doc/ifenslave-2.6.
1457
14583.6 Overriding Configuration for Special Cases
1459----------------------------------------------
1460
1461When using the bonding driver, the physical port which transmits a frame is
1462typically selected by the bonding driver, and is not relevant to the user or
1463system administrator.  The output port is simply selected using the policies of
1464the selected bonding mode.  On occasion however, it is helpful to direct certain
1465classes of traffic to certain physical interfaces on output to implement
1466slightly more complex policies.  For example, to reach a web server over a
1467bonded interface in which eth0 connects to a private network, while eth1
1468connects via a public network, it may be desirous to bias the bond to send said
1469traffic over eth0 first, using eth1 only as a fall back, while all other traffic
1470can safely be sent over either interface.  Such configurations may be achieved
1471using the traffic control utilities inherent in linux.
1472
1473By default the bonding driver is multiqueue aware and 16 queues are created
1474when the driver initializes (see Documentation/networking/multiqueue.txt
1475for details).  If more or less queues are desired the module parameter
1476tx_queues can be used to change this value.  There is no sysfs parameter
1477available as the allocation is done at module init time.
1478
1479The output of the file /proc/net/bonding/bondX has changed so the output Queue
1480ID is now printed for each slave:
1481
1482Bonding Mode: fault-tolerance (active-backup)
1483Primary Slave: None
1484Currently Active Slave: eth0
1485MII Status: up
1486MII Polling Interval (ms): 0
1487Up Delay (ms): 0
1488Down Delay (ms): 0
1489
1490Slave Interface: eth0
1491MII Status: up
1492Link Failure Count: 0
1493Permanent HW addr: 00:1a:a0:12:8f:cb
1494Slave queue ID: 0
1495
1496Slave Interface: eth1
1497MII Status: up
1498Link Failure Count: 0
1499Permanent HW addr: 00:1a:a0:12:8f:cc
1500Slave queue ID: 2
1501
1502The queue_id for a slave can be set using the command:
1503
1504# echo "eth1:2" > /sys/class/net/bond0/bonding/queue_id
1505
1506Any interface that needs a queue_id set should set it with multiple calls
1507like the one above until proper priorities are set for all interfaces.  On
1508distributions that allow configuration via initscripts, multiple 'queue_id'
1509arguments can be added to BONDING_OPTS to set all needed slave queues.
1510
1511These queue id's can be used in conjunction with the tc utility to configure
1512a multiqueue qdisc and filters to bias certain traffic to transmit on certain
1513slave devices.  For instance, say we wanted, in the above configuration to
1514force all traffic bound to 192.168.1.100 to use eth1 in the bond as its output
1515device. The following commands would accomplish this:
1516
1517# tc qdisc add dev bond0 handle 1 root multiq
1518
1519# tc filter add dev bond0 protocol ip parent 1: prio 1 u32 match ip dst \
1520        192.168.1.100 action skbedit queue_mapping 2
1521
1522These commands tell the kernel to attach a multiqueue queue discipline to the
1523bond0 interface and filter traffic enqueued to it, such that packets with a dst
1524ip of 192.168.1.100 have their output queue mapping value overwritten to 2.
1525This value is then passed into the driver, causing the normal output path
1526selection policy to be overridden, selecting instead qid 2, which maps to eth1.
1527
1528Note that qid values begin at 1.  Qid 0 is reserved to initiate to the driver
1529that normal output policy selection should take place.  One benefit to simply
1530leaving the qid for a slave to 0 is the multiqueue awareness in the bonding
1531driver that is now present.  This awareness allows tc filters to be placed on
1532slave devices as well as bond devices and the bonding driver will simply act as
1533a pass-through for selecting output queues on the slave device rather than 
1534output port selection.
1535
1536This feature first appeared in bonding driver version 3.7.0 and support for
1537output slave selection was limited to round-robin and active-backup modes.
1538
15394 Querying Bonding Configuration
1540=================================
1541
15424.1 Bonding Configuration
1543-------------------------
1544
1545        Each bonding device has a read-only file residing in the
1546/proc/net/bonding directory.  The file contents include information
1547about the bonding configuration, options and state of each slave.
1548
1549        For example, the contents of /proc/net/bonding/bond0 after the
1550driver is loaded with parameters of mode=0 and miimon=1000 is
1551generally as follows:
1552
1553        Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004)
1554        Bonding Mode: load balancing (round-robin)
1555        Currently Active Slave: eth0
1556        MII Status: up
1557        MII Polling Interval (ms): 1000
1558        Up Delay (ms): 0
1559        Down Delay (ms): 0
1560
1561        Slave Interface: eth1
1562        MII Status: up
1563        Link Failure Count: 1
1564
1565        Slave Interface: eth0
1566        MII Status: up
1567        Link Failure Count: 1
1568
1569        The precise format and contents will change depending upon the
1570bonding configuration, state, and version of the bonding driver.
1571
15724.2 Network configuration
1573-------------------------
1574
1575        The network configuration can be inspected using the ifconfig
1576command.  Bonding devices will have the MASTER flag set; Bonding slave
1577devices will have the SLAVE flag set.  The ifconfig output does not
1578contain information on which slaves are associated with which masters.
1579
1580        In the example below, the bond0 interface is the master
1581(MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of
1582bond0 have the same MAC address (HWaddr) as bond0 for all modes except
1583TLB and ALB that require a unique MAC address for each slave.
1584
1585# /sbin/ifconfig
1586bond0     Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1587          inet addr:XXX.XXX.XXX.YYY  Bcast:XXX.XXX.XXX.255  Mask:255.255.252.0
1588          UP BROADCAST RUNNING MASTER MULTICAST  MTU:1500  Metric:1
1589          RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0
1590          TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0
1591          collisions:0 txqueuelen:0
1592
1593eth0      Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1594          UP BROADCAST RUNNING SLAVE MULTICAST  MTU:1500  Metric:1
1595          RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0
1596          TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0
1597          collisions:0 txqueuelen:100
1598          Interrupt:10 Base address:0x1080
1599
1600eth1      Link encap:Ethernet  HWaddr 00:C0:F0:1F:37:B4
1601          UP BROADCAST RUNNING SLAVE MULTICAST  MTU:1500  Metric:1
1602          RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0
1603          TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0
1604          collisions:0 txqueuelen:100
1605          Interrupt:9 Base address:0x1400
1606
16075. Switch Configuration
1608=======================
1609
1610        For this section, "switch" refers to whatever system the
1611bonded devices are directly connected to (i.e., where the other end of
1612the cable plugs into).  This may be an actual dedicated switch device,
1613or it may be another regular system (e.g., another computer running
1614Linux),
1615
1616        The active-backup, balance-tlb and balance-alb modes do not
1617require any specific configuration of the switch.
1618
1619        The 802.3ad mode requires that the switch have the appropriate
1620ports configured as an 802.3ad aggregation.  The precise method used
1621to configure this varies from switch to switch, but, for example, a
1622Cisco 3550 series switch requires that the appropriate ports first be
1623grouped together in a single etherchannel instance, then that
1624etherchannel is set to mode "lacp" to enable 802.3ad (instead of
1625standard EtherChannel).
1626
1627        The balance-rr, balance-xor and broadcast modes generally
1628require that the switch have the appropriate ports grouped together.
1629The nomenclature for such a group differs between switches, it may be
1630called an "etherchannel" (as in the Cisco example, above), a "trunk
1631group" or some other similar variation.  For these modes, each switch
1632will also have its own configuration options for the switch's transmit
1633policy to the bond.  Typical choices include XOR of either the MAC or
1634IP addresses.  The transmit policy of the two peers does not need to
1635match.  For these three modes, the bonding mode really selects a
1636transmit policy for an EtherChannel group; all three will interoperate
1637with another EtherChannel group.
1638
1639
16406. 802.1q VLAN Support
1641======================
1642
1643        It is possible to configure VLAN devices over a bond interface
1644using the 8021q driver.  However, only packets coming from the 8021q
1645driver and passing through bonding will be tagged by default.  Self
1646generated packets, for example, bonding's learning packets or ARP
1647packets generated by either ALB mode or the ARP monitor mechanism, are
1648tagged internally by bonding itself.  As a result, bonding must
1649"learn" the VLAN IDs configured above it, and use those IDs to tag
1650self generated packets.
1651
1652        For reasons of simplicity, and to support the use of adapters
1653that can do VLAN hardware acceleration offloading, the bonding
1654interface declares itself as fully hardware offloading capable, it gets
1655the add_vid/kill_vid notifications to gather the necessary
1656information, and it propagates those actions to the slaves.  In case
1657of mixed adapter types, hardware accelerated tagged packets that
1658should go through an adapter that is not offloading capable are
1659"un-accelerated" by the bonding driver so the VLAN tag sits in the
1660regular location.
1661
1662        VLAN interfaces *must* be added on top of a bonding interface
1663only after enslaving at least one slave.  The bonding interface has a
1664hardware address of 00:00:00:00:00:00 until the first slave is added.
1665If the VLAN interface is created prior to the first enslavement, it
1666would pick up the all-zeroes hardware address.  Once the first slave
1667is attached to the bond, the bond device itself will pick up the
1668slave's hardware address, which is then available for the VLAN device.
1669
1670        Also, be aware that a similar problem can occur if all slaves
1671are released from a bond that still has one or more VLAN interfaces on
1672top of it.  When a new slave is added, the bonding interface will
1673obtain its hardware address from the first slave, which might not
1674match the hardware address of the VLAN interfaces (which was
1675ultimately copied from an earlier slave).
1676
1677        There are two methods to insure that the VLAN device operates
1678with the correct hardware address if all slaves are removed from a
1679bond interface:
1680
1681        1. Remove all VLAN interfaces then recreate them
1682
1683        2. Set the bonding interface's hardware address so that it
1684matches the hardware address of the VLAN interfaces.
1685
1686        Note that changing a VLAN interface's HW address would set the
1687underlying device -- i.e. the bonding interface -- to promiscuous
1688mode, which might not be what you want.
1689
1690
16917. Link Monitoring
1692==================
1693
1694        The bonding driver at present supports two schemes for
1695monitoring a slave device's link state: the ARP monitor and the MII
1696monitor.
1697
1698        At the present time, due to implementation restrictions in the
1699bonding driver itself, it is not possible to enable both ARP and MII
1700monitoring simultaneously.
1701
17027.1 ARP Monitor Operation
1703-------------------------
1704
1705        The ARP monitor operates as its name suggests: it sends ARP
1706queries to one or more designated peer systems on the network, and
1707uses the response as an indication that the link is operating.  This
1708gives some assurance that traffic is actually flowing to and from one
1709or more peers on the local network.
1710
1711        The ARP monitor relies on the device driver itself to verify
1712that traffic is flowing.  In particular, the driver must keep up to
1713date the last receive time, dev->last_rx, and transmit start time,
1714dev->trans_start.  If these are not updated by the driver, then the
1715ARP monitor will immediately fail any slaves using that driver, and
1716those slaves will stay down.  If networking monitoring (tcpdump, etc)
1717shows the ARP requests and replies on the network, then it may be that
1718your device driver is not updating last_rx and trans_start.
1719
17207.2 Configuring Multiple ARP Targets
1721------------------------------------
1722
1723        While ARP monitoring can be done with just one target, it can
1724be useful in a High Availability setup to have several targets to
1725monitor.  In the case of just one target, the target itself may go
1726down or have a problem making it unresponsive to ARP requests.  Having
1727an additional target (or several) increases the reliability of the ARP
1728monitoring.
1729
1730        Multiple ARP targets must be separated by commas as follows:
1731
1732# example options for ARP monitoring with three targets
1733alias bond0 bonding
1734options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9
1735
1736        For just a single target the options would resemble:
1737
1738# example options for ARP monitoring with one target
1739alias bond0 bonding
1740options bond0 arp_interval=60 arp_ip_target=192.168.0.100
1741
1742
17437.3 MII Monitor Operation
1744-------------------------
1745
1746        The MII monitor monitors only the carrier state of the local
1747network interface.  It accomplishes this in one of three ways: by
1748depending upon the device driver to maintain its carrier state, by
1749querying the device's MII registers, or by making an ethtool query to
1750the device.
1751
1752        If the use_carrier module parameter is 1 (the default value),
1753then the MII monitor will rely on the driver for carrier state
1754information (via the netif_carrier subsystem).  As explained in the
1755use_carrier parameter information, above, if the MII monitor fails to
1756detect carrier loss on the device (e.g., when the cable is physically
1757disconnected), it may be that the driver does not support
1758netif_carrier.
1759
1760        If use_carrier is 0, then the MII monitor will first query the
1761device's (via ioctl) MII registers and check the link state.  If that
1762request fails (not just that it returns carrier down), then the MII
1763monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain
1764the same information.  If both methods fail (i.e., the driver either
1765does not support or had some error in processing both the MII register
1766and ethtool requests), then the MII monitor will assume the link is
1767up.
1768
17698. Potential Sources of Trouble
1770===============================
1771
17728.1 Adventures in Routing
1773-------------------------
1774
1775        When bonding is configured, it is important that the slave
1776devices not have routes that supersede routes of the master (or,
1777generally, not have routes at all).  For example, suppose the bonding
1778device bond0 has two slaves, eth0 and eth1, and the routing table is
1779as follows:
1780
1781Kernel IP routing table
1782Destination     Gateway         Genmask         Flags   MSS Window  irtt Iface
178310.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 eth0
178410.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 eth1
178510.0.0.0        0.0.0.0         255.255.0.0     U        40 0          0 bond0
1786127.0.0.0       0.0.0.0         255.0.0.0       U        40 0          0 lo
1787
1788        This routing configuration will likely still update the
1789receive/transmit times in the driver (needed by the ARP monitor), but
1790may bypass the bonding driver (because outgoing traffic to, in this
1791case, another host on network 10 would use eth0 or eth1 before bond0).
1792
1793        The ARP monitor (and ARP itself) may become confused by this
1794configuration, because ARP requests (generated by the ARP monitor)
1795will be sent on one interface (bond0), but the corresponding reply
1796will arrive on a different interface (eth0).  This reply looks to ARP
1797as an unsolicited ARP reply (because ARP matches replies on an
1798interface basis), and is discarded.  The MII monitor is not affected
1799by the state of the routing table.
1800
1801        The solution here is simply to insure that slaves do not have
1802routes of their own, and if for some reason they must, those routes do
1803not supersede routes of their master.  This should generally be the
1804case, but unusual configurations or errant manual or automatic static
1805route additions may cause trouble.
1806
18078.2 Ethernet Device Renaming
1808----------------------------
1809
1810        On systems with network configuration scripts that do not
1811associate physical devices directly with network interface names (so
1812that the same physical device always has the same "ethX" name), it may
1813be necessary to add some special logic to config files in
1814/etc/modprobe.d/.
1815
1816        For example, given a modules.conf containing the following:
1817
1818alias bond0 bonding
1819options bond0 mode=some-mode miimon=50
1820alias eth0 tg3
1821alias eth1 tg3
1822alias eth2 e1000
1823alias eth3 e1000
1824
1825        If neither eth0 and eth1 are slaves to bond0, then when the
1826bond0 interface comes up, the devices may end up reordered.  This
1827happens because bonding is loaded first, then its slave device's
1828drivers are loaded next.  Since no other drivers have been loaded,
1829when the e1000 driver loads, it will receive eth0 and eth1 for its
1830devices, but the bonding configuration tries to enslave eth2 and eth3
1831(which may later be assigned to the tg3 devices).
1832
1833        Adding the following:
1834
1835add above bonding e1000 tg3
1836
1837        causes modprobe to load e1000 then tg3, in that order, when
1838bonding is loaded.  This command is fully documented in the
1839modules.conf manual page.
1840
1841        On systems utilizing modprobe an equivalent problem can occur.
1842In this case, the following can be added to config files in
1843/etc/modprobe.d/ as:
1844
1845softdep bonding pre: tg3 e1000
1846
1847        This will load tg3 and e1000 modules before loading the bonding one.
1848Full documentation on this can be found in the modprobe.d and modprobe
1849manual pages.
1850
18518.3. Painfully Slow Or No Failed Link Detection By Miimon
1852---------------------------------------------------------
1853
1854        By default, bonding enables the use_carrier option, which
1855instructs bonding to trust the driver to maintain carrier state.
1856
1857        As discussed in the options section, above, some drivers do
1858not support the netif_carrier_on/_off link state tracking system.
1859With use_carrier enabled, bonding will always see these links as up,
1860regardless of their actual state.
1861
1862        Additionally, other drivers do support netif_carrier, but do
1863not maintain it in real time, e.g., only polling the link state at
1864some fixed interval.  In this case, miimon will detect failures, but
1865only after some long period of time has expired.  If it appears that
1866miimon is very slow in detecting link failures, try specifying
1867use_carrier=0 to see if that improves the failure detection time.  If
1868it does, then it may be that the driver checks the carrier state at a
1869fixed interval, but does not cache the MII register values (so the
1870use_carrier=0 method of querying the registers directly works).  If
1871use_carrier=0 does not improve the failover, then the driver may cache
1872the registers, or the problem may be elsewhere.
1873
1874        Also, remember that miimon only checks for the device's
1875carrier state.  It has no way to determine the state of devices on or
1876beyond other ports of a switch, or if a switch is refusing to pass
1877traffic while still maintaining carrier on.
1878
18799. SNMP agents
1880===============
1881
1882        If running SNMP agents, the bonding driver should be loaded
1883before any network drivers participating in a bond.  This requirement
1884is due to the interface index (ipAdEntIfIndex) being associated to
1885the first interface found with a given IP address.  That is, there is
1886only one ipAdEntIfIndex for each IP address.  For example, if eth0 and
1887eth1 are slaves of bond0 and the driver for eth0 is loaded before the
1888bonding driver, the interface for the IP address will be associated
1889with the eth0 interface.  This configuration is shown below, the IP
1890address 192.168.1.1 has an interface index of 2 which indexes to eth0
1891in the ifDescr table (ifDescr.2).
1892
1893     interfaces.ifTable.ifEntry.ifDescr.1 = lo
1894     interfaces.ifTable.ifEntry.ifDescr.2 = eth0
1895     interfaces.ifTable.ifEntry.ifDescr.3 = eth1
1896     interfaces.ifTable.ifEntry.ifDescr.4 = eth2
1897     interfaces.ifTable.ifEntry.ifDescr.5 = eth3
1898     interfaces.ifTable.ifEntry.ifDescr.6 = bond0
1899     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5
1900     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1901     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4
1902     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1903
1904        This problem is avoided by loading the bonding driver before
1905any network drivers participating in a bond.  Below is an example of
1906loading the bonding driver first, the IP address 192.168.1.1 is
1907correctly associated with ifDescr.2.
1908
1909     interfaces.ifTable.ifEntry.ifDescr.1 = lo
1910     interfaces.ifTable.ifEntry.ifDescr.2 = bond0
1911     interfaces.ifTable.ifEntry.ifDescr.3 = eth0
1912     interfaces.ifTable.ifEntry.ifDescr.4 = eth1
1913     interfaces.ifTable.ifEntry.ifDescr.5 = eth2
1914     interfaces.ifTable.ifEntry.ifDescr.6 = eth3
1915     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6
1916     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2
1917     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5
1918     ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1
1919
1920        While some distributions may not report the interface name in
1921ifDescr, the association between the IP address and IfIndex remains
1922and SNMP functions such as Interface_Scan_Next will report that
1923association.
1924
192510. Promiscuous mode
1926====================
1927
1928        When running network monitoring tools, e.g., tcpdump, it is
1929common to enable promiscuous mode on the device, so that all traffic
1930is seen (instead of seeing only traffic destined for the local host).
1931The bonding driver handles promiscuous mode changes to the bonding
1932master device (e.g., bond0), and propagates the setting to the slave
1933devices.
1934
1935        For the balance-rr, balance-xor, broadcast, and 802.3ad modes,
1936the promiscuous mode setting is propagated to all slaves.
1937
1938        For the active-backup, balance-tlb and balance-alb modes, the
1939promiscuous mode setting is propagated only to the active slave.
1940
1941        For balance-tlb mode, the active slave is the slave currently
1942receiving inbound traffic.
1943
1944        For balance-alb mode, the active slave is the slave used as a
1945"primary."  This slave is used for mode-specific control traffic, for
1946sending to peers that are unassigned or if the load is unbalanced.
1947
1948        For the active-backup, balance-tlb and balance-alb modes, when
1949the active slave changes (e.g., due to a link failure), the
1950promiscuous setting will be propagated to the new active slave.
1951
195211. Configuring Bonding for High Availability
1953=============================================
1954
1955        High Availability refers to configurations that provide
1956maximum network availability by having redundant or backup devices,
1957links or switches between the host and the rest of the world.  The
1958goal is to provide the maximum availability of network connectivity
1959(i.e., the network always works), even though other configurations
1960could provide higher throughput.
1961
196211.1 High Availability in a Single Switch Topology
1963--------------------------------------------------
1964
1965        If two hosts (or a host and a single switch) are directly
1966connected via multiple physical links, then there is no availability
1967penalty to optimizing for maximum bandwidth.  In this case, there is
1968only one switch (or peer), so if it fails, there is no alternative
1969access to fail over to.  Additionally, the bonding load balance modes
1970support link monitoring of their members, so if individual links fail,
1971the load will be rebalanced across the remaining devices.
1972
1973        See Section 12, "Configuring Bonding for Maximum Throughput"
1974for information on configuring bonding with one peer device.
1975
197611.2 High Availability in a Multiple Switch Topology
1977----------------------------------------------------
1978
1979        With multiple switches, the configuration of bonding and the
1980network changes dramatically.  In multiple switch topologies, there is
1981a trade off between network availability and usable bandwidth.
1982
1983        Below is a sample network, configured to maximize the
1984availability of the network:
1985
1986                |                                     |
1987                |port3                           port3|
1988          +-----+----+                          +-----+----+
1989          |          |port2       ISL      port2|          |
1990          | switch A +--------------------------+ switch B |
1991          |          |                          |          |
1992          +-----+----+                          +-----++---+
1993                |port1                           port1|
1994                |             +-------+               |
1995                +-------------+ host1 +---------------+
1996                         eth0 +-------+ eth1
1997
1998        In this configuration, there is a link between the two
1999switches (ISL, or inter switch link), and multiple ports connecting to
2000the outside world ("port3" on each switch).  There is no technical
2001reason that this could not be extended to a third switch.
2002
200311.2.1 HA Bonding Mode Selection for Multiple Switch Topology
2004-------------------------------------------------------------
2005
2006        In a topology such as the example above, the active-backup and
2007broadcast modes are the only useful bonding modes when optimizing for
2008availability; the other modes require all links to terminate on the
2009same peer for them to behave rationally.
2010
2011active-backup: This is generally the preferred mode, particularly if
2012        the switches have an ISL and play together well.  If the
2013        network configuration is such that one switch is specifically
2014        a backup switch (e.g., has lower capacity, higher cost, etc),
2015        then the primary option can be used to insure that the
2016        preferred link is always used when it is available.
2017
2018broadcast: This mode is really a special purpose mode, and is suitable
2019        only for very specific needs.  For example, if the two
2020        switches are not connected (no ISL), and the networks beyond
2021        them are totally independent.  In this case, if it is
2022        necessary for some specific one-way traffic to reach both
2023        independent networks, then the broadcast mode may be suitable.
2024
202511.2.2 HA Link Monitoring Selection for Multiple Switch Topology
2026----------------------------------------------------------------
2027
2028        The choice of link monitoring ultimately depends upon your
2029switch.  If the switch can reliably fail ports in response to other
2030failures, then either the MII or ARP monitors should work.  For
2031example, in the above example, if the "port3" link fails at the remote
2032end, the MII monitor has no direct means to detect this.  The ARP
2033monitor could be configured with a target at the remote end of port3,
2034thus detecting that failure without switch support.
2035
2036        In general, however, in a multiple switch topology, the ARP
2037monitor can provide a higher level of reliability in detecting end to
2038end connectivity failures (which may be caused by the failure of any
2039individual component to pass traffic for any reason).  Additionally,
2040the ARP monitor should be configured with multiple targets (at least
2041one for each switch in the network).  This will insure that,
2042regardless of which switch is active, the ARP monitor has a suitable
2043target to query.
2044
2045        Note, also, that of late many switches now support a functionality
2046generally referred to as "trunk failover."  This is a feature of the
2047switch that causes the link state of a particular switch port to be set
2048down (or up) when the state of another switch port goes down (or up).
2049Its purpose is to propagate link failures from logically "exterior" ports
2050to the logically "interior" ports that bonding is able to monitor via
2051miimon.  Availability and configuration for trunk failover varies by
2052switch, but this can be a viable alternative to the ARP monitor when using
2053suitable switches.
2054
205512. Configuring Bonding for Maximum Throughput
2056==============================================
2057
205812.1 Maximizing Throughput in a Single Switch Topology
2059------------------------------------------------------
2060
2061        In a single switch configuration, the best method to maximize
2062throughput depends upon the application and network environment.  The
2063various load balancing modes each have strengths and weaknesses in
2064different environments, as detailed below.
2065
2066        For this discussion, we will break down the topologies into
2067two categories.  Depending upon the destination of most traffic, we
2068categorize them into either "gatewayed" or "local" configurations.
2069
2070        In a gatewayed configuration, the "switch" is acting primarily
2071as a router, and the majority of traffic passes through this router to
2072other networks.  An example would be the following:
2073
2074
2075     +----------+                     +----------+
2076     |          |eth0            port1|          | to other networks
2077     | Host A   +---------------------+ router   +------------------->
2078     |          +---------------------+          | Hosts B and C are out
2079     |          |eth1            port2|          | here somewhere
2080     +----------+                     +----------+
2081
2082        The router may be a dedicated router device, or another host
2083acting as a gateway.  For our discussion, the important point is that
2084the majority of traffic from Host A will pass through the router to
2085some other network before reaching its final destination.
2086
2087        In a gatewayed network configuration, although Host A may
2088communicate with many other systems, all of its traffic will be sent
2089and received via one other peer on the local network, the router.
2090
2091        Note that the case of two systems connected directly via
2092multiple physical links is, for purposes of configuring bonding, the
2093same as a gatewayed configuration.  In that case, it happens that all
2094traffic is destined for the "gateway" itself, not some other network
2095beyond the gateway.
2096
2097        In a local configuration, the "switch" is acting primarily as
2098a switch, and the majority of traffic passes through this switch to
2099reach other stations on the same network.  An example would be the
2100following:
2101
2102    +----------+            +----------+       +--------+
2103    |          |eth0   port1|          +-------+ Host B |
2104    |  Host A  +------------+  switch  |port3  +--------+
2105    |          +------------+          |                  +--------+
2106    |          |eth1   port2|          +------------------+ Host C |
2107    +----------+            +----------+port4             +--------+
2108
2109
2110        Again, the switch may be a dedicated switch device, or another
2111host acting as a gateway.  For our discussion, the important point is
2112that the majority of traffic from Host A is destined for other hosts
2113on the same local network (Hosts B and C in the above example).
2114
2115        In summary, in a gatewayed configuration, traffic to and from
2116the bonded device will be to the same MAC level peer on the network
2117(the gateway itself, i.e., the router), regardless of its final
2118destination.  In a local configuration, traffic flows directly to and
2119from the final destinations, thus, each destination (Host B, Host C)
2120will be addressed directly by their individual MAC addresses.
2121
2122        This distinction between a gatewayed and a local network
2123configuration is important because many of the load balancing modes
2124available use the MAC addresses of the local network source and
2125destination to make load balancing decisions.  The behavior of each
2126mode is described below.
2127
2128
212912.1.1 MT Bonding Mode Selection for Single Switch Topology
2130-----------------------------------------------------------
2131
2132        This configuration is the easiest to set up and to understand,
2133although you will have to decide which bonding mode best suits your
2134needs.  The trade offs for each mode are detailed below:
2135
2136balance-rr: This mode is the only mode that will permit a single
2137        TCP/IP connection to stripe traffic across multiple
2138        interfaces. It is therefore the only mode that will allow a
2139        single TCP/IP stream to utilize more than one interface's
2140        worth of throughput.  This comes at a cost, however: the
2141        striping generally results in peer systems receiving packets out
2142        of order, causing TCP/IP's congestion control system to kick
2143        in, often by retransmitting segments.
2144
2145        It is possible to adjust TCP/IP's congestion limits by
2146        altering the net.ipv4.tcp_reordering sysctl parameter.  The
2147        usual default value is 3, and the maximum useful value is 127.
2148        For a four interface balance-rr bond, expect that a single
2149        TCP/IP stream will utilize no more than approximately 2.3
2150        interface's worth of throughput, even after adjusting
2151        tcp_reordering.
2152
2153        Note that the fraction of packets that will be delivered out of
2154        order is highly variable, and is unlikely to be zero.  The level
2155        of reordering depends upon a variety of factors, including the
2156        networking interfaces, the switch, and the topology of the
2157        configuration.  Speaking in general terms, higher speed network
2158        cards produce more reordering (due to factors such as packet
2159        coalescing), and a "many to many" topology will reorder at a
2160        higher rate than a "many slow to one fast" configuration.
2161
2162        Many switches do not support any modes that stripe traffic
2163        (instead choosing a port based upon IP or MAC level addresses);
2164        for those devices, traffic for a particular connection flowing
2165        through the switch to a balance-rr bond will not utilize greater
2166        than one interface's worth of bandwidth.
2167
2168        If you are utilizing protocols other than TCP/IP, UDP for
2169        example, and your application can tolerate out of order
2170        delivery, then this mode can allow for single stream datagram
2171        performance that scales near linearly as interfaces are added
2172        to the bond.
2173
2174        This mode requires the switch to have the appropriate ports
2175        configured for "etherchannel" or "trunking."
2176
2177active-backup: There is not much advantage in this network topology to
2178        the active-backup mode, as the inactive backup devices are all
2179        connected to the same peer as the primary.  In this case, a
2180        load balancing mode (with link monitoring) will provide the
2181        same level of network availability, but with increased
2182        available bandwidth.  On the plus side, active-backup mode
2183        does not require any configuration of the switch, so it may
2184        have value if the hardware available does not support any of
2185        the load balance modes.
2186
2187balance-xor: This mode will limit traffic such that packets destined
2188        for specific peers will always be sent over the same
2189        interface.  Since the destination is determined by the MAC
2190        addresses involved, this mode works best in a "local" network
2191        configuration (as described above), with destinations all on
2192        the same local network.  This mode is likely to be suboptimal
2193        if all your traffic is passed through a single router (i.e., a
2194        "gatewayed" network configuration, as described above).
2195
2196        As with balance-rr, the switch ports need to be configured for
2197        "etherchannel" or "trunking."
2198
2199broadcast: Like active-backup, there is not much advantage to this
2200        mode in this type of network topology.
2201
2202802.3ad: This mode can be a good choice for this type of network
2203        topology.  The 802.3ad mode is an IEEE standard, so all peers
2204        that implement 802.3ad should interoperate well.  The 802.3ad
2205        protocol includes automatic configuration of the aggregates,
2206        so minimal manual configuration of the switch is needed
2207        (typically only to designate that some set of devices is
2208        available for 802.3ad).  The 802.3ad standard also mandates
2209        that frames be delivered in order (within certain limits), so
2210        in general single connections will not see misordering of
2211        packets.  The 802.3ad mode does have some drawbacks: the
2212        standard mandates that all devices in the aggregate operate at
2213        the same speed and duplex.  Also, as with all bonding load
2214        balance modes other than balance-rr, no single connection will
2215        be able to utilize more than a single interface's worth of
2216        bandwidth.  
2217
2218        Additionally, the linux bonding 802.3ad implementation
2219        distributes traffic by peer (using an XOR of MAC addresses),
2220        so in a "gatewayed" configuration, all outgoing traffic will
2221        generally use the same device.  Incoming traffic may also end
2222        up on a single device, but that is dependent upon the
2223        balancing policy of the peer's 8023.ad implementation.  In a
2224        "local" configuration, traffic will be distributed across the
2225        devices in the bond.
2226
2227        Finally, the 802.3ad mode mandates the use of the MII monitor,
2228        therefore, the ARP monitor is not available in this mode.
2229
2230balance-tlb: The balance-tlb mode balances outgoing traffic by peer.
2231        Since the balancing is done according to MAC address, in a
2232        "gatewayed" configuration (as described above), this mode will
2233        send all traffic across a single device.  However, in a
2234        "local" network configuration, this mode balances multiple
2235        local network peers across devices in a vaguely intelligent
2236        manner (not a simple XOR as in balance-xor or 802.3ad mode),
2237        so that mathematically unlucky MAC addresses (i.e., ones that
2238        XOR to the same value) will not all "bunch up" on a single
2239        interface.
2240
2241        Unlike 802.3ad, interfaces may be of differing speeds, and no
2242        special switch configuration is required.  On the down side,
2243        in this mode all incoming traffic arrives over a single
2244        interface, this mode requires certain ethtool support in the
2245        network device driver of the slave interfaces, and the ARP
2246        monitor is not available.
2247
2248balance-alb: This mode is everything that balance-tlb is, and more.
2249        It has all of the features (and restrictions) of balance-tlb,
2250        and will also balance incoming traffic from local network
2251        peers (as described in the Bonding Module Options section,
2252        above).
2253
2254        The only additional down side to this mode is that the network
2255        device driver must support changing the hardware address while
2256        the device is open.
2257
225812.1.2 MT Link Monitoring for Single Switch Topology
2259----------------------------------------------------
2260
2261        The choice of link monitoring may largely depend upon which
2262mode you choose to use.  The more advanced load balancing modes do not
2263support the use of the ARP monitor, and are thus restricted to using
2264the MII monitor (which does not provide as high a level of end to end
2265assurance as the ARP monitor).
2266
226712.2 Maximum Throughput in a Multiple Switch Topology
2268-----------------------------------------------------
2269
2270        Multiple switches may be utilized to optimize for throughput
2271when they are configured in parallel as part of an isolated network
2272between two or more systems, for example:
2273
2274                       +-----------+
2275                       |  Host A   | 
2276                       +-+---+---+-+
2277                         |   |   |
2278                +--------+   |   +---------+
2279                |            |             |
2280         +------+---+  +-----+----+  +-----+----+
2281         | Switch A |  | Switch B |  | Switch C |
2282         +------+---+  +-----+----+  +-----+----+
2283                |            |             |
2284                +--------+   |   +---------+
2285                         |   |   |
2286                       +-+---+---+-+
2287                       |  Host B   | 
2288                       +-----------+
2289
2290        In this configuration, the switches are isolated from one
2291another.  One reason to employ a topology such as this is for an
2292isolated network with many hosts (a cluster configured for high
2293performance, for example), using multiple smaller switches can be more
2294cost effective than a single larger switch, e.g., on a network with 24
2295hosts, three 24 port switches can be significantly less expensive than
2296a single 72 port switch.
2297
2298        If access beyond the network is required, an individual host
2299can be equipped with an additional network device connected to an
2300external network; this host then additionally acts as a gateway.
2301
230212.2.1 MT Bonding Mode Selection for Multiple Switch Topology
2303-------------------------------------------------------------
2304
2305        In actual practice, the bonding mode typically employed in
2306configurations of this type is balance-rr.  Historically, in this
2307network configuration, the usual caveats about out of order packet
2308delivery are mitigated by the use of network adapters that do not do
2309any kind of packet coalescing (via the use of NAPI, or because the
2310device itself does not generate interrupts until some number of
2311packets has arrived).  When employed in this fashion, the balance-rr
2312mode allows individual connections between two hosts to effectively
2313utilize greater than one interface's bandwidth.
2314
231512.2.2 MT Link Monitoring for Multiple Switch Topology
2316------------------------------------------------------
2317
2318        Again, in actual practice, the MII monitor is most often used
2319in this configuration, as performance is given preference over
2320availability.  The ARP monitor will function in this topology, but its
2321advantages over the MII monitor are mitigated by the volume of probes
2322needed as the number of systems involved grows (remember that each
2323host in the network is configured with bonding).
2324
232513. Switch Behavior Issues
2326==========================
2327
232813.1 Link Establishment and Failover Delays
2329-------------------------------------------
2330
2331        Some switches exhibit undesirable behavior with regard to the
2332timing of link up and down reporting by the switch.
2333
2334        First, when a link comes up, some switches may indicate that
2335the link is up (carrier available), but not pass traffic over the
2336interface for some period of time.  This delay is typically due to
2337some type of autonegotiation or routing protocol, but may also occur
2338during switch initialization (e.g., during recovery after a switch
2339failure).  If you find this to be a problem, specify an appropriate
2340value to the updelay bonding module option to delay the use of the
2341relevant interface(s).
2342
2343        Second, some switches may "bounce" the link state one or more
2344times while a link is changing state.  This occurs most commonly while
2345the switch is initializing.  Again, an appropriate updelay value may
2346help.
2347
2348        Note that when a bonding interface has no active links, the
2349driver will immediately reuse the first link that goes up, even if the
2350updelay parameter has been specified (the updelay is ignored in this
2351case).  If there are slave interfaces waiting for the updelay timeout
2352to expire, the interface that first went into that state will be
2353immediately reused.  This reduces down time of the network if the
2354value of updelay has been overestimated, and since this occurs only in
2355cases with no connectivity, there is no additional penalty for
2356ignoring the updelay.
2357
2358        In addition to the concerns about switch timings, if your
2359switches take a long time to go into backup mode, it may be desirable
2360to not activate a backup interface immediately after a link goes down.
2361Failover may be delayed via the downdelay bonding module option.
2362
236313.2 Duplicated Incoming Packets
2364--------------------------------
2365
2366        NOTE: Starting with version 3.0.2, the bonding driver has logic to
2367suppress duplicate packets, which should largely eliminate this problem.
2368The following description is kept for reference.
2369
2370        It is not uncommon to observe a short burst of duplicated
2371traffic when the bonding device is first used, or after it has been
2372idle for some period of time.  This is most easily observed by issuing
2373a "ping" to some other host on the network, and noticing that the
2374output from ping flags duplicates (typically one per slave).
2375
2376        For example, on a bond in active-backup mode with five slaves
2377all connected to one switch, the output may appear as follows:
2378
2379# ping -n 10.0.4.2
2380PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data.
238164 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms
238264 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
238364 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
238464 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
238564 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!)
238664 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms
238764 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms
238864 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms
2389
2390        This is not due to an error in the bonding driver, rather, it
2391is a side effect of how many switches update their MAC forwarding
2392tables.  Initially, the switch does not associate the MAC address in
2393the packet with a particular switch port, and so it may send the
2394traffic to all ports until its MAC forwarding table is updated.  Since
2395the interfaces attached to the bond may occupy multiple ports on a
2396single switch, when the switch (temporarily) floods the traffic to all
2397ports, the bond device receives multiple copies of the same packet
2398(one per slave device).
2399
2400        The duplicated packet behavior is switch dependent, some
2401switches exhibit this, and some do not.  On switches that display this
2402behavior, it can be induced by clearing the MAC forwarding table (on
2403most Cisco switches, the privileged command "clear mac address-table
2404dynamic" will accomplish this).
2405
240614. Hardware Specific Considerations
2407====================================
2408
2409        This section contains additional information for configuring
2410bonding on specific hardware platforms, or for interfacing bonding
2411with particular switches or other devices.
2412
241314.1 IBM BladeCenter
2414--------------------
2415
2416        This applies to the JS20 and similar systems.
2417
2418        On the JS20 blades, the bonding driver supports only
2419balance-rr, active-backup, balance-tlb and balance-alb modes.  This is
2420largely due to the network topology inside the BladeCenter, detailed
2421below.
2422
2423JS20 network adapter information
2424--------------------------------
2425
2426        All JS20s come with two Broadcom Gigabit Ethernet ports
2427integrated on the planar (that's "motherboard" in IBM-speak).  In the
2428BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to
2429I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2.
2430An add-on Broadcom daughter card can be installed on a JS20 to provide
2431two more Gigabit Ethernet ports.  These ports, eth2 and eth3, are
2432wired to I/O Modules 3 and 4, respectively.
2433
2434        Each I/O Module may contain either a switch or a passthrough
2435module (which allows ports to be directly connected to an external
2436switch).  Some bonding modes require a specific BladeCenter internal
2437network topology in order to function; these are detailed below.
2438
2439        Additional BladeCenter-specific networking information can be
2440found in two IBM Redbooks (www.ibm.com/redbooks):
2441
2442"IBM eServer BladeCenter Networking Options"
2443"IBM eServer BladeCenter Layer 2-7 Network Switching"
2444
2445BladeCenter networking configuration
2446------------------------------------
2447
2448        Because a BladeCenter can be configured in a very large number
2449of ways, this discussion will be confined to describing basic
2450configurations.
2451
2452        Normally, Ethernet Switch Modules (ESMs) are used in I/O
2453modules 1 and 2.  In this configuration, the eth0 and eth1 ports of a
2454JS20 will be connected to different internal switches (in the
2455respective I/O modules).
2456
2457        A passthrough module (OPM or CPM, optical or copper,
2458passthrough module) connects the I/O module directly to an external
2459switch.  By using PMs in I/O module #1 and #2, the eth0 and eth1
2460interfaces of a JS20 can be redirected to the outside world and
2461connected to a common external switch.
2462
2463        Depending upon the mix of ESMs and PMs, the network will
2464appear to bonding as either a single switch topology (all PMs) or as a
2465multiple switch topology (one or more ESMs, zero or more PMs).  It is
2466also possible to connect ESMs together, resulting in a configuration
2467much like the example in "High Availability in a Multiple Switch
2468Topology," above.
2469
2470Requirements for specific modes
2471-------------------------------
2472
2473        The balance-rr mode requires the use of passthrough modules
2474for devices in the bond, all connected to an common external switch.
2475That switch must be configured for "etherchannel" or "trunking" on the
2476appropriate ports, as is usual for balance-rr.
2477
2478        The balance-alb and balance-tlb modes will function with
2479either switch modules or passthrough modules (or a mix).  The only
2480specific requirement for these modes is that all network interfaces
2481must be able to reach all destinations for traffic sent over the
2482bonding device (i.e., the network must converge at some point outside
2483the BladeCenter).
2484
2485        The active-backup mode has no additional requirements.
2486
2487Link monitoring issues
2488----------------------
2489
2490        When an Ethernet Switch Module is in place, only the ARP
2491monitor will reliably detect link loss to an external switch.  This is
2492nothing unusual, but examination of the BladeCenter cabinet would
2493suggest that the "external" network ports are the ethernet ports for
2494the system, when it fact there is a switch between these "external"
2495ports and the devices on the JS20 system itself.  The MII monitor is
2496only able to detect link failures between the ESM and the JS20 system.
2497
2498        When a passthrough module is in place, the MII monitor does
2499detect failures to the "external" port, which is then directly
2500connected to the JS20 system.
2501
2502Other concerns
2503--------------
2504
2505        The Serial Over LAN (SoL) link is established over the primary
2506ethernet (eth0) only, therefore, any loss of link to eth0 will result
2507in losing your SoL connection.  It will not fail over with other
2508network traffic, as the SoL system is beyond the control of the
2509bonding driver.
2510
2511        It may be desirable to disable spanning tree on the switch
2512(either the internal Ethernet Switch Module, or an external switch) to
2513avoid fail-over delay issues when using bonding.
2514
2515        
251615. Frequently Asked Questions
2517==============================
2518
25191.  Is it SMP safe?
2520
2521        Yes. The old 2.0.xx channel bonding patch was not SMP safe.
2522The new driver was designed to be SMP safe from the start.
2523
25242.  What type of cards will work with it?
2525
2526        Any Ethernet type cards (you can even mix cards - a Intel
2527EtherExpress PRO/100 and a 3com 3c905b, for example).  For most modes,
2528devices need not be of the same speed.
2529
2530        Starting with version 3.2.1, bonding also supports Infiniband
2531slaves in active-backup mode.
2532
25333.  How many bonding devices can I have?
2534
2535        There is no limit.
2536
25374.  How many slaves can a bonding device have?
2538
2539        This is limited only by the number of network interfaces Linux
2540supports and/or the number of network cards you can place in your
2541system.
2542
25435.  What happens when a slave link dies?
2544
2545        If link monitoring is enabled, then the failing device will be
2546disabled.  The active-backup mode will fail over to a backup link, and
2547other modes will ignore the failed link.  The link will continue to be
2548monitored, and should it recover, it will rejoin the bond (in whatever
2549manner is appropriate for the mode). See the sections on High
2550Availability and the documentation for each mode for additional
2551information.
2552        
2553        Link monitoring can be enabled via either the miimon or
2554arp_interval parameters (described in the module parameters section,
2555above).  In general, miimon monitors the carrier state as sensed by
2556the underlying network device, and the arp monitor (arp_interval)
2557monitors connectivity to another host on the local network.
2558
2559        If no link monitoring is configured, the bonding driver will
2560be unable to detect link failures, and will assume that all links are
2561always available.  This will likely result in lost packets, and a
2562resulting degradation of performance.  The precise performance loss
2563depends upon the bonding mode and network configuration.
2564
25656.  Can bonding be used for High Availability?
2566
2567        Yes.  See the section on High Availability for details.
2568
25697.  Which switches/systems does it work with?
2570
2571        The full answer to this depends upon the desired mode.
2572
2573        In the basic balance modes (balance-rr and balance-xor), it
2574works with any system that supports etherchannel (also called
2575trunking).  Most managed switches currently available have such
2576support, and many unmanaged switches as well.
2577
2578        The advanced balance modes (balance-tlb and balance-alb) do
2579not have special switch requirements, but do need device drivers that
2580support specific features (described in the appropriate section under
2581module parameters, above).
2582
2583        In 802.3ad mode, it works with systems that support IEEE
2584802.3ad Dynamic Link Aggregation.  Most managed and many unmanaged
2585switches currently available support 802.3ad.
2586
2587        The active-backup mode should work with any Layer-II switch.
2588
25898.  Where does a bonding device get its MAC address from?
2590
2591        When using slave devices that have fixed MAC addresses, or when
2592the fail_over_mac option is enabled, the bonding device's MAC address is
2593the MAC address of the active slave.
2594
2595        For other configurations, if not explicitly configured (with
2596ifconfig or ip link), the MAC address of the bonding device is taken from
2597its first slave device.  This MAC address is then passed to all following
2598slaves and remains persistent (even if the first slave is removed) until
2599the bonding device is brought down or reconfigured.
2600
2601        If you wish to change the MAC address, you can set it with
2602ifconfig or ip link:
2603
2604# ifconfig bond0 hw ether 00:11:22:33:44:55
2605
2606# ip link set bond0 address 66:77:88:99:aa:bb
2607
2608        The MAC address can be also changed by bringing down/up the
2609device and then changing its slaves (or their order):
2610
2611# ifconfig bond0 down ; modprobe -r bonding
2612# ifconfig bond0 .... up
2613# ifenslave bond0 eth...
2614
2615        This method will automatically take the address from the next
2616slave that is added.
2617
2618        To restore your slaves' MAC addresses, you need to detach them
2619from the bond (`ifenslave -d bond0 eth0'). The bonding driver will
2620then restore the MAC addresses that the slaves had before they were
2621enslaved.
2622
262316. Resources and Links
2624=======================
2625
2626        The latest version of the bonding driver can be found in the latest
2627version of the linux kernel, found on http://kernel.org
2628
2629        The latest version of this document can be found in the latest kernel
2630source (named Documentation/networking/bonding.txt).
2631
2632        Discussions regarding the usage of the bonding driver take place on the
2633bonding-devel mailing list, hosted at sourceforge.net. If you have questions or
2634problems, post them to the list.  The list address is:
2635
2636bonding-devel@lists.sourceforge.net
2637
2638        The administrative interface (to subscribe or unsubscribe) can
2639be found at:
2640
2641https://lists.sourceforge.net/lists/listinfo/bonding-devel
2642
2643        Discussions regarding the development of the bonding driver take place
2644on the main Linux network mailing list, hosted at vger.kernel.org. The list
2645address is:
2646
2647netdev@vger.kernel.org
2648
2649        The administrative interface (to subscribe or unsubscribe) can
2650be found at:
2651
2652http://vger.kernel.org/vger-lists.html#netdev
2653
2654Donald Becker's Ethernet Drivers and diag programs may be found at :
2655 - http://web.archive.org/web/*/http://www.scyld.com/network/ 
2656
2657You will also find a lot of information regarding Ethernet, NWay, MII,
2658etc. at www.scyld.com.
2659
2660-- END --
2661
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.