1 2 Linux Ethernet Bonding Driver HOWTO 3 4 Latest update: 24 April 2006 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 52 534. Querying Bonding Configuration 544.1 Bonding Configuration 554.2 Network Configuration 56 575. Switch Configuration 58 596. 802.1q VLAN Support 60 617. Link Monitoring 627.1 ARP Monitor Operation 637.2 Configuring Multiple ARP Targets 647.3 MII Monitor Operation 65 668. Potential Trouble Sources 678.1 Adventures in Routing 688.2 Ethernet Device Renaming 698.3 Painfully Slow Or No Failed Link Detection By Miimon 70 719. SNMP agents 72 7310. Promiscuous mode 74 7511. Configuring Bonding for High Availability 7611.1 High Availability in a Single Switch Topology 7711.2 High Availability in a Multiple Switch Topology 7811.2.1 HA Bonding Mode Selection for Multiple Switch Topology 7911.2.2 HA Link Monitoring for Multiple Switch Topology 80 8112. Configuring Bonding for Maximum Throughput 8212.1 Maximum Throughput in a Single Switch Topology 8312.1.1 MT Bonding Mode Selection for Single Switch Topology 8412.1.2 MT Link Monitoring for Single Switch Topology 8512.2 Maximum Throughput in a Multiple Switch Topology 8612.2.1 MT Bonding Mode Selection for Multiple Switch Topology 8712.2.2 MT Link Monitoring for Multiple Switch Topology 88 8913. Switch Behavior Issues 9013.1 Link Establishment and Failover Delays 9113.2 Duplicated Incoming Packets 92 9314. Hardware Specific Considerations 9414.1 IBM BladeCenter 95 9615. Frequently Asked Questions 97 9816. Resources and Links 99 100 1011. Bonding Driver Installation 102============================== 103 104 Most popular distro kernels ship with the bonding driver 105already available as a module and the ifenslave user level control 106program installed and ready for use. If your distro does not, or you 107have need to compile bonding from source (e.g., configuring and 108installing a mainline kernel from kernel.org), you'll need to perform 109the following steps: 110 1111.1 Configure and build the kernel with bonding 112----------------------------------------------- 113 114 The current version of the bonding driver is available in the 115drivers/net/bonding subdirectory of the most recent kernel source 116(which is available on http://kernel.org). Most users "rolling their 117own" will want to use the most recent kernel from kernel.org. 118 119 Configure kernel with "make menuconfig" (or "make xconfig" or 120"make config"), then select "Bonding driver support" in the "Network 121device support" section. It is recommended that you configure the 122driver as module since it is currently the only way to pass parameters 123to the driver or configure more than one bonding device. 124 125 Build and install the new kernel and modules, then continue 126below to install ifenslave. 127 1281.2 Install ifenslave Control Utility 129------------------------------------- 130 131 The ifenslave user level control program is included in the 132kernel source tree, in the file Documentation/networking/ifenslave.c. 133It is generally recommended that you use the ifenslave that 134corresponds to the kernel that you are using (either from the same 135source tree or supplied with the distro), however, ifenslave 136executables from older kernels should function (but features newer 137than the ifenslave release are not supported). Running an ifenslave 138that is newer than the kernel is not supported, and may or may not 139work. 140 141 To install ifenslave, do the following: 142 143# gcc -Wall -O -I/usr/src/linux/include ifenslave.c -o ifenslave 144# cp ifenslave /sbin/ifenslave 145 146 If your kernel source is not in "/usr/src/linux," then replace 147"/usr/src/linux/include" in the above with the location of your kernel 148source include directory. 149 150 You may wish to back up any existing /sbin/ifenslave, or, for 151testing or informal use, tag the ifenslave to the kernel version 152(e.g., name the ifenslave executable /sbin/ifenslave-2.6.10). 153 154IMPORTANT NOTE: 155 156 If you omit the "-I" or specify an incorrect directory, you 157may end up with an ifenslave that is incompatible with the kernel 158you're trying to build it for. Some distros (e.g., Red Hat from 7.1 159onwards) do not have /usr/include/linux symbolically linked to the 160default kernel source include directory. 161 162SECOND IMPORTANT NOTE: 163 If you plan to configure bonding using sysfs, you do not need 164to use ifenslave. 165 1662. Bonding Driver Options 167========================= 168 169 Options for the bonding driver are supplied as parameters to 170the bonding module at load time. They may be given as command line 171arguments to the insmod or modprobe command, but are usually specified 172in either the /etc/modules.conf or /etc/modprobe.conf configuration 173file, or in a distro-specific configuration file (some of which are 174detailed in the next section). 175 176 The available bonding driver parameters are listed below. If a 177parameter is not specified the default value is used. When initially 178configuring a bond, it is recommended "tail -f /var/log/messages" be 179run in a separate window to watch for bonding driver error messages. 180 181 It is critical that either the miimon or arp_interval and 182arp_ip_target parameters be specified, otherwise serious network 183degradation will occur during link failures. Very few devices do not 184support at least miimon, so there is really no reason not to use it. 185 186 Options with textual values will accept either the text name 187or, for backwards compatibility, the option value. E.g., 188"mode=802.3ad" and "mode=4" set the same mode. 189 190 The parameters are as follows: 191 192arp_interval 193 194 Specifies the ARP link monitoring frequency in milliseconds. 195 196 The ARP monitor works by periodically checking the slave 197 devices to determine whether they have sent or received 198 traffic recently (the precise criteria depends upon the 199 bonding mode, and the state of the slave). Regular traffic is 200 generated via ARP probes issued for the addresses specified by 201 the arp_ip_target option. 202 203 This behavior can be modified by the arp_validate option, 204 below. 205 206 If ARP monitoring is used in an etherchannel compatible mode 207 (modes 0 and 2), the switch should be configured in a mode 208 that evenly distributes packets across all links. If the 209 switch is configured to distribute the packets in an XOR 210 fashion, all replies from the ARP targets will be received on 211 the same link which could cause the other team members to 212 fail. ARP monitoring should not be used in conjunction with 213 miimon. A value of 0 disables ARP monitoring. The default 214 value is 0. 215 216arp_ip_target 217 218 Specifies the IP addresses to use as ARP monitoring peers when 219 arp_interval is > 0. These are the targets of the ARP request 220 sent to determine the health of the link to the targets. 221 Specify these values in ddd.ddd.ddd.ddd format. Multiple IP 222 addresses must be separated by a comma. At least one IP 223 address must be given for ARP monitoring to function. The 224 maximum number of targets that can be specified is 16. The 225 default value is no IP addresses. 226 227arp_validate 228 229 Specifies whether or not ARP probes and replies should be 230 validated in the active-backup mode. This causes the ARP 231 monitor to examine the incoming ARP requests and replies, and 232 only consider a slave to be up if it is receiving the 233 appropriate ARP traffic. 234 235 Possible values are: 236 237 none or 0 238 239 No validation is performed. This is the default. 240 241 active or 1 242 243 Validation is performed only for the active slave. 244 245 backup or 2 246 247 Validation is performed only for backup slaves. 248 249 all or 3 250 251 Validation is performed for all slaves. 252 253 For the active slave, the validation checks ARP replies to 254 confirm that they were generated by an arp_ip_target. Since 255 backup slaves do not typically receive these replies, the 256 validation performed for backup slaves is on the ARP request 257 sent out via the active slave. It is possible that some 258 switch or network configurations may result in situations 259 wherein the backup slaves do not receive the ARP requests; in 260 such a situation, validation of backup slaves must be 261 disabled. 262 263 This option is useful in network configurations in which 264 multiple bonding hosts are concurrently issuing ARPs to one or 265 more targets beyond a common switch. Should the link between 266 the switch and target fail (but not the switch itself), the 267 probe traffic generated by the multiple bonding instances will 268 fool the standard ARP monitor into considering the links as 269 still up. Use of the arp_validate option can resolve this, as 270 the ARP monitor will only consider ARP requests and replies 271 associated with its own instance of bonding. 272 273 This option was added in bonding version 3.1.0. 274 275downdelay 276 277 Specifies the time, in milliseconds, to wait before disabling 278 a slave after a link failure has been detected. This option 279 is only valid for the miimon link monitor. The downdelay 280 value should be a multiple of the miimon value; if not, it 281 will be rounded down to the nearest multiple. The default 282 value is 0. 283 284lacp_rate 285 286 Option specifying the rate in which we'll ask our link partner 287 to transmit LACPDU packets in 802.3ad mode. Possible values 288 are: 289 290 slow or 0 291 Request partner to transmit LACPDUs every 30 seconds 292 293 fast or 1 294 Request partner to transmit LACPDUs every 1 second 295 296 The default is slow. 297 298max_bonds 299 300 Specifies the number of bonding devices to create for this 301 instance of the bonding driver. E.g., if max_bonds is 3, and 302 the bonding driver is not already loaded, then bond0, bond1 303 and bond2 will be created. The default value is 1. 304 305miimon 306 307 Specifies the MII link monitoring frequency in milliseconds. 308 This determines how often the link state of each slave is 309 inspected for link failures. A value of zero disables MII 310 link monitoring. A value of 100 is a good starting point. 311 The use_carrier option, below, affects how the link state is 312 determined. See the High Availability section for additional 313 information. The default value is 0. 314 315mode 316 317 Specifies one of the bonding policies. The default is 318 balance-rr (round robin). Possible values are: 319 320 balance-rr or 0 321 322 Round-robin policy: Transmit packets in sequential 323 order from the first available slave through the 324 last. This mode provides load balancing and fault 325 tolerance. 326 327 active-backup or 1 328 329 Active-backup policy: Only one slave in the bond is 330 active. A different slave becomes active if, and only 331 if, the active slave fails. The bond's MAC address is 332 externally visible on only one port (network adapter) 333 to avoid confusing the switch. 334 335 In bonding version 2.6.2 or later, when a failover 336 occurs in active-backup mode, bonding will issue one 337 or more gratuitous ARPs on the newly active slave. 338 One gratuitous ARP is issued for the bonding master 339 interface and each VLAN interfaces configured above 340 it, provided that the interface has at least one IP 341 address configured. Gratuitous ARPs issued for VLAN 342 interfaces are tagged with the appropriate VLAN id. 343 344 This mode provides fault tolerance. The primary 345 option, documented below, affects the behavior of this 346 mode. 347 348 balance-xor or 2 349 350 XOR policy: Transmit based on the selected transmit 351 hash policy. The default policy is a simple [(source 352 MAC address XOR'd with destination MAC address) modulo 353 slave count]. Alternate transmit policies may be 354 selected via the xmit_hash_policy option, described 355 below. 356 357 This mode provides load balancing and fault tolerance. 358 359 broadcast or 3 360 361 Broadcast policy: transmits everything on all slave 362 interfaces. This mode provides fault tolerance. 363 364 802.3ad or 4 365 366 IEEE 802.3ad Dynamic link aggregation. Creates 367 aggregation groups that share the same speed and 368 duplex settings. Utilizes all slaves in the active 369 aggregator according to the 802.3ad specification. 370 371 Slave selection for outgoing traffic is done according 372 to the transmit hash policy, which may be changed from 373 the default simple XOR policy via the xmit_hash_policy 374 option, documented below. Note that not all transmit 375 policies may be 802.3ad compliant, particularly in 376 regards to the packet mis-ordering requirements of 377 section 43.2.4 of the 802.3ad standard. Differing 378 peer implementations will have varying tolerances for 379 noncompliance. 380 381 Prerequisites: 382 383 1. Ethtool support in the base drivers for retrieving 384 the speed and duplex of each slave. 385 386 2. A switch that supports IEEE 802.3ad Dynamic link 387 aggregation. 388 389 Most switches will require some type of configuration 390 to enable 802.3ad mode. 391 392 balance-tlb or 5 393 394 Adaptive transmit load balancing: channel bonding that 395 does not require any special switch support. The 396 outgoing traffic is distributed according to the 397 current load (computed relative to the speed) on each 398 slave. Incoming traffic is received by the current 399 slave. If the receiving slave fails, another slave 400 takes over the MAC address of the failed receiving 401 slave. 402 403 Prerequisite: 404 405 Ethtool support in the base drivers for retrieving the 406 speed of each slave. 407 408 balance-alb or 6 409 410 Adaptive load balancing: includes balance-tlb plus 411 receive load balancing (rlb) for IPV4 traffic, and 412 does not require any special switch support. The 413 receive load balancing is achieved by ARP negotiation. 414 The bonding driver intercepts the ARP Replies sent by 415 the local system on their way out and overwrites the 416 source hardware address with the unique hardware 417 address of one of the slaves in the bond such that 418 different peers use different hardware addresses for 419 the server. 420 421 Receive traffic from connections created by the server 422 is also balanced. When the local system sends an ARP 423 Request the bonding driver copies and saves the peer's 424 IP information from the ARP packet. When the ARP 425 Reply arrives from the peer, its hardware address is 426 retrieved and the bonding driver initiates an ARP 427 reply to this peer assigning it to one of the slaves 428 in the bond. A problematic outcome of using ARP 429 negotiation for balancing is that each time that an 430 ARP request is broadcast it uses the hardware address 431 of the bond. Hence, peers learn the hardware address 432 of the bond and the balancing of receive traffic 433 collapses to the current slave. This is handled by 434 sending updates (ARP Replies) to all the peers with 435 their individually assigned hardware address such that 436 the traffic is redistributed. Receive traffic is also 437 redistributed when a new slave is added to the bond 438 and when an inactive slave is re-activated. The 439 receive load is distributed sequentially (round robin) 440 among the group of highest speed slaves in the bond. 441 442 When a link is reconnected or a new slave joins the 443 bond the receive traffic is redistributed among all 444 active slaves in the bond by initiating ARP Replies 445 with the selected MAC address to each of the 446 clients. The updelay parameter (detailed below) must 447 be set to a value equal or greater than the switch's 448 forwarding delay so that the ARP Replies sent to the 449 peers will not be blocked by the switch. 450 451 Prerequisites: 452 453 1. Ethtool support in the base drivers for retrieving 454 the speed of each slave. 455 456 2. Base driver support for setting the hardware 457 address of a device while it is open. This is 458 required so that there will always be one slave in the 459 team using the bond hardware address (the 460 curr_active_slave) while having a unique hardware 461 address for each slave in the bond. If the 462 curr_active_slave fails its hardware address is 463 swapped with the new curr_active_slave that was 464 chosen. 465 466primary 467 468 A string (eth0, eth2, etc) specifying which slave is the 469 primary device. The specified device will always be the 470 active slave while it is available. Only when the primary is 471 off-line will alternate devices be used. This is useful when 472 one slave is preferred over another, e.g., when one slave has 473 higher throughput than another. 474 475 The primary option is only valid for active-backup mode. 476 477updelay 478 479 Specifies the time, in milliseconds, to wait before enabling a 480 slave after a link recovery has been detected. This option is 481 only valid for the miimon link monitor. The updelay value 482 should be a multiple of the miimon value; if not, it will be 483 rounded down to the nearest multiple. The default value is 0. 484 485use_carrier 486 487 Specifies whether or not miimon should use MII or ETHTOOL 488 ioctls vs. netif_carrier_ok() to determine the link 489 status. The MII or ETHTOOL ioctls are less efficient and 490 utilize a deprecated calling sequence within the kernel. The 491 netif_carrier_ok() relies on the device driver to maintain its 492 state with netif_carrier_on/off; at this writing, most, but 493 not all, device drivers support this facility. 494 495 If bonding insists that the link is up when it should not be, 496 it may be that your network device driver does not support 497 netif_carrier_on/off. The default state for netif_carrier is 498 "carrier on," so if a driver does not support netif_carrier, 499 it will appear as if the link is always up. In this case, 500 setting use_carrier to 0 will cause bonding to revert to the 501 MII / ETHTOOL ioctl method to determine the link state. 502 503 A value of 1 enables the use of netif_carrier_ok(), a value of 504 0 will use the deprecated MII / ETHTOOL ioctls. The default 505 value is 1. 506 507xmit_hash_policy 508 509 Selects the transmit hash policy to use for slave selection in 510 balance-xor and 802.3ad modes. Possible values are: 511 512 layer2 513 514 Uses XOR of hardware MAC addresses to generate the 515 hash. The formula is 516 517 (source MAC XOR destination MAC) modulo slave count 518 519 This algorithm will place all traffic to a particular 520 network peer on the same slave. 521 522 This algorithm is 802.3ad compliant. 523 524 layer3+4 525 526 This policy uses upper layer protocol information, 527 when available, to generate the hash. This allows for 528 traffic to a particular network peer to span multiple 529 slaves, although a single connection will not span 530 multiple slaves. 531 532 The formula for unfragmented TCP and UDP packets is 533 534 ((source port XOR dest port) XOR 535 ((source IP XOR dest IP) AND 0xffff) 536 modulo slave count 537 538 For fragmented TCP or UDP packets and all other IP 539 protocol traffic, the source and destination port 540 information is omitted. For non-IP traffic, the 541 formula is the same as for the layer2 transmit hash 542 policy. 543 544 This policy is intended to mimic the behavior of 545 certain switches, notably Cisco switches with PFC2 as 546 well as some Foundry and IBM products. 547 548 This algorithm is not fully 802.3ad compliant. A 549 single TCP or UDP conversation containing both 550 fragmented and unfragmented packets will see packets 551 striped across two interfaces. This may result in out 552 of order delivery. Most traffic types will not meet 553 this criteria, as TCP rarely fragments traffic, and 554 most UDP traffic is not involved in extended 555 conversations. Other implementations of 802.3ad may 556 or may not tolerate this noncompliance. 557 558 The default value is layer2. This option was added in bonding 559version 2.6.3. In earlier versions of bonding, this parameter does 560not exist, and the layer2 policy is the only policy. 561 562 5633. Configuring Bonding Devices 564============================== 565 566 You can configure bonding using either your distro's network 567initialization scripts, or manually using either ifenslave or the 568sysfs interface. Distros generally use one of two packages for the 569network initialization scripts: initscripts or sysconfig. Recent 570versions of these packages have support for bonding, while older 571versions do not. 572 573 We will first describe the options for configuring bonding for 574distros using versions of initscripts and sysconfig with full or 575partial support for bonding, then provide information on enabling 576bonding without support from the network initialization scripts (i.e., 577older versions of initscripts or sysconfig). 578 579 If you're unsure whether your distro uses sysconfig or 580initscripts, or don't know if it's new enough, have no fear. 581Determining this is fairly straightforward. 582 583 First, issue the command: 584 585$ rpm -qf /sbin/ifup 586 587 It will respond with a line of text starting with either 588"initscripts" or "sysconfig," followed by some numbers. This is the 589package that provides your network initialization scripts. 590 591 Next, to determine if your installation supports bonding, 592issue the command: 593 594$ grep ifenslave /sbin/ifup 595 596 If this returns any matches, then your initscripts or 597sysconfig has support for bonding. 598 5993.1 Configuration with Sysconfig Support 600---------------------------------------- 601 602 This section applies to distros using a version of sysconfig 603with bonding support, for example, SuSE Linux Enterprise Server 9. 604 605 SuSE SLES 9's networking configuration system does support 606bonding, however, at this writing, the YaST system configuration 607front end does not provide any means to work with bonding devices. 608Bonding devices can be managed by hand, however, as follows. 609 610 First, if they have not already been configured, configure the 611slave devices. On SLES 9, this is most easily done by running the 612yast2 sysconfig configuration utility. The goal is for to create an 613ifcfg-id file for each slave device. The simplest way to accomplish 614this is to configure the devices for DHCP (this is only to get the 615file ifcfg-id file created; see below for some issues with DHCP). The 616name of the configuration file for each device will be of the form: 617 618ifcfg-id-xx:xx:xx:xx:xx:xx 619 620 Where the "xx" portion will be replaced with the digits from 621the device's permanent MAC address. 622 623 Once the set of ifcfg-id-xx:xx:xx:xx:xx:xx files has been 624created, it is necessary to edit the configuration files for the slave 625devices (the MAC addresses correspond to those of the slave devices). 626Before editing, the file will contain multiple lines, and will look 627something like this: 628 629BOOTPROTO='dhcp' 630STARTMODE='on' 631USERCTL='no' 632UNIQUE='XNzu.WeZGOGF+4wE' 633_nm_name='bus-pci-0001:61:01.0' 634 635 Change the BOOTPROTO and STARTMODE lines to the following: 636 637BOOTPROTO='none' 638STARTMODE='off' 639 640 Do not alter the UNIQUE or _nm_name lines. Remove any other 641lines (USERCTL, etc). 642 643 Once the ifcfg-id-xx:xx:xx:xx:xx:xx files have been modified, 644it's time to create the configuration file for the bonding device 645itself. This file is named ifcfg-bondX, where X is the number of the 646bonding device to create, starting at 0. The first such file is 647ifcfg-bond0, the second is ifcfg-bond1, and so on. The sysconfig 648network configuration system will correctly start multiple instances 649of bonding. 650 651 The contents of the ifcfg-bondX file is as follows: 652 653BOOTPROTO="static" 654BROADCAST="10.0.2.255" 655IPADDR="10.0.2.10" 656NETMASK="255.255.0.0" 657NETWORK="10.0.2.0" 658REMOTE_IPADDR="" 659STARTMODE="onboot" 660BONDING_MASTER="yes" 661BONDING_MODULE_OPTS="mode=active-backup miimon=100" 662BONDING_SLAVE0="eth0" 663BONDING_SLAVE1="bus-pci-0000:06:08.1" 664 665 Replace the sample BROADCAST, IPADDR, NETMASK and NETWORK 666values with the appropriate values for your network. 667 668 The STARTMODE specifies when the device is brought online. 669The possible values are: 670 671 onboot: The device is started at boot time. If you're not 672 sure, this is probably what you want. 673 674 manual: The device is started only when ifup is called 675 manually. Bonding devices may be configured this 676 way if you do not wish them to start automatically 677 at boot for some reason. 678 679 hotplug: The device is started by a hotplug event. This is not 680 a valid choice for a bonding device. 681 682 off or ignore: The device configuration is ignored. 683 684 The line BONDING_MASTER='yes' indicates that the device is a 685bonding master device. The only useful value is "yes." 686 687 The contents of BONDING_MODULE_OPTS are supplied to the 688instance of the bonding module for this device. Specify the options 689for the bonding mode, link monitoring, and so on here. Do not include 690the max_bonds bonding parameter; this will confuse the configuration 691system if you have multiple bonding devices. 692 693 Finally, supply one BONDING_SLAVEn="slave device" for each 694slave. where "n" is an increasing value, one for each slave. The 695"slave device" is either an interface name, e.g., "eth0", or a device 696specifier for the network device. The interface name is easier to 697find, but the ethN names are subject to change at boot time if, e.g., 698a device early in the sequence has failed. The device specifiers 699(bus-pci-0000:06:08.1 in the example above) specify the physical 700network device, and will not change unless the device's bus location 701changes (for example, it is moved from one PCI slot to another). The 702example above uses one of each type for demonstration purposes; most 703configurations will choose one or the other for all slave devices. 704 705 When all configuration files have been modified or created, 706networking must be restarted for the configuration changes to take 707effect. This can be accomplished via the following: 708 709# /etc/init.d/network restart 710 711 Note that the network control script (/sbin/ifdown) will 712remove the bonding module as part of the network shutdown processing, 713so it is not necessary to remove the module by hand if, e.g., the 714module parameters have changed. 715 716 Also, at this writing, YaST/YaST2 will not manage bonding 717devices (they do not show bonding interfaces on its list of network 718devices). It is necessary to edit the configuration file by hand to 719change the bonding configuration. 720 721 Additional general options and details of the ifcfg file 722format can be found in an example ifcfg template file: 723 724/etc/sysconfig/network/ifcfg.template 725 726 Note that the template does not document the various BONDING_ 727settings described above, but does describe many of the other options. 728 7293.1.1 Using DHCP with Sysconfig 730------------------------------- 731 732 Under sysconfig, configuring a device with BOOTPROTO='dhcp' 733will cause it to query DHCP for its IP address information. At this 734writing, this does not function for bonding devices; the scripts 735attempt to obtain the device address from DHCP prior to adding any of 736the slave devices. Without active slaves, the DHCP requests are not 737sent to the network. 738 7393.1.2 Configuring Multiple Bonds with Sysconfig 740----------------------------------------------- 741 742 The sysconfig network initialization system is capable of 743handling multiple bonding devices. All that is necessary is for each 744bonding instance to have an appropriately configured ifcfg-bondX file 745(as described above). Do not specify the "max_bonds" parameter to any 746instance of bonding, as this will confuse sysconfig. If you require 747multiple bonding devices with identical parameters, create multiple 748ifcfg-bondX files. 749 750 Because the sysconfig scripts supply the bonding module 751options in the ifcfg-bondX file, it is not necessary to add them to 752the system /etc/modules.conf or /etc/modprobe.conf configuration file. 753 7543.2 Configuration with Initscripts Support 755------------------------------------------ 756 757 This section applies to distros using a version of initscripts 758with bonding support, for example, Red Hat Linux 9 or Red Hat 759Enterprise Linux version 3 or 4. On these systems, the network 760initialization scripts have some knowledge of bonding, and can be 761configured to control bonding devices. 762 763 These distros will not automatically load the network adapter 764driver unless the ethX device is configured with an IP address. 765Because of this constraint, users must manually configure a 766network-script file for all physical adapters that will be members of 767a bondX link. Network script files are located in the directory: 768 769/etc/sysconfig/network-scripts 770 771 The file name must be prefixed with "ifcfg-eth" and suffixed 772with the adapter's physical adapter number. For example, the script 773for eth0 would be named /etc/sysconfig/network-scripts/ifcfg-eth0. 774Place the following text in the file: 775 776DEVICE=eth0 777USERCTL=no 778ONBOOT=yes 779MASTER=bond0 780SLAVE=yes 781BOOTPROTO=none 782 783 The DEVICE= line will be different for every ethX device and 784must correspond with the name of the file, i.e., ifcfg-eth1 must have 785a device line of DEVICE=eth1. The setting of the MASTER= line will 786also depend on the final bonding interface name chosen for your bond. 787As with other network devices, these typically start at 0, and go up 788one for each device, i.e., the first bonding instance is bond0, the 789second is bond1, and so on. 790 791 Next, create a bond network script. The file name for this 792script will be /etc/sysconfig/network-scripts/ifcfg-bondX where X is 793the number of the bond. For bond0 the file is named "ifcfg-bond0", 794for bond1 it is named "ifcfg-bond1", and so on. Within that file, 795place the following text: 796 797DEVICE=bond0 798IPADDR=192.168.1.1 799NETMASK=255.255.255.0 800NETWORK=192.168.1.0 801BROADCAST=192.168.1.255 802ONBOOT=yes 803BOOTPROTO=none 804USERCTL=no 805 806 Be sure to change the networking specific lines (IPADDR, 807NETMASK, NETWORK and BROADCAST) to match your network configuration. 808 809 Finally, it is necessary to edit /etc/modules.conf (or 810/etc/modprobe.conf, depending upon your distro) to load the bonding 811module with your desired options when the bond0 interface is brought 812up. The following lines in /etc/modules.conf (or modprobe.conf) will 813load the bonding module, and select its options: 814 815alias bond0 bonding 816options bond0 mode=balance-alb miimon=100 817 818 Replace the sample parameters with the appropriate set of 819options for your configuration. 820 821 Finally run "/etc/rc.d/init.d/network restart" as root. This 822will restart the networking subsystem and your bond link should be now 823up and running. 824 8253.2.1 Using DHCP with Initscripts 826--------------------------------- 827 828 Recent versions of initscripts (the version supplied with 829Fedora Core 3 and Red Hat Enterprise Linux 4 is reported to work) do 830have support for assigning IP information to bonding devices via DHCP. 831 832 To configure bonding for DHCP, configure it as described 833above, except replace the line "BOOTPROTO=none" with "BOOTPROTO=dhcp" 834and add a line consisting of "TYPE=Bonding". Note that the TYPE value 835is case sensitive. 836 8373.2.2 Configuring Multiple Bonds with Initscripts 838------------------------------------------------- 839 840 At this writing, the initscripts package does not directly 841support loading the bonding driver multiple times, so the process for 842doing so is the same as described in the "Configuring Multiple Bonds 843Manually" section, below. 844 845 NOTE: It has been observed that some Red Hat supplied kernels 846are apparently unable to rename modules at load time (the "-o bond1" 847part). Attempts to pass that option to modprobe will produce an 848"Operation not permitted" error. This has been reported on some 849Fedora Core kernels, and has been seen on RHEL 4 as well. On kernels 850exhibiting this problem, it will be impossible to configure multiple 851bonds with differing parameters. 852 8533.3 Configuring Bonding Manually with Ifenslave 854----------------------------------------------- 855 856 This section applies to distros whose network initialization 857scripts (the sysconfig or initscripts package) do not have specific 858knowledge of bonding. One such distro is SuSE Linux Enterprise Server 859version 8. 860 861 The general method for these systems is to place the bonding 862module parameters into /etc/modules.conf or /etc/modprobe.conf (as 863appropriate for the installed distro), then add modprobe and/or 864ifenslave commands to the system's global init script. The name of 865the global init script differs; for sysconfig, it is 866/etc/init.d/boot.local and for initscripts it is /etc/rc.d/rc.local. 867 868 For example, if you wanted to make a simple bond of two e100 869devices (presumed to be eth0 and eth1), and have it persist across 870reboots, edit the appropriate file (/etc/init.d/boot.local or 871/etc/rc.d/rc.local), and add the following: 872 873modprobe bonding mode=balance-alb miimon=100 874modprobe e100 875ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up 876ifenslave bond0 eth0 877ifenslave bond0 eth1 878 879 Replace the example bonding module parameters and bond0 880network configuration (IP address, netmask, etc) with the appropriate 881values for your configuration. 882 883 Unfortunately, this method will not provide support for the 884ifup and ifdown scripts on the bond devices. To reload the bonding 885configuration, it is necessary to run the initialization script, e.g., 886 887# /etc/init.d/boot.local 888 889 or 890 891# /etc/rc.d/rc.local 892 893 It may be desirable in such a case to create a separate script 894which only initializes the bonding configuration, then call that 895separate script from within boot.local. This allows for bonding to be 896enabled without re-running the entire global init script. 897 898 To shut down the bonding devices, it is necessary to first 899mark the bonding device itself as being down, then remove the 900appropriate device driver modules. For our example above, you can do 901the following: 902 903# ifconfig bond0 down 904# rmmod bonding 905# rmmod e100 906 907 Again, for convenience, it may be desirable to create a script 908with these commands. 909 910 9113.3.1 Configuring Multiple Bonds Manually 912----------------------------------------- 913 914 This section contains information on configuring multiple 915bonding devices with differing options for those systems whose network 916initialization scripts lack support for configuring multiple bonds. 917 918 If you require multiple bonding devices, but all with the same 919options, you may wish to use the "max_bonds" module parameter, 920documented above. 921 922 To create multiple bonding devices with differing options, it 923is necessary to use bonding parameters exported by sysfs, documented 924in the section below. 925 926 9273.4 Configuring Bonding Manually via Sysfs 928------------------------------------------ 929 930 Starting with version 3.0, Channel Bonding may be configured 931via the sysfs interface. This interface allows dynamic configuration 932of all bonds in the system without unloading the module. It also 933allows for adding and removing bonds at runtime. Ifenslave is no 934longer required, though it is still supported. 935 936 Use of the sysfs interface allows you to use multiple bonds 937with different configurations without having to reload the module. 938It also allows you to use multiple, differently configured bonds when 939bonding is compiled into the kernel. 940 941 You must have the sysfs filesystem mounted to configure 942bonding this way. The examples in this document assume that you 943are using the standard mount point for sysfs, e.g. /sys. If your 944sysfs filesystem is mounted elsewhere, you will need to adjust the 945example paths accordingly. 946 947Creating and Destroying Bonds 948----------------------------- 949To add a new bond foo: 950# echo +foo > /sys/class/net/bonding_masters 951 952To remove an existing bond bar: 953# echo -bar > /sys/class/net/bonding_masters 954 955To show all existing bonds: 956# cat /sys/class/net/bonding_masters 957 958NOTE: due to 4K size limitation of sysfs files, this list may be 959truncated if you have more than a few hundred bonds. This is unlikely 960to occur under normal operating conditions. 961 962Adding and Removing Slaves 963-------------------------- 964 Interfaces may be enslaved to a bond using the file 965/sys/class/net/<bond>/bonding/slaves. The semantics for this file 966are the same as for the bonding_masters file. 967 968To enslave interface eth0 to bond bond0: 969# ifconfig bond0 up 970# echo +eth0 > /sys/class/net/bond0/bonding/slaves 971 972To free slave eth0 from bond bond0: 973# echo -eth0 > /sys/class/net/bond0/bonding/slaves 974 975 NOTE: The bond must be up before slaves can be added. All 976slaves are freed when the interface is brought down. 977 978 When an interface is enslaved to a bond, symlinks between the 979two are created in the sysfs filesystem. In this case, you would get 980/sys/class/net/bond0/slave_eth0 pointing to /sys/class/net/eth0, and 981/sys/class/net/eth0/master pointing to /sys/class/net/bond0. 982 983 This means that you can tell quickly whether or not an 984interface is enslaved by looking for the master symlink. Thus: 985# echo -eth0 > /sys/class/net/eth0/master/bonding/slaves 986will free eth0 from whatever bond it is enslaved to, regardless of 987the name of the bond interface. 988 989Changing a Bond's Configuration 990------------------------------- 991 Each bond may be configured individually by manipulating the 992files located in /sys/class/net/<bond name>/bonding 993 994 The names of these files correspond directly with the command- 995line parameters described elsewhere in this file, and, with the 996exception of arp_ip_target, they accept the same values. To see the 997current setting, simply cat the appropriate file. 998 999 A few examples will be given here; for specific usage 1000guidelines for each parameter, see the appropriate section in this 1001document. 1002 1003To configure bond0 for balance-alb mode: 1004# ifconfig bond0 down 1005# echo 6 > /sys/class/net/bond0/bonding/mode 1006 - or - 1007# echo balance-alb > /sys/class/net/bond0/bonding/mode 1008 NOTE: The bond interface must be down before the mode can be 1009changed. 1010 1011To enable MII monitoring on bond0 with a 1 second interval: 1012# echo 1000 > /sys/class/net/bond0/bonding/miimon 1013 NOTE: If ARP monitoring is enabled, it will disabled when MII 1014monitoring is enabled, and vice-versa. 1015 1016To add ARP targets: 1017# echo +192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target 1018# echo +192.168.0.101 > /sys/class/net/bond0/bonding/arp_ip_target 1019 NOTE: up to 10 target addresses may be specified. 1020 1021To remove an ARP target: 1022# echo -192.168.0.100 > /sys/class/net/bond0/bonding/arp_ip_target 1023 1024Example Configuration 1025--------------------- 1026 We begin with the same example that is shown in section 3.3, 1027executed with sysfs, and without using ifenslave. 1028 1029 To make a simple bond of two e100 devices (presumed to be eth0 1030and eth1), and have it persist across reboots, edit the appropriate 1031file (/etc/init.d/boot.local or /etc/rc.d/rc.local), and add the 1032following: 1033 1034modprobe bonding 1035modprobe e100 1036echo balance-alb > /sys/class/net/bond0/bonding/mode 1037ifconfig bond0 192.168.1.1 netmask 255.255.255.0 up 1038echo 100 > /sys/class/net/bond0/bonding/miimon 1039echo +eth0 > /sys/class/net/bond0/bonding/slaves 1040echo +eth1 > /sys/class/net/bond0/bonding/slaves 1041 1042 To add a second bond, with two e1000 interfaces in 1043active-backup mode, using ARP monitoring, add the following lines to 1044your init script: 1045 1046modprobe e1000 1047echo +bond1 > /sys/class/net/bonding_masters 1048echo active-backup > /sys/class/net/bond1/bonding/mode 1049ifconfig bond1 192.168.2.1 netmask 255.255.255.0 up 1050echo +192.168.2.100 /sys/class/net/bond1/bonding/arp_ip_target 1051echo 2000 > /sys/class/net/bond1/bonding/arp_interval 1052echo +eth2 > /sys/class/net/bond1/bonding/slaves 1053echo +eth3 > /sys/class/net/bond1/bonding/slaves 1054 1055 10564. Querying Bonding Configuration 1057================================= 1058 10594.1 Bonding Configuration 1060------------------------- 1061 1062 Each bonding device has a read-only file residing in the 1063/proc/net/bonding directory. The file contents include information 1064about the bonding configuration, options and state of each slave. 1065 1066 For example, the contents of /proc/net/bonding/bond0 after the 1067driver is loaded with parameters of mode=0 and miimon=1000 is 1068generally as follows: 1069 1070 Ethernet Channel Bonding Driver: 2.6.1 (October 29, 2004) 1071 Bonding Mode: load balancing (round-robin) 1072 Currently Active Slave: eth0 1073 MII Status: up 1074 MII Polling Interval (ms): 1000 1075 Up Delay (ms): 0 1076 Down Delay (ms): 0 1077 1078 Slave Interface: eth1 1079 MII Status: up 1080 Link Failure Count: 1 1081 1082 Slave Interface: eth0 1083 MII Status: up 1084 Link Failure Count: 1 1085 1086 The precise format and contents will change depending upon the 1087bonding configuration, state, and version of the bonding driver. 1088 10894.2 Network configuration 1090------------------------- 1091 1092 The network configuration can be inspected using the ifconfig 1093command. Bonding devices will have the MASTER flag set; Bonding slave 1094devices will have the SLAVE flag set. The ifconfig output does not 1095contain information on which slaves are associated with which masters. 1096 1097 In the example below, the bond0 interface is the master 1098(MASTER) while eth0 and eth1 are slaves (SLAVE). Notice all slaves of 1099bond0 have the same MAC address (HWaddr) as bond0 for all modes except 1100TLB and ALB that require a unique MAC address for each slave. 1101 1102# /sbin/ifconfig 1103bond0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4 1104 inet addr:XXX.XXX.XXX.YYY Bcast:XXX.XXX.XXX.255 Mask:255.255.252.0 1105 UP BROADCAST RUNNING MASTER MULTICAST MTU:1500 Metric:1 1106 RX packets:7224794 errors:0 dropped:0 overruns:0 frame:0 1107 TX packets:3286647 errors:1 dropped:0 overruns:1 carrier:0 1108 collisions:0 txqueuelen:0 1109 1110eth0 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4 1111 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1 1112 RX packets:3573025 errors:0 dropped:0 overruns:0 frame:0 1113 TX packets:1643167 errors:1 dropped:0 overruns:1 carrier:0 1114 collisions:0 txqueuelen:100 1115 Interrupt:10 Base address:0x1080 1116 1117eth1 Link encap:Ethernet HWaddr 00:C0:F0:1F:37:B4 1118 UP BROADCAST RUNNING SLAVE MULTICAST MTU:1500 Metric:1 1119 RX packets:3651769 errors:0 dropped:0 overruns:0 frame:0 1120 TX packets:1643480 errors:0 dropped:0 overruns:0 carrier:0 1121 collisions:0 txqueuelen:100 1122 Interrupt:9 Base address:0x1400 1123 11245. Switch Configuration 1125======================= 1126 1127 For this section, "switch" refers to whatever system the 1128bonded devices are directly connected to (i.e., where the other end of 1129the cable plugs into). This may be an actual dedicated switch device, 1130or it may be another regular system (e.g., another computer running 1131Linux), 1132 1133 The active-backup, balance-tlb and balance-alb modes do not 1134require any specific configuration of the switch. 1135 1136 The 802.3ad mode requires that the switch have the appropriate 1137ports configured as an 802.3ad aggregation. The precise method used 1138to configure this varies from switch to switch, but, for example, a 1139Cisco 3550 series switch requires that the appropriate ports first be 1140grouped together in a single etherchannel instance, then that 1141etherchannel is set to mode "lacp" to enable 802.3ad (instead of 1142standard EtherChannel). 1143 1144 The balance-rr, balance-xor and broadcast modes generally 1145require that the switch have the appropriate ports grouped together. 1146The nomenclature for such a group differs between switches, it may be 1147called an "etherchannel" (as in the Cisco example, above), a "trunk 1148group" or some other similar variation. For these modes, each switch 1149will also have its own configuration options for the switch's transmit 1150policy to the bond. Typical choices include XOR of either the MAC or 1151IP addresses. The transmit policy of the two peers does not need to 1152match. For these three modes, the bonding mode really selects a 1153transmit policy for an EtherChannel group; all three will interoperate 1154with another EtherChannel group. 1155 1156 11576. 802.1q VLAN Support 1158====================== 1159 1160 It is possible to configure VLAN devices over a bond interface 1161using the 8021q driver. However, only packets coming from the 8021q 1162driver and passing through bonding will be tagged by default. Self 1163generated packets, for example, bonding's learning packets or ARP 1164packets generated by either ALB mode or the ARP monitor mechanism, are 1165tagged internally by bonding itself. As a result, bonding must 1166"learn" the VLAN IDs configured above it, and use those IDs to tag 1167self generated packets. 1168 1169 For reasons of simplicity, and to support the use of adapters 1170that can do VLAN hardware acceleration offloading, the bonding 1171interface declares itself as fully hardware offloading capable, it gets 1172the add_vid/kill_vid notifications to gather the necessary 1173information, and it propagates those actions to the slaves. In case 1174of mixed adapter types, hardware accelerated tagged packets that 1175should go through an adapter that is not offloading capable are 1176"un-accelerated" by the bonding driver so the VLAN tag sits in the 1177regular location. 1178 1179 VLAN interfaces *must* be added on top of a bonding interface 1180only after enslaving at least one slave. The bonding interface has a 1181hardware address of 00:00:00:00:00:00 until the first slave is added. 1182If the VLAN interface is created prior to the first enslavement, it 1183would pick up the all-zeroes hardware address. Once the first slave 1184is attached to the bond, the bond device itself will pick up the 1185slave's hardware address, which is then available for the VLAN device. 1186 1187 Also, be aware that a similar problem can occur if all slaves 1188are released from a bond that still has one or more VLAN interfaces on 1189top of it. When a new slave is added, the bonding interface will 1190obtain its hardware address from the first slave, which might not 1191match the hardware address of the VLAN interfaces (which was 1192ultimately copied from an earlier slave). 1193 1194 There are two methods to insure that the VLAN device operates 1195with the correct hardware address if all slaves are removed from a 1196bond interface: 1197 1198 1. Remove all VLAN interfaces then recreate them 1199 1200 2. Set the bonding interface's hardware address so that it 1201matches the hardware address of the VLAN interfaces. 1202 1203 Note that changing a VLAN interface's HW address would set the 1204underlying device -- i.e. the bonding interface -- to promiscuous 1205mode, which might not be what you want. 1206 1207 12087. Link Monitoring 1209================== 1210 1211 The bonding driver at present supports two schemes for 1212monitoring a slave device's link state: the ARP monitor and the MII 1213monitor. 1214 1215 At the present time, due to implementation restrictions in the 1216bonding driver itself, it is not possible to enable both ARP and MII 1217monitoring simultaneously. 1218 12197.1 ARP Monitor Operation 1220------------------------- 1221 1222 The ARP monitor operates as its name suggests: it sends ARP 1223queries to one or more designated peer systems on the network, and 1224uses the response as an indication that the link is operating. This 1225gives some assurance that traffic is actually flowing to and from one 1226or more peers on the local network. 1227 1228 The ARP monitor relies on the device driver itself to verify 1229that traffic is flowing. In particular, the driver must keep up to 1230date the last receive time, dev->last_rx, and transmit start time, 1231dev->trans_start. If these are not updated by the driver, then the 1232ARP monitor will immediately fail any slaves using that driver, and 1233those slaves will stay down. If networking monitoring (tcpdump, etc) 1234shows the ARP requests and replies on the network, then it may be that 1235your device driver is not updating last_rx and trans_start. 1236 12377.2 Configuring Multiple ARP Targets 1238------------------------------------ 1239 1240 While ARP monitoring can be done with just one target, it can 1241be useful in a High Availability setup to have several targets to 1242monitor. In the case of just one target, the target itself may go 1243down or have a problem making it unresponsive to ARP requests. Having 1244an additional target (or several) increases the reliability of the ARP 1245monitoring. 1246 1247 Multiple ARP targets must be separated by commas as follows: 1248 1249# example options for ARP monitoring with three targets 1250alias bond0 bonding 1251options bond0 arp_interval=60 arp_ip_target=192.168.0.1,192.168.0.3,192.168.0.9 1252 1253 For just a single target the options would resemble: 1254 1255# example options for ARP monitoring with one target 1256alias bond0 bonding 1257options bond0 arp_interval=60 arp_ip_target=192.168.0.100 1258 1259 12607.3 MII Monitor Operation 1261------------------------- 1262 1263 The MII monitor monitors only the carrier state of the local 1264network interface. It accomplishes this in one of three ways: by 1265depending upon the device driver to maintain its carrier state, by 1266querying the device's MII registers, or by making an ethtool query to 1267the device. 1268 1269 If the use_carrier module parameter is 1 (the default value), 1270then the MII monitor will rely on the driver for carrier state 1271information (via the netif_carrier subsystem). As explained in the 1272use_carrier parameter information, above, if the MII monitor fails to 1273detect carrier loss on the device (e.g., when the cable is physically 1274disconnected), it may be that the driver does not support 1275netif_carrier. 1276 1277 If use_carrier is 0, then the MII monitor will first query the 1278device's (via ioctl) MII registers and check the link state. If that 1279request fails (not just that it returns carrier down), then the MII 1280monitor will make an ethtool ETHOOL_GLINK request to attempt to obtain 1281the same information. If both methods fail (i.e., the driver either 1282does not support or had some error in processing both the MII register 1283and ethtool requests), then the MII monitor will assume the link is 1284up. 1285 12868. Potential Sources of Trouble 1287=============================== 1288 12898.1 Adventures in Routing 1290------------------------- 1291 1292 When bonding is configured, it is important that the slave 1293devices not have routes that supersede routes of the master (or, 1294generally, not have routes at all). For example, suppose the bonding 1295device bond0 has two slaves, eth0 and eth1, and the routing table is 1296as follows: 1297 1298Kernel IP routing table 1299Destination Gateway Genmask Flags MSS Window irtt Iface 130010.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth0 130110.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 eth1 130210.0.0.0 0.0.0.0 255.255.0.0 U 40 0 0 bond0 1303127.0.0.0 0.0.0.0 255.0.0.0 U 40 0 0 lo 1304 1305 This routing configuration will likely still update the 1306receive/transmit times in the driver (needed by the ARP monitor), but 1307may bypass the bonding driver (because outgoing traffic to, in this 1308case, another host on network 10 would use eth0 or eth1 before bond0). 1309 1310 The ARP monitor (and ARP itself) may become confused by this 1311configuration, because ARP requests (generated by the ARP monitor) 1312will be sent on one interface (bond0), but the corresponding reply 1313will arrive on a different interface (eth0). This reply looks to ARP 1314as an unsolicited ARP reply (because ARP matches replies on an 1315interface basis), and is discarded. The MII monitor is not affected 1316by the state of the routing table. 1317 1318 The solution here is simply to insure that slaves do not have 1319routes of their own, and if for some reason they must, those routes do 1320not supersede routes of their master. This should generally be the 1321case, but unusual configurations or errant manual or automatic static 1322route additions may cause trouble. 1323 13248.2 Ethernet Device Renaming 1325---------------------------- 1326 1327 On systems with network configuration scripts that do not 1328associate physical devices directly with network interface names (so 1329that the same physical device always has the same "ethX" name), it may 1330be necessary to add some special logic to either /etc/modules.conf or 1331/etc/modprobe.conf (depending upon which is installed on the system). 1332 1333 For example, given a modules.conf containing the following: 1334 1335alias bond0 bonding 1336options bond0 mode=some-mode miimon=50 1337alias eth0 tg3 1338alias eth1 tg3 1339alias eth2 e1000 1340alias eth3 e1000 1341 1342 If neither eth0 and eth1 are slaves to bond0, then when the 1343bond0 interface comes up, the devices may end up reordered. This 1344happens because bonding is loaded first, then its slave device's 1345drivers are loaded next. Since no other drivers have been loaded, 1346when the e1000 driver loads, it will receive eth0 and eth1 for its 1347devices, but the bonding configuration tries to enslave eth2 and eth3 1348(which may later be assigned to the tg3 devices). 1349 1350 Adding the following: 1351 1352add above bonding e1000 tg3 1353 1354 causes modprobe to load e1000 then tg3, in that order, when 1355bonding is loaded. This command is fully documented in the 1356modules.conf manual page. 1357 1358 On systems utilizing modprobe.conf (or modprobe.conf.local), 1359an equivalent problem can occur. In this case, the following can be 1360added to modprobe.conf (or modprobe.conf.local, as appropriate), as 1361follows (all on one line; it has been split here for clarity): 1362 1363install bonding /sbin/modprobe tg3; /sbin/modprobe e1000; 1364 /sbin/modprobe --ignore-install bonding 1365 1366 This will, when loading the bonding module, rather than 1367performing the normal action, instead execute the provided command. 1368This command loads the device drivers in the order needed, then calls 1369modprobe with --ignore-install to cause the normal action to then take 1370place. Full documentation on this can be found in the modprobe.conf 1371and modprobe manual pages. 1372 13738.3. Painfully Slow Or No Failed Link Detection By Miimon 1374--------------------------------------------------------- 1375 1376 By default, bonding enables the use_carrier option, which 1377instructs bonding to trust the driver to maintain carrier state. 1378 1379 As discussed in the options section, above, some drivers do 1380not support the netif_carrier_on/_off link state tracking system. 1381With use_carrier enabled, bonding will always see these links as up, 1382regardless of their actual state. 1383 1384 Additionally, other drivers do support netif_carrier, but do 1385not maintain it in real time, e.g., only polling the link state at 1386some fixed interval. In this case, miimon will detect failures, but 1387only after some long period of time has expired. If it appears that 1388miimon is very slow in detecting link failures, try specifying 1389use_carrier=0 to see if that improves the failure detection time. If 1390it does, then it may be that the driver checks the carrier state at a 1391fixed interval, but does not cache the MII register values (so the 1392use_carrier=0 method of querying the registers directly works). If 1393use_carrier=0 does not improve the failover, then the driver may cache 1394the registers, or the problem may be elsewhere. 1395 1396 Also, remember that miimon only checks for the device's 1397carrier state. It has no way to determine the state of devices on or 1398beyond other ports of a switch, or if a switch is refusing to pass 1399traffic while still maintaining carrier on. 1400 14019. SNMP agents 1402=============== 1403 1404 If running SNMP agents, the bonding driver should be loaded 1405before any network drivers participating in a bond. This requirement 1406is due to the interface index (ipAdEntIfIndex) being associated to 1407the first interface found with a given IP address. That is, there is 1408only one ipAdEntIfIndex for each IP address. For example, if eth0 and 1409eth1 are slaves of bond0 and the driver for eth0 is loaded before the 1410bonding driver, the interface for the IP address will be associated 1411with the eth0 interface. This configuration is shown below, the IP 1412address 192.168.1.1 has an interface index of 2 which indexes to eth0 1413in the ifDescr table (ifDescr.2). 1414 1415 interfaces.ifTable.ifEntry.ifDescr.1 = lo 1416 interfaces.ifTable.ifEntry.ifDescr.2 = eth0 1417 interfaces.ifTable.ifEntry.ifDescr.3 = eth1 1418 interfaces.ifTable.ifEntry.ifDescr.4 = eth2 1419 interfaces.ifTable.ifEntry.ifDescr.5 = eth3 1420 interfaces.ifTable.ifEntry.ifDescr.6 = bond0 1421 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 5 1422 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2 1423 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 4 1424 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1 1425 1426 This problem is avoided by loading the bonding driver before 1427any network drivers participating in a bond. Below is an example of 1428loading the bonding driver first, the IP address 192.168.1.1 is 1429correctly associated with ifDescr.2. 1430 1431 interfaces.ifTable.ifEntry.ifDescr.1 = lo 1432 interfaces.ifTable.ifEntry.ifDescr.2 = bond0 1433 interfaces.ifTable.ifEntry.ifDescr.3 = eth0 1434 interfaces.ifTable.ifEntry.ifDescr.4 = eth1 1435 interfaces.ifTable.ifEntry.ifDescr.5 = eth2 1436 interfaces.ifTable.ifEntry.ifDescr.6 = eth3 1437 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.10.10.10 = 6 1438 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.192.168.1.1 = 2 1439 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.10.74.20.94 = 5 1440 ip.ipAddrTable.ipAddrEntry.ipAdEntIfIndex.127.0.0.1 = 1 1441 1442 While some distributions may not report the interface name in 1443ifDescr, the association between the IP address and IfIndex remains 1444and SNMP functions such as Interface_Scan_Next will report that 1445association. 1446 144710. Promiscuous mode 1448==================== 1449 1450 When running network monitoring tools, e.g., tcpdump, it is 1451common to enable promiscuous mode on the device, so that all traffic 1452is seen (instead of seeing only traffic destined for the local host). 1453The bonding driver handles promiscuous mode changes to the bonding 1454master device (e.g., bond0), and propagates the setting to the slave 1455devices. 1456 1457 For the balance-rr, balance-xor, broadcast, and 802.3ad modes, 1458the promiscuous mode setting is propagated to all slaves. 1459 1460 For the active-backup, balance-tlb and balance-alb modes, the 1461promiscuous mode setting is propagated only to the active slave. 1462 1463 For balance-tlb mode, the active slave is the slave currently 1464receiving inbound traffic. 1465 1466 For balance-alb mode, the active slave is the slave used as a 1467"primary." This slave is used for mode-specific control traffic, for 1468sending to peers that are unassigned or if the load is unbalanced. 1469 1470 For the active-backup, balance-tlb and balance-alb modes, when 1471the active slave changes (e.g., due to a link failure), the 1472promiscuous setting will be propagated to the new active slave. 1473 147411. Configuring Bonding for High Availability 1475============================================= 1476 1477 High Availability refers to configurations that provide 1478maximum network availability by having redundant or backup devices, 1479links or switches between the host and the rest of the world. The 1480goal is to provide the maximum availability of network connectivity 1481(i.e., the network always works), even though other configurations 1482could provide higher throughput. 1483 148411.1 High Availability in a Single Switch Topology 1485-------------------------------------------------- 1486 1487 If two hosts (or a host and a single switch) are directly 1488connected via multiple physical links, then there is no availability 1489penalty to optimizing for maximum bandwidth. In this case, there is 1490only one switch (or peer), so if it fails, there is no alternative 1491access to fail over to. Additionally, the bonding load balance modes 1492support link monitoring of their members, so if individual links fail, 1493the load will be rebalanced across the remaining devices. 1494 1495 See Section 13, "Configuring Bonding for Maximum Throughput" 1496for information on configuring bonding with one peer device. 1497 149811.2 High Availability in a Multiple Switch Topology 1499---------------------------------------------------- 1500 1501 With multiple switches, the configuration of bonding and the 1502network changes dramatically. In multiple switch topologies, there is 1503a trade off between network availability and usable bandwidth. 1504 1505 Below is a sample network, configured to maximize the 1506availability of the network: 1507 1508 | | 1509 |port3 port3| 1510 +-----+----+ +-----+----+ 1511 | |port2 ISL port2| | 1512 | switch A +--------------------------+ switch B | 1513 | | | | 1514 +-----+----+ +-----++---+ 1515 |port1 port1| 1516 | +-------+ | 1517 +-------------+ host1 +---------------+ 1518 eth0 +-------+ eth1 1519 1520 In this configuration, there is a link between the two 1521switches (ISL, or inter switch link), and multiple ports connecting to 1522the outside world ("port3" on each switch). There is no technical 1523reason that this could not be extended to a third switch. 1524 152511.2.1 HA Bonding Mode Selection for Multiple Switch Topology 1526------------------------------------------------------------- 1527 1528 In a topology such as the example above, the active-backup and 1529broadcast modes are the only useful bonding modes when optimizing for 1530availability; the other modes require all links to terminate on the 1531same peer for them to behave rationally. 1532 1533active-backup: This is generally the preferred mode, particularly if 1534 the switches have an ISL and play together well. If the 1535 network configuration is such that one switch is specifically 1536 a backup switch (e.g., has lower capacity, higher cost, etc), 1537 then the primary option can be used to insure that the 1538 preferred link is always used when it is available. 1539 1540broadcast: This mode is really a special purpose mode, and is suitable 1541 only for very specific needs. For example, if the two 1542 switches are not connected (no ISL), and the networks beyond 1543 them are totally independent. In this case, if it is 1544 necessary for some specific one-way traffic to reach both 1545 independent networks, then the broadcast mode may be suitable. 1546 154711.2.2 HA Link Monitoring Selection for Multiple Switch Topology 1548---------------------------------------------------------------- 1549 1550 The choice of link monitoring ultimately depends upon your 1551switch. If the switch can reliably fail ports in response to other 1552failures, then either the MII or ARP monitors should work. For 1553example, in the above example, if the "port3" link fails at the remote 1554end, the MII monitor has no direct means to detect this. The ARP 1555monitor could be configured with a target at the remote end of port3, 1556thus detecting that failure without switch support. 1557 1558 In general, however, in a multiple switch topology, the ARP 1559monitor can provide a higher level of reliability in detecting end to 1560end connectivity failures (which may be caused by the failure of any 1561individual component to pass traffic for any reason). Additionally, 1562the ARP monitor should be configured with multiple targets (at least 1563one for each switch in the network). This will insure that, 1564regardless of which switch is active, the ARP monitor has a suitable 1565target to query. 1566 1567 156812. Configuring Bonding for Maximum Throughput 1569============================================== 1570 157112.1 Maximizing Throughput in a Single Switch Topology 1572------------------------------------------------------ 1573 1574 In a single switch configuration, the best method to maximize 1575throughput depends upon the application and network environment. The 1576various load balancing modes each have strengths and weaknesses in 1577different environments, as detailed below. 1578 1579 For this discussion, we will break down the topologies into 1580two categories. Depending upon the destination of most traffic, we 1581categorize them into either "gatewayed" or "local" configurations. 1582 1583 In a gatewayed configuration, the "switch" is acting primarily 1584as a router, and the majority of traffic passes through this router to 1585other networks. An example would be the following: 1586 1587 1588 +----------+ +----------+ 1589 | |eth0 port1| | to other networks 1590 | Host A +---------------------+ router +-------------------> 1591 | +---------------------+ | Hosts B and C are out 1592 | |eth1 port2| | here somewhere 1593 +----------+ +----------+ 1594 1595 The router may be a dedicated router device, or another host 1596acting as a gateway. For our discussion, the important point is that 1597the majority of traffic from Host A will pass through the router to 1598some other network before reaching its final destination. 1599 1600 In a gatewayed network configuration, although Host A may 1601communicate with many other systems, all of its traffic will be sent 1602and received via one other peer on the local network, the router. 1603 1604 Note that the case of two systems connected directly via 1605multiple physical links is, for purposes of configuring bonding, the 1606same as a gatewayed configuration. In that case, it happens that all 1607traffic is destined for the "gateway" itself, not some other network 1608beyond the gateway. 1609 1610 In a local configuration, the "switch" is acting primarily as 1611a switch, and the majority of traffic passes through this switch to 1612reach other stations on the same network. An example would be the 1613following: 1614 1615 +----------+ +----------+ +--------+ 1616 | |eth0 port1| +-------+ Host B | 1617 | Host A +------------+ switch |port3 +--------+ 1618 | +------------+ | +--------+ 1619 | |eth1 port2| +------------------+ Host C | 1620 +----------+ +----------+port4 +--------+ 1621 1622 1623 Again, the switch may be a dedicated switch device, or another 1624host acting as a gateway. For our discussion, the important point is 1625that the majority of traffic from Host A is destined for other hosts 1626on the same local network (Hosts B and C in the above example). 1627 1628 In summary, in a gatewayed configuration, traffic to and from 1629the bonded device will be to the same MAC level peer on the network 1630(the gateway itself, i.e., the router), regardless of its final 1631destination. In a local configuration, traffic flows directly to and 1632from the final destinations, thus, each destination (Host B, Host C) 1633will be addressed directly by their individual MAC addresses. 1634 1635 This distinction between a gatewayed and a local network 1636configuration is important because many of the load balancing modes 1637available use the MAC addresses of the local network source and 1638destination to make load balancing decisions. The behavior of each 1639mode is described below. 1640 1641 164212.1.1 MT Bonding Mode Selection for Single Switch Topology 1643----------------------------------------------------------- 1644 1645 This configuration is the easiest to set up and to understand, 1646although you will have to decide which bonding mode best suits your 1647needs. The trade offs for each mode are detailed below: 1648 1649balance-rr: This mode is the only mode that will permit a single 1650 TCP/IP connection to stripe traffic across multiple 1651 interfaces. It is therefore the only mode that will allow a 1652 single TCP/IP stream to utilize more than one interface's 1653 worth of throughput. This comes at a cost, however: the 1654 striping often results in peer systems receiving packets out 1655 of order, causing TCP/IP's congestion control system to kick 1656 in, often by retransmitting segments. 1657 1658 It is possible to adjust TCP/IP's congestion limits by 1659 altering the net.ipv4.tcp_reordering sysctl parameter. The 1660 usual default value is 3, and the maximum useful value is 127. 1661 For a four interface balance-rr bond, expect that a single 1662 TCP/IP stream will utilize no more than approximately 2.3 1663 interface's worth of throughput, even after adjusting 1664 tcp_reordering. 1665 1666 Note that this out of order delivery occurs when both the 1667 sending and receiving systems are utilizing a multiple 1668 interface bond. Consider a configuration in which a 1669 balance-rr bond feeds into a single higher capacity network 1670 channel (e.g., multiple 100Mb/sec ethernets feeding a single 1671 gigabit ethernet via an etherchannel capable switch). In this 1672 configuration, traffic sent from the multiple 100Mb devices to 1673 a destination connected to the gigabit device will not see 1674 packets out of order. However, traffic sent from the gigabit 1675 device to the multiple 100Mb devices may or may not see 1676 traffic out of order, depending upon the balance policy of the 1677 switch. Many switches do not support any modes that stripe 1678 traffic (instead choosing a port based upon IP or MAC level 1679 addresses); for those devices, traffic flowing from the 1680 gigabit device to the many 100Mb devices will only utilize one 1681 interface. 1682 1683 If you are utilizing protocols other than TCP/IP, UDP for 1684 example, and your application can tolerate out of order 1685 delivery, then this mode can allow for single stream datagram 1686 performance that scales near linearly as interfaces are added 1687 to the bond. 1688 1689 This mode requires the switch to have the appropriate ports 1690 configured for "etherchannel" or "trunking." 1691 1692active-backup: There is not much advantage in this network topology to 1693 the active-backup mode, as the inactive backup devices are all 1694 connected to the same peer as the primary. In this case, a 1695 load balancing mode (with link monitoring) will provide the 1696 same level of network availability, but with increased 1697 available bandwidth. On the plus side, active-backup mode 1698 does not require any configuration of the switch, so it may 1699 have value if the hardware available does not support any of 1700 the load balance modes. 1701 1702balance-xor: This mode will limit traffic such that packets destined 1703 for specific peers will always be sent over the same 1704 interface. Since the destination is determined by the MAC 1705 addresses involved, this mode works best in a "local" network 1706 configuration (as described above), with destinations all on 1707 the same local network. This mode is likely to be suboptimal 1708 if all your traffic is passed through a single router (i.e., a 1709 "gatewayed" network configuration, as described above). 1710 1711 As with balance-rr, the switch ports need to be configured for 1712 "etherchannel" or "trunking." 1713 1714broadcast: Like active-backup, there is not much advantage to this 1715 mode in this type of network topology. 1716 1717802.3ad: This mode can be a good choice for this type of network 1718 topology. The 802.3ad mode is an IEEE standard, so all peers 1719 that implement 802.3ad should interoperate well. The 802.3ad 1720 protocol includes automatic configuration of the aggregates, 1721 so minimal manual configuration of the switch is needed 1722 (typically only to designate that some set of devices is 1723 available for 802.3ad). The 802.3ad standard also mandates 1724 that frames be delivered in order (within certain limits), so 1725 in general single connections will not see misordering of 1726 packets. The 802.3ad mode does have some drawbacks: the 1727 standard mandates that all devices in the aggregate operate at 1728 the same speed and duplex. Also, as with all bonding load 1729 balance modes other than balance-rr, no single connection will 1730 be able to utilize more than a single interface's worth of 1731 bandwidth. 1732 1733 Additionally, the linux bonding 802.3ad implementation 1734 distributes traffic by peer (using an XOR of MAC addresses), 1735 so in a "gatewayed" configuration, all outgoing traffic will 1736 generally use the same device. Incoming traffic may also end 1737 up on a single device, but that is dependent upon the 1738 balancing policy of the peer's 8023.ad implementation. In a 1739 "local" configuration, traffic will be distributed across the 1740 devices in the bond. 1741 1742 Finally, the 802.3ad mode mandates the use of the MII monitor, 1743 therefore, the ARP monitor is not available in this mode. 1744 1745balance-tlb: The balance-tlb mode balances outgoing traffic by peer. 1746 Since the balancing is done according to MAC address, in a 1747 "gatewayed" configuration (as described above), this mode will 1748 send all traffic across a single device. However, in a 1749 "local" network configuration, this mode balances multiple 1750 local network peers across devices in a vaguely intelligent 1751 manner (not a simple XOR as in balance-xor or 802.3ad mode), 1752 so that mathematically unlucky MAC addresses (i.e., ones that 1753 XOR to the same value) will not all "bunch up" on a single 1754 interface. 1755 1756 Unlike 802.3ad, interfaces may be of differing speeds, and no 1757 special switch configuration is required. On the down side, 1758 in this mode all incoming traffic arrives over a single 1759 interface, this mode requires certain ethtool support in the 1760 network device driver of the slave interfaces, and the ARP 1761 monitor is not available. 1762 1763balance-alb: This mode is everything that balance-tlb is, and more. 1764 It has all of the features (and restrictions) of balance-tlb, 1765 and will also balance incoming traffic from local network 1766 peers (as described in the Bonding Module Options section, 1767 above). 1768 1769 The only additional down side to this mode is that the network 1770 device driver must support changing the hardware address while 1771 the device is open. 1772 177312.1.2 MT Link Monitoring for Single Switch Topology 1774---------------------------------------------------- 1775 1776 The choice of link monitoring may largely depend upon which 1777mode you choose to use. The more advanced load balancing modes do not 1778support the use of the ARP monitor, and are thus restricted to using 1779the MII monitor (which does not provide as high a level of end to end 1780assurance as the ARP monitor). 1781 178212.2 Maximum Throughput in a Multiple Switch Topology 1783----------------------------------------------------- 1784 1785 Multiple switches may be utilized to optimize for throughput 1786when they are configured in parallel as part of an isolated network 1787between two or more systems, for example: 1788 1789 +-----------+ 1790 | Host A | 1791 +-+---+---+-+ 1792 | | | 1793 +--------+ | +---------+ 1794 | | | 1795 +------+---+ +-----+----+ +-----+----+ 1796 | Switch A | | Switch B | | Switch C | 1797 +------+---+ +-----+----+ +-----+----+ 1798 | | | 1799 +--------+ | +---------+ 1800 | | | 1801 +-+---+---+-+ 1802 | Host B | 1803 +-----------+ 1804 1805 In this configuration, the switches are isolated from one 1806another. One reason to employ a topology such as this is for an 1807isolated network with many hosts (a cluster configured for high 1808performance, for example), using multiple smaller switches can be more 1809cost effective than a single larger switch, e.g., on a network with 24 1810hosts, three 24 port switches can be significantly less expensive than 1811a single 72 port switch. 1812 1813 If access beyond the network is required, an individual host 1814can be equipped with an additional network device connected to an 1815external network; this host then additionally acts as a gateway. 1816 181712.2.1 MT Bonding Mode Selection for Multiple Switch Topology 1818------------------------------------------------------------- 1819 1820 In actual practice, the bonding mode typically employed in 1821configurations of this type is balance-rr. Historically, in this 1822network configuration, the usual caveats about out of order packet 1823delivery are mitigated by the use of network adapters that do not do 1824any kind of packet coalescing (via the use of NAPI, or because the 1825device itself does not generate interrupts until some number of 1826packets has arrived). When employed in this fashion, the balance-rr 1827mode allows individual connections between two hosts to effectively 1828utilize greater than one interface's bandwidth. 1829 183012.2.2 MT Link Monitoring for Multiple Switch Topology 1831------------------------------------------------------ 1832 1833 Again, in actual practice, the MII monitor is most often used 1834in this configuration, as performance is given preference over 1835availability. The ARP monitor will function in this topology, but its 1836advantages over the MII monitor are mitigated by the volume of probes 1837needed as the number of systems involved grows (remember that each 1838host in the network is configured with bonding). 1839 184013. Switch Behavior Issues 1841========================== 1842 184313.1 Link Establishment and Failover Delays 1844------------------------------------------- 1845 1846 Some switches exhibit undesirable behavior with regard to the 1847timing of link up and down reporting by the switch. 1848 1849 First, when a link comes up, some switches may indicate that 1850the link is up (carrier available), but not pass traffic over the 1851interface for some period of time. This delay is typically due to 1852some type of autonegotiation or routing protocol, but may also occur 1853during switch initialization (e.g., during recovery after a switch 1854failure). If you find this to be a problem, specify an appropriate 1855value to the updelay bonding module option to delay the use of the 1856relevant interface(s). 1857 1858 Second, some switches may "bounce" the link state one or more 1859times while a link is changing state. This occurs most commonly while 1860the switch is initializing. Again, an appropriate updelay value may 1861help. 1862 1863 Note that when a bonding interface has no active links, the 1864driver will immediately reuse the first link that goes up, even if the 1865updelay parameter has been specified (the updelay is ignored in this 1866case). If there are slave interfaces waiting for the updelay timeout 1867to expire, the interface that first went into that state will be 1868immediately reused. This reduces down time of the network if the 1869value of updelay has been overestimated, and since this occurs only in 1870cases with no connectivity, there is no additional penalty for 1871ignoring the updelay. 1872 1873 In addition to the concerns about switch timings, if your 1874switches take a long time to go into backup mode, it may be desirable 1875to not activate a backup interface immediately after a link goes down. 1876Failover may be delayed via the downdelay bonding module option. 1877 187813.2 Duplicated Incoming Packets 1879-------------------------------- 1880 1881 It is not uncommon to observe a short burst of duplicated 1882traffic when the bonding device is first used, or after it has been 1883idle for some period of time. This is most easily observed by issuing 1884a "ping" to some other host on the network, and noticing that the 1885output from ping flags duplicates (typically one per slave). 1886 1887 For example, on a bond in active-backup mode with five slaves 1888all connected to one switch, the output may appear as follows: 1889 1890# ping -n 10.0.4.2 1891PING 10.0.4.2 (10.0.4.2) from 10.0.3.10 : 56(84) bytes of data. 189264 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.7 ms 189364 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) 189464 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) 189564 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) 189664 bytes from 10.0.4.2: icmp_seq=1 ttl=64 time=13.8 ms (DUP!) 189764 bytes from 10.0.4.2: icmp_seq=2 ttl=64 time=0.216 ms 189864 bytes from 10.0.4.2: icmp_seq=3 ttl=64 time=0.267 ms 189964 bytes from 10.0.4.2: icmp_seq=4 ttl=64 time=0.222 ms 1900 1901 This is not due to an error in the bonding driver, rather, it 1902is a side effect of how many switches update their MAC forwarding 1903tables. Initially, the switch does not associate the MAC address in 1904the packet with a particular switch port, and so it may send the 1905traffic to all ports until its MAC forwarding table is updated. Since 1906the interfaces attached to the bond may occupy multiple ports on a 1907single switch, when the switch (temporarily) floods the traffic to all 1908ports, the bond device receives multiple copies of the same packet 1909(one per slave device). 1910 1911 The duplicated packet behavior is switch dependent, some 1912switches exhibit this, and some do not. On switches that display this 1913behavior, it can be induced by clearing the MAC forwarding table (on 1914most Cisco switches, the privileged command "clear mac address-table 1915dynamic" will accomplish this). 1916 191714. Hardware Specific Considerations 1918==================================== 1919 1920 This section contains additional information for configuring 1921bonding on specific hardware platforms, or for interfacing bonding 1922with particular switches or other devices. 1923 192414.1 IBM BladeCenter 1925-------------------- 1926 1927 This applies to the JS20 and similar systems. 1928 1929 On the JS20 blades, the bonding driver supports only 1930balance-rr, active-backup, balance-tlb and balance-alb modes. This is 1931largely due to the network topology inside the BladeCenter, detailed 1932below. 1933 1934JS20 network adapter information 1935-------------------------------- 1936 1937 All JS20s come with two Broadcom Gigabit Ethernet ports 1938integrated on the planar (that's "motherboard" in IBM-speak). In the 1939BladeCenter chassis, the eth0 port of all JS20 blades is hard wired to 1940I/O Module #1; similarly, all eth1 ports are wired to I/O Module #2. 1941An add-on Broadcom daughter card can be installed on a JS20 to provide 1942two more Gigabit Ethernet ports. These ports, eth2 and eth3, are 1943wired to I/O Modules 3 and 4, respectively. 1944 1945 Each I/O Module may contain either a switch or a passthrough 1946module (which allows ports to be directly connected to an external 1947switch). Some bonding modes require a specific BladeCenter internal 1948network topology in order to function; these are detailed below. 1949 1950 Additional BladeCenter-specific networking information can be 1951found in two IBM Redbooks (www.ibm.com/redbooks): 1952 1953"IBM eServer BladeCenter Networking Options" 1954"IBM eServer BladeCenter Layer 2-7 Network Switching" 1955 1956BladeCenter networking configuration 1957------------------------------------ 1958 1959 Because a BladeCenter can be configured in a very large number 1960of ways, this discussion will be confined to describing basic 1961configurations. 1962 1963 Normally, Ethernet Switch Modules (ESMs) are used in I/O 1964modules 1 and 2. In this configuration, the eth0 and eth1 ports of a 1965JS20 will be connected to different internal switches (in the 1966respective I/O modules). 1967 1968 A passthrough module (OPM or CPM, optical or copper, 1969passthrough module) connects the I/O module directly to an external 1970switch. By using PMs in I/O module #1 and #2, the eth0 and eth1 1971interfaces of a JS20 can be redirected to the outside world and 1972connected to a common external switch. 1973 1974 Depending upon the mix of ESMs and PMs, the network will 1975appear to bonding as either a single switch topology (all PMs) or as a 1976multiple switch topology (one or more ESMs, zero or more PMs). It is 1977also possible to connect ESMs together, resulting in a configuration 1978much like the example in "High Availability in a Multiple Switch 1979Topology," above. 1980 1981Requirements for specific modes 1982------------------------------- 1983 1984 The balance-rr mode requires the use of passthrough modules 1985for devices in the bond, all connected to an common external switch. 1986That switch must be configured for "etherchannel" or "trunking" on the 1987appropriate ports, as is usual for balance-rr. 1988 1989 The balance-alb and balance-tlb modes will function with 1990either switch modules or passthrough modules (or a mix). The only 1991specific requirement for these modes is that all network interfaces 1992must be able to reach all destinations for traffic sent over the 1993bonding device (i.e., the network must converge at some point outside 1994the BladeCenter). 1995 1996 The active-backup mode has no additional requirements. 1997 1998Link monitoring issues 1999---------------------- 2000 2001 When an Ethernet Switch Module is in place, only the ARP 2002monitor will reliably detect link loss to an external switch. This is 2003nothing unusual, but examination of the BladeCenter cabinet would 2004suggest that the "external" network ports are the ethernet ports for 2005the system, when it fact there is a switch between these "external" 2006ports and the devices on the JS20 system itself. The MII monitor is 2007only able to detect link failures between the ESM and the JS20 system. 2008 2009 When a passthrough module is in place, the MII monitor does 2010detect failures to the "external" port, which is then directly 2011connected to the JS20 system. 2012 2013Other concerns 2014-------------- 2015 2016 The Serial Over LAN (SoL) link is established over the primary 2017ethernet (eth0) only, therefore, any loss of link to eth0 will result 2018in losing your SoL connection. It will not fail over with other 2019network traffic, as the SoL system is beyond the control of the 2020bonding driver. 2021 2022 It may be desirable to disable spanning tree on the switch 2023(either the internal Ethernet Switch Module, or an external switch) to 2024avoid fail-over delay issues when using bonding. 2025 2026 202715. Frequently Asked Questions 2028============================== 2029 20301. Is it SMP safe? 2031 2032 Yes. The old 2.0.xx channel bonding patch was not SMP safe. 2033The new driver was designed to be SMP safe from the start. 2034 20352. What type of cards will work with it? 2036 2037 Any Ethernet type cards (you can even mix cards - a Intel 2038EtherExpress PRO/100 and a 3com 3c905b, for example). For most modes, 2039devices need not be of the same speed. 2040 20413. How many bonding devices can I have? 2042 2043 There is no limit. 2044 20454. How many slaves can a bonding device have? 2046 2047 This is limited only by the number of network interfaces Linux 2048supports and/or the number of network cards you can place in your 2049system. 2050 20515. What happens when a slave link dies? 2052 2053 If link monitoring is enabled, then the failing device will be 2054disabled. The active-backup mode will fail over to a backup link, and 2055other modes will ignore the failed link. The link will continue to be 2056monitored, and should it recover, it will rejoin the bond (in whatever 2057manner is appropriate for the mode). See the sections on High 2058Availability and the documentation for each mode for additional 2059information. 2060 2061 Link monitoring can be enabled via either the miimon or 2062arp_interval parameters (described in the module parameters section, 2063above). In general, miimon monitors the carrier state as sensed by 2064the underlying network device, and the arp monitor (arp_interval) 2065monitors connectivity to another host on the local network. 2066 2067 If no link monitoring is configured, the bonding driver will 2068be unable to detect link failures, and will assume that all links are 2069always available. This will likely result in lost packets, and a 2070resulting degradation of performance. The precise performance loss 2071depends upon the bonding mode and network configuration. 2072 20736. Can bonding be used for High Availability? 2074 2075 Yes. See the section on High Availability for details. 2076 20777. Which switches/systems does it work with? 2078 2079 The full answer to this depends upon the desired mode. 2080 2081 In the basic balance modes (balance-rr and balance-xor), it 2082works with any system that supports etherchannel (also called 2083trunking). Most managed switches currently available have such 2084support, and many unmanaged switches as well. 2085 2086 The advanced balance modes (balance-tlb and balance-alb) do 2087not have special switch requirements, but do need device drivers that 2088support specific features (described in the appropriate section under 2089module parameters, above). 2090 2091 In 802.3ad mode, it works with systems that support IEEE 2092802.3ad Dynamic Link Aggregation. Most managed and many unmanaged 2093switches currently available support 802.3ad. 2094 2095 The active-backup mode should work with any Layer-II switch. 2096 20978. Where does a bonding device get its MAC address from? 2098 2099 If not explicitly configured (with ifconfig or ip link), the 2100MAC address of the bonding device is taken from its first slave 2101device. This MAC address is then passed to all following slaves and 2102remains persistent (even if the first slave is removed) until the 2103bonding device is brought down or reconfigured. 2104 2105 If you wish to change the MAC address, you can set it with 2106ifconfig or ip link: 2107 2108# ifconfig bond0 hw ether 00:11:22:33:44:55 2109 2110# ip link set bond0 address 66:77:88:99:aa:bb 2111 2112 The MAC address can be also changed by bringing down/up the 2113device and then changing its slaves (or their order): 2114 2115# ifconfig bond0 down ; modprobe -r bonding 2116# ifconfig bond0 .... up 2117# ifenslave bond0 eth... 2118 2119 This method will automatically take the address from the next 2120slave that is added. 2121 2122 To restore your slaves' MAC addresses, you need to detach them 2123from the bond (`ifenslave -d bond0 eth0'). The bonding driver will 2124then restore the MAC addresses that the slaves had before they were 2125enslaved. 2126 212716. Resources and Links 2128======================= 2129 2130The latest version of the bonding driver can be found in the latest 2131version of the linux kernel, found on http://kernel.org 2132 2133The latest version of this document can be found in either the latest 2134kernel source (named Documentation/networking/bonding.txt), or on the 2135bonding sourceforge site: 2136 2137http://www.sourceforge.net/projects/bonding 2138 2139Discussions regarding the bonding driver take place primarily on the 2140bonding-devel mailing list, hosted at sourceforge.net. If you have 2141questions or problems, post them to the list. The list address is: 2142 2143bonding-devel@lists.sourceforge.net 2144 2145 The administrative interface (to subscribe or unsubscribe) can 2146be found at: 2147 2148https://lists.sourceforge.net/lists/listinfo/bonding-devel 2149 2150Donald Becker's Ethernet Drivers and diag programs may be found at : 2151 - http://www.scyld.com/network/ 2152 2153You will also find a lot of information regarding Ethernet, NWay, MII, 2154etc. at www.scyld.com. 2155 2156-- END -- 2157