help with ethernet load-balancing (em)

David Smithson david at
Thu Apr 24 12:16:44 PDT 2003


I have a mission-critical file-server with two on-board Intel Gig-copper
NICs, em0 and em1, configured as follows:

        inet netmask 0xffffff00 broadcast
        inet6 fe80::230:48ff:fe27:7868%em0 prefixlen 64 scopeid 0x1
        ether 00:30:48:27:78:68
        media: Ethernet autoselect (1000baseTX <full-duplex>)
        status: active
        inet netmask 0xffffffff broadcast
        inet6 fe80::230:48ff:fe27:7869%em1 prefixlen 64 scopeid 0x2
        ether 00:30:48:27:78:69
        media: Ethernet autoselect (1000baseTX <full-duplex>)
        status: active

I would like network load-balancing to happen on this server and would
appreciate your help in this matter.  So far, most network clients
attach to em0.  arp -a returns:

? ( at 00:07:e9:0c:52:6d on em0 [ethernet]
? ( at 00:07:e9:0a:22:6a on em0 [ethernet]
? ( at 00:04:76:f5:88:9f on em0 [ethernet]
? ( at 00:04:76:e6:15:62 on em0 [ethernet]
? ( at 00:04:5a:59:65:1f on em0 [ethernet]
? ( at 00:04:5a:59:54:e7 on em0 [ethernet]
? ( at 00:04:5a:59:7a:dd on em0 [ethernet]
? ( at 00:04:5a:59:64:f9 on em0 [ethernet]
? ( at 00:04:5a:59:54:97 on em0 [ethernet]
? ( at 00:d0:b7:9e:ad:d0 on em0 [ethernet]
? ( at 00:01:02:83:de:f0 on em0 [ethernet]
? ( at 00:40:f4:21:c2:d2 on em0 [ethernet]
? ( at 00:30:48:27:78:68 on em0 permanent [ethernet]
? ( at 00:30:48:27:78:69 on em1 permanent [ethernet]
? ( at 00:e0:81:02:79:da on em0 [ethernet]
? ( at 00:e0:81:02:89:90 on em0 [ethernet]
? ( at 00:e0:81:02:87:b3 on em0 [ethernet]
? ( at 00:e0:81:02:64:6f on em0 [ethernet]
? ( at 00:e0:81:02:16:df on em0 [ethernet]
? ( at 00:e0:81:02:16:e1 on em0 [ethernet]
? ( at 00:d0:b7:b7:3b:3d on em0 [ethernet]
? ( at ff:ff:ff:ff:ff:ff on em0 permanent [ethernet]

Sorry about the wrap, but as you can see, all clients are attached to
em0.  If I do arp -a at the NT shell prompt on the Win2k Pro clients, I
get:	00:30:48:27:78:68	dynamic

My internal DNS server has round-robin turned on and has entries for
both and  However, clients consistently
resolve name-to-IP

What am I missing?  Can someone point me to some useful information
regarding load-balancing?  What is the mechanism here?

Thanks for your time.

David Smithson

More information about the freebsd-questions mailing list