Hi there...!
I'm a long-time reader, but this is my first post.. :)
I set up our old server (an Acer Altos 11000) with Release 6.0
two days ago (cvsup'ed and installed new world and custom kernel
(see below)) and I keep getting a huge amounts of these message:
-----
Feb 17 12:39:46 scorpio kernel: arp: 192.168.1.103 is on fxp0
but got reply from 00:10:dc:7b:91:12 on re0
Feb 17 12:40:40 scorpio kernel: arp: 192.168.1.254 is on fxp0
but got reply from 00:a0:c5:44:a0:30 on re0
Feb 17 12:41:50 scorpio kernel: arp: 192.168.1.103 is on fxp0
but got reply from 00:10:dc:7b:91:12 on re0
Feb 17 12:42:16 scorpio kernel: arp: 192.168.1.103 is on fxp0
but got reply from 00:10:dc:7b:91:12 on re0
Feb 17 12:42:16 scorpio kernel: arp: 192.168.1.254 is on fxp0
but got reply from 00:a0:c5:44:a0:30 on re0
Feb 17 12:44:34 scorpio kernel: arp: 192.168.1.103 is on fxp0
but got reply from 00:10:dc:7b:91:12 on re0
Feb 17 12:46:45 scorpio kernel: arp: 192.168.1.103 is on fxp0
but got reply from 00:10:dc:7b:91:12 on re0
-----
There's an fxp onboard and an added realtek Gigabit card.
They are configured for two different subnets but are connected
to the same physical network.
The re0 is on the 192.168.100.0/24 network and the fxp0 on the
192.168.1.0/24 network.. the .254 is our gateway and the .103 is
me..
this is the ifconfig output of the cards:
-----
re0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
options=18<VLAN_MTU,VLAN_HWTAGGING>
inet6 fe80::240:f4ff:fed1:cee7%re0 prefixlen 64 scopeid
0x1
inet 192.168.100.15 netmask 0xffffff00 broadcast
192.168.100.255
ether 00:40:f4:d1:ce:e7
media: Ethernet autoselect (1000baseTX <full-duplex>)
status: active
fxp0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu
1500
options=8<VLAN_MTU>
inet6 fe80::200:e2ff:fe22:f419%fxp0 prefixlen 64 scopeid
0x2
inet 192.168.1.15 netmask 0xffffff00 broadcast
192.168.1.255
ether 00:00:e2:22:f4:19
media: Ethernet autoselect (100baseTX <full-duplex>)
status: active
-----
I compiled a custom kernel for SMP support (and commented out a
lot devices we don't have (other SCSI and RAID controllers))
It doesn't seem to affect the function of the server, but it's
mighty irritating and blows up the logs a lot... plus, I don't
think it's supposed to show this behaviour.. :)
I've going through the archives & web but the threads I found
didn't fit my case.. :(
Any ideas about that? Your help is very appreciated..
thank you.. :)
- Thomas
On 17 Feb 2006 at 12:26, Dominic Marks wrote:> Tried these sysctls? > > net.link.ether.inet.log_arp_wrong_iface > net.link.ether.inet.log_arp_movements > net.link.ether.inet.log_arp_permanent_modifyI set this in sysctl.conf now and did a reboot.. net.link.ether.inet.log_arp_wrong_iface = 0 and it works like expected.. no more messages of that kind... Thank you... :) Say.. this this flag is probably aware of run-time changes, too, hmm? so a "sysctl net.link.ether.inet.log_arp_wrong_iface=0" would have turned it off without reboot as well, right? How can an arp reply be received by the wrong interface, though? Isn't the request broadcast and the reply MAC addressed? I'm amazed - problem solved in less than 30 mins.. - Thomas
Thomas Franck wrote: [ ... ]> It doesn't seem to affect the function of the server, but it's > mighty irritating and blows up the logs a lot... plus, I don't > think it's supposed to show this behaviour.. :) > > I've going through the archives & web but the threads I found > didn't fit my case.. :(The first two hits from Google were informative, but this is what you want: sysctl net.link.ether.inet.log_arp_wrong_iface=0 Your network would be better configured if separate subnets were actually in separate collision domains, by using VLANs or an individual hub/switch for each subnet. The error message is useful to those people for whom it would be a genuine sign of problems... -- -Chuck
On 17 Feb 2006 at 8:07, Chuck Swiger wrote:> Thomas Franck wrote: > [ ... ] > > It doesn't seem to affect the function of the server, but it's > > mighty irritating and blows up the logs a lot... plus, I don't > > think it's supposed to show this behaviour.. :) > > > > I've going through the archives & web but the threads I found > > didn't fit my case.. :( > > The first two hits from Google were informative, but this is what > you want:the ones I found where about routing or two NICs on the same subnet.. anyway, none seem to relate to the behaviour I had.. or rather, give the sysctl mib in question..> sysctl net.link.ether.inet.log_arp_wrong_iface=0Yes.. that worked well..> Your network would be better configured if separate subnets were > actually in separate collision domains, by using VLANs or an > individual hub/switch for each subnet. The error message is > useful to those people for whom it would be a genuine sign of > problems...they will be seperated again once our firewall is back & running and the standard router is taken off the net.. still, as I wrote in the reply to Dominic Marks: Isn't the request broadcast and the reply MAC addressed? if I'm right on that, the "problem" (it's more cosmetic, really) shouldn't have happened, right..? (at the moment the net looks like that: {FreeBSD} | | [sw1] [sw2] | | [switch3] | | | | | {me} | [router] the switch3 will be VLAN'ed again and properly connected once the firewall is back..) Thanks for the help.. - Thomas