[j-nsp] ARP on unnumbered interfaces
Mihai
mihaigabriel at gmail.com
Sat Jan 10 10:34:55 EST 2015
Hello,
After the migration of a large network from a Cisco 7600 to a MX104 a
lot of users started to have random problems with their connection.
The setup is based on unnumbered interfaces and /32 static routes
through IFLs.
Basically, all clients with Cisco routers will have at some point a
missing ARP entry for their default gateway because the MX is changing
the ARP source address from the gw_addr to the primary address.On Cisco
i see the well known 'wrong cable' error.
Does anyone have a clue why is this happening beside a bug? I've made
some tests on MX960,MX480 and MX5 and didn't see this behavior.
This is a lab simulation:
mx# show
interfaces {
ge-1/1/8 {
unit 55 {
vlan-id 55;
proxy-arp unrestricted;
family inet {
unnumbered-address lo0.55;
}
}
unit 56 {
vlan-id 56;
proxy-arp unrestricted;
family inet {
unnumbered-address lo0.55;
}
}
}
lo0 {
unit 55 {
family inet {
address 5.5.5.5/32 {
primary;
}
address 10.10.10.1/24;
address 20.20.20.1/24;
}
}
}
}
routing-options {
static {
route 20.20.20.2/32 {
qualified-next-hop ge-1/1/8.55;
}
route 10.10.10.2/32 {
qualified-next-hop ge-1/1/8.56;
}
}
router-id 5.5.5.5;
}
mx> monitor traffic interface ge-1/1/8.55 detail no-resolve matching arp
Address resolution is OFF.
Listening on ge-1/1/8.55, capture size 1514 bytes
17:28:11.105586 Out arp who-has 20.20.20.2 tell 20.20.20.1
17:28:11.106100 In arp reply 20.20.20.2 is-at 00:1e:4a:fc:44:84
17:29:20.504891 Out arp who-has 20.20.20.2 tell 20.20.20.1
17:29:20.505375 In arp reply 20.20.20.2 is-at 00:1e:4a:fc:44:84
17:30:30.104188 Out arp who-has 20.20.20.2 tell 20.20.20.1
17:30:30.104632 In arp reply 20.20.20.2 is-at 00:1e:4a:fc:44:84
.....
17:53:01.790690 Out arp who-has 20.20.20.2 tell 5.5.5.5
17:54:05.690056 Out arp who-has 20.20.20.2 tell 5.5.5.5
Thanks!
More information about the juniper-nsp
mailing list