Routing

Hi, 

I'm struggling to get my head around this and hoping someone can help. 

I'm building out a POC using vMXs implementing an EVPN-MPLS based leaf/spine architecture. I've reached the point where most things work but there is one thing I can't understand. 

I have a 2 leaf' connected to 2 hosts via multihoming with a trunk interface carrying vlan 7 and 10. on the leaf' I use virtual switches to ingest these frames and transport them up to PE devices. The PE devices contain the default gateways. The hosts can reach the default gateways. 

The odd behaviour is that the hosts have an interface in vlans 7 & 10. If I ping host1's vlan 7 address from host2's vlan 10 address the traffic only goes to the leaf and then gets transported back down to the host. As it's crossing broadcast domains I would have expected it to travel up to the PE, be routed and then come back down to the leaf.

I've attached a screengrab of the topology. I've also attached configs for leaf1 and PE1, the 2s are just mirrors. The spines are just participating in the SR-ISIS domain.

Many thanks

Hi,

Looks like on PE1 your rib-group to import routes from VPN10.inet.0 to VPN7.inet.0 seems to be wrong, you're using same rib-group under both routing instances - 

routing-options {
    rib-groups {
        7->10 {
            import-rib [ VPN7.inet.0 VPN10.inet.0 ];
            import-policy ACCEPT_ALL;
        }
    }
}

So, your second rib-group, to be applied under VPN10 routing-instance should be something like

routing-options {
    rib-groups {
        10->7 {
            import-rib [ VPN10.inet.0 VPN7.inet.0 ];
            import-policy ACCEPT_ALL;
        }
    }
}

Hi, 

I'm struggling to get my head around this and hoping someone can help. 

I'm building out a POC using vMXs implementing an EVPN-MPLS based leaf/spine architecture. I've reached the point where most things work but there is one thing I can't understand. 

I have a 2 leaf' connected to 2 hosts via multihoming with a trunk interface carrying vlan 7 and 10. on the leaf' I use virtual switches to ingest these frames and transport them up to PE devices. The PE devices contain the default gateways. The hosts can reach the default gateways. 

The odd behaviour is that the hosts have an interface in vlans 7 & 10. If I ping host1's vlan 7 address from host2's vlan 10 address the traffic only goes to the leaf and then gets transported back down to the host. As it's crossing broadcast domains I would have expected it to travel up to the PE, be routed and then come back down to the leaf.

I've attached a screengrab of the topology. I've also attached configs for leaf1 and PE1, the 2s are just mirrors. The spines are just participating in the SR-ISIS domain.

Many thanks

Hi,

Looks like on PE1 your rib-group to import routes from VPN10.inet.0 to VPN7.inet.0 seems to be wrong, you're using same rib-group under both routing instances - 

routing-options {    rib-groups {        7->10 {            import-rib [ VPN7.inet.0 VPN10.inet.0 ];            import-policy ACCEPT_ALL;        }    }}

So, your second rib-group, to be applied under VPN10 routing-instance should be something like

routing-options {    rib-groups {        10->7 {            import-rib [ VPN10.inet.0 VPN7.inet.0 ];            import-policy ACCEPT_ALL;        }    }}

Hi, 

I'm struggling to get my head around this and hoping someone can help. 

I'm building out a POC using vMXs implementing an EVPN-MPLS based leaf/spine architecture. I've reached the point where most things work but there is one thing I can't understand. 

I have a 2 leaf' connected to 2 hosts via multihoming with a trunk interface carrying vlan 7 and 10. on the leaf' I use virtual switches to ingest these frames and transport them up to PE devices. The PE devices contain the default gateways. The hosts can reach the default gateways. 

The odd behaviour is that the hosts have an interface in vlans 7 & 10. If I ping host1's vlan 7 address from host2's vlan 10 address the traffic only goes to the leaf and then gets transported back down to the host. As it's crossing broadcast domains I would have expected it to travel up to the PE, be routed and then come back down to the leaf.

I've attached a screengrab of the topology. I've also attached configs for leaf1 and PE1, the 2s are just mirrors. The spines are just participating in the SR-ISIS domain.

Many thanks

Well, that RIB group takes routes from VPN7 RIB and imports it into VPN10 RIB.

Therefore, to import routes from VPN10 RIB into VPN7 RIB, you should do the reverse.

Hi,

Looks like on PE1 your rib-group to import routes from VPN10.inet.0 to VPN7.inet.0 seems to be wrong, you're using same rib-group under both routing instances - 

routing-options {    rib-groups {        7->10 {            import-rib [ VPN7.inet.0 VPN10.inet.0 ];            import-policy ACCEPT_ALL;        }    }}

So, your second rib-group, to be applied under VPN10 routing-instance should be something like

routing-options {    rib-groups {        10->7 {            import-rib [ VPN10.inet.0 VPN7.inet.0 ];            import-policy ACCEPT_ALL;        }    }}

Hi, 

I'm struggling to get my head around this and hoping someone can help. 

I'm building out a POC using vMXs implementing an EVPN-MPLS based leaf/spine architecture. I've reached the point where most things work but there is one thing I can't understand. 

I have a 2 leaf' connected to 2 hosts via multihoming with a trunk interface carrying vlan 7 and 10. on the leaf' I use virtual switches to ingest these frames and transport them up to PE devices. The PE devices contain the default gateways. The hosts can reach the default gateways. 

The odd behaviour is that the hosts have an interface in vlans 7 & 10. If I ping host1's vlan 7 address from host2's vlan 10 address the traffic only goes to the leaf and then gets transported back down to the host. As it's crossing broadcast domains I would have expected it to travel up to the PE, be routed and then come back down to the leaf.

I've attached a screengrab of the topology. I've also attached configs for leaf1 and PE1, the 2s are just mirrors. The spines are just participating in the SR-ISIS domain.

Many thanks