ASR9901 BNG running 7.1.3 routing-rib error message affecting services

[RogerLienert15023]

Hi,

ASR9901 running 7.1.3 software, fairly new install with only 67 IPoE DHCP connected subscribers.

Pretty vanilla configuration using bundle interfaces mapped to a loopback with AAA auth, DHCP proxy to an external Kea DHCP server.

Am receiving the below error messages in the logs, customers/services with an IP allocated as per the message are having their DHCP ack messages dropped hence their service goes offline since they can't be allocated an IP address. Seems to be related to the number of subscribers although the active subs count is very low.

ipv4_rib[1185]: %ROUTING-RIB-3-ECMP_ERR_ADD : Path add exceed max number of paths supported by protocol. Table 0xe0000000, prefix xxx.xxx.236.98/32, protocol subscriber, intf 0xb260, tunnelid 0, nexthop_table 0xe0000000, nexthop 0.0.0.0

I also see the below error messages when running debug dhcp ipv4 error which I feel are related.

dhcpd[1054]: DHCPD ERROR: TP3647: FSM call returned error for chaddr_string: 6ccd.d624.5f44, msg_type:5, mode: 4, event: 2
dhcpd[1054]: DHCPD ERROR: TP1665: Proxy process client request packet failed for chaddr 6ccd.d624.5f44
dhcpd[1054]: DHCPD ERROR: TP2995: ARP bulker active on new event cycle: 1
dhcpd[1054]: DHCPD ERROR: TP3677: wildcard list is not initialized, mode 4
dhcpd[1054]: DHCPD ERROR: TP3784: Base profile: high-speed not found
dhcpd[1054]: DHCPD ERROR: TP2751: Unable to add ARP for client 6ccd.d624.5f44 on interface Bundle-Ether10.400, because ARP is not supported on ambiguous VLAN intf.
dhcpd[1054]: DHCPD ERROR: TP3677: wildcard list is not initialized, mode 4
dhcpd[1054]: DHCPD ERROR: TP485: Packet drop

Has anybody come across this before? Is there special memory configuration I need to setup on the ASR9901?

Appreciate any help.

Regards,

Roger

[Giuseppe Larosa]

Hello @RogerLienert15023 ,

>> ipv4_rib[1185]: %ROUTING-RIB-3-ECMP_ERR_ADD : Path add exceed max number of paths supported by protocol. Table 0xe0000000, prefix xxx.xxx.236.98/32, protocol subscriber, intf 0xb260, tunnelid 0, nexthop_table 0xe0000000, nexthop 0.0.0.0

>> fairly new install with only 67 IPoE DHCP connected subscribers.

as far as I know max number of equal cost multipath in IOS XR should be 64.

You are likely assigning the same LAN subnet to each CE node .

>> dhcpd[1054]: DHCPD ERROR: TP2751: Unable to add ARP for client 6ccd.d624.5f44 on interface Bundle-Ether10.400, because ARP is not supported on ambiguous VLAN intf.

give a static MAC address to bundle-ethernet10.400

Hope to help

Giuseppe

[RogerLienert15023]

Hi @Giuseppe Larosa 

With TACs help I've worked out that the routing rib error is due to my Kea DHCP server trying to allocate the same IP address to a different MAC address.

The situation is a customer replacing his router with a new router hence different MAC address. Our Kea DHCP server is configured to allocate an IP based on the Option 82 remote-id field (attached via gfast dslam), so that we can assign static IP's to customers (via DHCP) if required. Each customer has a unique remote-id configured on the dslam.

An existing session already exists on the BNG, for say IP 10.10.236.98/32, the customer changes their router hence new mac-address comes through, Kea allocates the same IP 10.10.236.98/32 to the different MAC and this is where the BNG complains and logs the earlier mentioned message, since the BNG already has this IP address assigned to an active session. The BNG sends a DHCP NACK to the client, the client then sends another discover, repeating the process all over.

This is resolved by manually clearing the original MAC address session on the BNG.

I'd like the BNG to see this new DHCP discover/offer/request message with the same Option 82 remote-id and delete the old session, and then create a new session with the new MAC address.

The current situation means any customers changing/routers cannot get an IP until the lease expires on the BNG.

Not sure if this is possible, any thoughts/comments appreciated.

Regards,

Roger

[marce1000]

     - FYI : https://bst.cloudapps.cisco.com/bugsearch/bug/CSCvy11744

 M.