Bgp ipv6 over ipv4 between cisco 2900 and juniper MX

[unfraget1]

hi guys
I've issue with implement bgp IPv6 over ipv4 between cisco 2921 and juniper MX240.
Session is establishing and at once fails because juniper and cisco couldn't confirm field NLRI inet6-unicast in update messages:

error log with juniper:

bgp_read_v4_update:10317: NOTIFICATION sent to 77.xx.xx.xx (Internal AS XXXX): code 3 (Update Message Error) subcode 9 (error with optional attribute), Reason: peer 77.xx.xx.xx (Internal AS XXXX) UPDATE - NLRI inet6-unicast not negotiated 

bgp_process_caps: mismatch NLRI with 77.xx.xx.xx  (Internal AS XXXX): peer: <inet-unicast inet-vpn-unicast inet6-unicast>(21) us: <inet-unicast inet-vpn-unicast inet-vpn-multicast l2vpn inet6-labeled-unicast inet6-vpn-unicast inet6-vpn-multicast>(3661) 

debug log with cisco:

BGP(1): 77.yy.yy.yy NEXT_HOP is set to self for net 2A02:xxx::y:xx/128,
BGP(0): 77.yy.yy.yy NEXT_HOP is set to self for net 77.xx.xx.xx/31,
BGP(0): (base) 77.yy.yy.yy send UPDATE (format) 77.77.xx.xx.xx/31, next 77.xx.xx.xx, metric 0, path Local
BGP: 77.yy.yy.yy rcv message type 3, length (excl. header) 2
%BGP-3-NOTIFICATION: received from neighbor 77.yy.yy.yy 3/9 (unsupported option specified) 0 bytes
BGP: ses global 77.yy.yy.yy (0x2396AE30:1) Receive NOTIFICATION 3/9 (unsupported option specified) 0 bytes
%BGP-5-NBR_RESET: Neighbor 77.yy.yy.yy reset (BGP Notification received)
BGP: ses global 77.yy.yy.yy (0x2396AE30:1) Reset (BGP Notification received).
BGP: 77.yy.yy.yy went from Established to Closing

i've tried to change IOS version from c2900-universalk9_npe-mz.SPA.153-1.T.bin to c2900-universalk9_npe-mz.SPA.157-3.M2.bin, but to no avail.

How can I solve this problem?

similar problem was discussed here - https://lists.gt.net/nsp/juniper/53080 , but, sollution isn't found,.

[rais]

Do you have family inet6 unicast enabled on Juniper side?

Thanks.

[unfraget1]

no,  i don't 

Ok, i've understood you, but, in network backbone we use inet6 labeled-unicast between PE, juniper can't use inet6 unicast and inet6 labeled-unicast together :(

[Julio E. Moisa]

Hi

Im not really sure about your topology but have you tried with 6PE solution?

[unfraget1]

yes i have. I'm trying to implement 6PE solution.

[rais · Accepted solution]

Then you would need to have send-label under cisco bgp address-family ipv6.

Regards.

[unfraget1]

I configured AF ipv6 on cisco device 

 address-family ipv6
  redistribute connected route-map ipv6_in_bgp
  redistribute static route-map ipv6_in_bgp
  network 2A02/128
  neighbor 77.xx.xx.xx activate
  neighbor 77.xx.xx.xx send-community
  neighbor 77.xx.xx.xx next-hop-self
  neighbor 77.xx.xx.xx soft-reconfiguration inbound
 exit-address-family

juniper side

show configuration protocols bgp group Edge
type internal;
local-address 77.XX.XX.XX;
family inet {
    unicast;
}
family inet-vpn {
    any;
}
family inet6 {
    labeled-unicast {
        explicit-null;
    }
}
family inet6-vpn {
    any;
}
family l2vpn {
    signaling;
}

[rais]

But there is no: 

neighbor 77.xx.xx.xx send-label

 if this an edge setup you should have Juniper/PE side ipv6 under a vrf.

[Julio E. Moisa · Accepted solution]

That is correct, also the following line should be configured on the IPv4 AF only: 

neighbor 77.xx.xx.xx next-hop-self

 And the send-label under the IPv6 AF only. 

neighbor 77.xx.xx.xx send-label

[unfraget1]

rais and Julio Moisa
It worked! bgp session is up.
thank you !!!

[Julio E. Moisa]

Excellent! Im happy to see it is working!

Please dont't forget to mark as answered the comments for future references.

Have a great day!

:-)