Exploring BGP with IPMininet
Exploring BGP with IPMininet
IPMininetsupports various routing protocols. In this post, we use it to study how the Border Gateway Protocol operates in a simple network containing only BGP routers. Our virtual lab contains four routers and four hosts:
h2
||
h1 = ra ----- rb ----- rd = h4
| |
+------ rc = h3
TheIPMininetscript for this lab is available fromhttps://github.com/obonaventure/RoutingExamples/blob/master/bgp/simple-bgp.py. There are four main parts in the definition of this topology.
First, we create the BGP routers and specify the prefixes that each router advertises.
# BGP routers
as1ra = self.bgp('as1ra',['2001:1234:1::/64'])
as2rb = self.bgp('as2rb',['2001:1234:2::/64'])
as3rc = self.bgp('as3rc',['2001:1234:3::/64'])
as4rd = self.bgp('as4rd',['2001:1234:4::/64'])
The first router,as1rawill use BGP to advertise prefix2001:1234:1::/64that is assigned to the LAN containing hosth1. Then, we associate each router to an Autonomous System.
# Set AS-ownerships
self.addOverlay(AS(1, (as1ra,)))
self.addOverlay(AS(2, (as2rb,)))
self.addOverlay(AS(3, (as3rc,)))
self.addOverlay(AS(4, (as4rd,)))
This information is used byIPMininetto create the BGP configuration for theQuagga dameon. We can then create the links between the different routers. We configure each interdomain link with its own IPv6 prefix. Kato and Manning have proposed to uselink-local addresses for the eBGP sessionsand several vendors have implemented this feature, but we do not analyse it here.
# Inter-AS links
self.addLink(as1ra, as2rb,
params1={"ip": "2001:12::a/64"},
params2={"ip": "2001:12::b/64"})
self.addLink(as1ra, as3rc,
params1={"ip": "2001:13::a/64"},
params2={"ip": "2001:13::c/64"})
self.addLink(as2rb, as3rc,
params1={"ip": "2001:23::b/64"},
params2={"ip": "2001:23::c/64"})
self.addLink(as2rb, as4rd,
params1={"ip": "2001:24::c/64"},
params2={"ip": "2001:24::d/64"})
The topology and the IPv6 addresses have been allocated. We can now create the peering links.
# Add eBGP peering
bgp_peering(self, as1ra, as2rb)
bgp_peering(self, as1ra, as3rc)
bgp_peering(self, as2rb, as3rc)
bgp_peering(self, as2rb, as4rd)
These methods create the configuration files for the Quagga BGP daemon based on the parameters that we have passed. For example, let us look at the BGP configuration file foras1ra.
hostname as1ra
password zebra
log file /tmp/bgpd_as1ra.log
router bgp 1
bgp router-id 0.0.0.2
bgp bestpath compare-routerid
no auto-summary
neighbor 2001:12::b remote-as 2
neighbor 2001:12::b port 179
neighbor 2001:12::b description as2rb (eBGP)
neighbor 2001:12::b ebgp-multihop
no auto-summary
neighbor 2001:13::c remote-as 3
neighbor 2001:13::c port 179
neighbor 2001:13::c description as3rc (eBGP)
neighbor 2001:13::c ebgp-multihop
address-family ipv6
network 2001:1234:1::/64
neighbor 2001:12::b activate
neighbor 2001:12::b next-hop-self all
neighbor 2001:13::c activate
neighbor 2001:13::c next-hop-self all
exit-address-family
!
!
After having defined thehostname, default password and log file, we start the BGP part of the configuration by indicating the AS number ofas1rawithrouter bgp 1. The next line specifies therouter-idof the BGP router. Each BGP router must have a unique router-id which is usually one of its IPv4 addresses. In our examples, we do not assign IPv4 addresses to the routers andIPMininetautomatically assigns router-id0.0.0.2toas1ra. The lines starting with theneighborcommands define the eBGP sessions betweenas1raandas2rbandas3rc. Each session is an external BGP session and uses port 179 which is the standard BGP port. The last part of the configuration specifies that routeras1rawill advertise the2001:1234:1::/64IPv6 prefix.
We can now start the network withsudo python bgp/simple-bgp.py. Let us first check that all hosts can reach the others. The latest version ofIPMininetincludes an updated version of thepingallcommand which supports IPv4 and IPv6.
mininet>pingall
*** Ping: testing reachability over IPv4 and IPv6
h1 --IPv6--> h4 h3 h2
h2 --IPv6--> h4 h1 h3
h3 --IPv6--> h4 h1 h2
h4 --IPv6--> h1 h2 h3
*** Results: 0% dropped (12/12 received)
This confirms that the routing tables are correct. We can verify some of the paths by usingtraceroute6.
mininet> h1 traceroute6 h2
traceroute to 2001:1234:2::2 (2001:1234:2::2) from 2001:1234:1::1, 30 hops max, 16 byte packets
1 2001:1234:1::a (2001:1234:1::a) 0.088 ms 0.131 ms 0.063 ms
2 2001:12::b (2001:12::b) 0.056 ms 0.045 ms 0.035 ms
3 2001:1234:2::2 (2001:1234:2::2) 0.096 ms 0.039 ms 0.106 ms
mininet> h4 traceroute6 h1
traceroute to 2001:1234:1::1 (2001:1234:1::1) from 2001:1234:4::4, 30 hops max, 16 byte packets
1 2001:1234:4::d (2001:1234:4::d) 0.059 ms 0.057 ms 0.025 ms
2 2001:24::c (2001:24::c) 0.037 ms 0.131 ms 0.045 ms
3 2001:12::a (2001:12::a) 0.053 ms 0.046 ms 0.043 ms
4 2001:1234:1::1 (2001:1234:1::1) 0.043 ms 0.087 ms 0.048 ms
The BGP daemon which runs on each router provides a command line interface (CLI) which can be used to issue commands but also to check the status of the routing protocols. We can start this CLI withtelnet ::1 bgpd. In this command,bgpdis translated into port number 2605 which is defined in the/etc/servicesfile onas1ra.
mininet> as1ra telnet ::1 bgpd
Trying ::1...
Connected to ::1.
Escape character is '^]'.
Hello, this is Quagga (version 1.2.2).
Copyright 1996-2005 Kunihiro Ishiguro, et al.
User Access Verification
Password: zebra
as1ra>A first command isshow bgp summarywhich outputs a summary of the status of the BGP sessions on the local router.
as1ra> sh bgp summary
...
IPv6 Unicast Summary:
---------------------
BGP router identifier 0.0.0.2, local AS number 1
RIB entries 7, using 504 bytes of memory
Peers 2, using 9888 bytes of memory
Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd
2001:12::b 4 2 31 33 0 0 0 00:25:26 3
2001:13::c 4 3 31 34 0 0 0 00:25:26 3
Total number of neighbors 2
Total num. Established sessions 2
Total num. of routes received 6
...
The output is verbose. We only list above the information related to the IPv6 routes that we use. The two eBGP sessions that we have defined in the configuration are active and 3 prefixes have been learned from each of these sessions.
We can obtain more information about the BGP sessions with theshow bgp neighborscommand.
as1ra> show bgp neighbors 2001:13::c
BGP neighbor is 2001:13::c, remote AS 3, local AS 1, external link
Description: as3rc (eBGP)
BGP version 4, remote router ID 0.0.0.4
BGP state = Established, up for 00:46:29
Last read 00:00:29, hold time is 180, keepalive interval is 60 seconds
Neighbor capabilities:
4 Byte AS: advertised and received
Route refresh: advertised and received(old & new)
Address family IPv4 Unicast: advertised and received
Address family IPv6 Unicast: advertised and received
Graceful Restart Capabilty: advertised and received
Remote Restart timer is 120 seconds
Address families by peer:
none
Graceful restart informations:
End-of-RIB send: IPv4 Unicast, IPv6 Unicast
End-of-RIB received: IPv4 Unicast
Message statistics:
Inq depth is 0
Outq depth is 0
Sent Rcvd
Opens: 2 0
Notifications: 0 0
Updates: 5 5
Keepalives: 48 47
Route Refresh: 0 0
Capability: 0 0
Total: 55 52
Minimum time between advertisement runs is 3 seconds
For address family: IPv4 Unicast
Community attribute sent to this neighbor(all)
0 accepted prefixes
For address family: IPv6 Unicast
NEXT_HOP is always this router
Community attribute sent to this neighbor(all)
3 accepted prefixes
Connections established 1; dropped 0
Last reset never
External BGP neighbor may be up to 255 hops away.
Local host: 2001:13::a, Local port: 179
Foreign host: 2001:13::c, Foreign port: 46614
Nexthop: 0.0.0.2
Nexthop global: 2001:13::a
Nexthop local: fe80::9442:91ff:fe35:b5b1
BGP connection: shared network
Read thread: on Write thread: off
Another very important command isshow ipv6 bgp. It outputs the BGP routing table of the router.
as1ra> show ipv6 bgp
BGP table version is 0, local router ID is 0.0.0.2
Status codes: s suppressed, d damped, h history, * valid, > best, = multipath,
i internal, r RIB-failure, S Stale, R Removed
Origin codes: i - IGP, e - EGP, ? - incomplete
Network Next Hop Metric LocPrf Weight Path
*> 2001:1234:1::/64 :: 0 32768 i
* 2001:1234:2::/64 2001:13::c 0 3 2 i
*> 2001:12::b 0 0 2 i
* 2001:1234:3::/64 2001:12::b 0 2 3 i
*> 2001:13::c 0 0 3 i
* 2001:1234:4::/64 2001:13::c 0 3 2 4 i
*> 2001:12::b 0 2 4 i
Displayed 4 out of 7 total prefixes
It is important to understand the information in the BGP routing table. The first prefix,2001:1234:1::/64is the one which is originated byas1ra. Its origin is marked as being interal (see characteriat the end of the line). The first prefix that we learn from BGP is2001:1234:2::/64.as1rahas received two paths for this prefix :AS3:AS2via routeras3rc(nexthop2001:13::c) andAS2via routeras2rb(nexthop2001:12::b). According to theBGP Decision Process, the path viaas2rbis preferred since it has the shortes AS-Path. The*character at the beginning of each line indicates that the paths are valid and the>character indicates the best path which is installed in the FIB and used to forwarded data plane packets. Prefix2001:1234:3::/64is learned from bothas2rbandas3rcand the latter path is preferred. Prefix2001:1234:4::/64is learned from bothas2rbandas3rcand in this case the former is preffered.
Once the BGP sessions are up, it is interesting to observe the BGP messages that are exchanged. Routers only exchange BGP routes when they changed. In our stable topology, we only observe BGP Update messages when sessions are created or stopped. The most frequent BGP messages that we can observe are the BGP Keepalive messages that are exchanged regularly between routers to check that the BGP sessions remain up. As an example, here are the BGP Update messages observed on the BGP session betweenas1raandas2rb.
14:14:40.122437 IP6 (class 0xc0, flowlabel 0x147fb, hlim 255, next-header TCP (6) payload length: 51) 2001:12::a.179 > 2001:12::b.54230: Flags [P.], cksum 0x4074 (incorrect -> 0x6e6e), seq 2508:2527, ack 2509, win 60, options [nop,nop,TS val 7219330 ecr 7204330], length 19: BGP
Keepalive Message (4), length: 19
14:14:40.122637 IP6 (class 0xc0, flowlabel 0x67094, hlim 255, next-header TCP (6) payload length: 51) 2001:12::b.54230 > 2001:12::a.179: Flags [P.], cksum 0x4074 (incorrect -> 0x33c4), seq 2509:2528, ack 2527, win 59, options [nop,nop,TS val 7219330 ecr 7219330], length 19: BGP
Keepalive Message (4), length: 19
14:14:40.122647 IP6 (class 0xc0, flowlabel 0x147fb, hlim 255, next-header TCP (6) payload length: 32) 2001:12::a.179 > 2001:12::b.54230: Flags [.], cksum 0x4061 (incorrect -> 0x37de), seq 2527, ack 2528, win 60, options [nop,nop,TS val 7219330 ecr 7219330], length 0
14:15:40.122793 IP6 (class 0xc0, flowlabel 0x67094, hlim 255, next-header TCP (6) payload length: 51) 2001:12::b.54230 > 2001:12::a.179: Flags [P.], cksum 0x4074 (incorrect -> 0xf918), seq 2528:2547, ack 2527, win 59, options [nop,nop,TS val 7234330 ecr 7219330], length 19: BGP
Keepalive Message (4), length: 19
14:15:40.122827 IP6 (class 0xc0, flowlabel 0x147fb, hlim 255, next-header TCP (6) payload length: 32) 2001:12::a.179 > 2001:12::b.54230: Flags [.], cksum 0x4061 (incorrect -> 0xc29a), seq 2527, ack 2547, win 60, options [nop,nop,TS val 7234330 ecr 7234330], length 0
14:15:40.123176 IP6 (class 0xc0, flowlabel 0x147fb, hlim 255, next-header TCP (6) payload length: 51) 2001:12::a.179 > 2001:12::b.54230: Flags [P.], cksum 0x4074 (incorrect -> 0xbe6c), seq 2527:2546, ack 2547, win 60, options [nop,nop,TS val 7234330 ecr 7234330], length 19: BGP
Looking at the timing between the messages, we can easily observe that BGP Keepalive messages are exchanged every minute. This is the default timer period with our version of Quagga.
Let us now explore what happens when we disable the link betweenas1raandas2rb. For this, we executeip link set dev as2rb-eth0 downon routeras2rband observe the impact of this command on routeras1ra. After having issued this command the link between the two routers appears to freeze as it does not anymore carry packets. If we runtraceroute6from hosth1, we see clearly that the packets have been rerouted via routeras3rc.
mininet> h1 traceroute6 h4
traceroute to 2001:1234:4::4 (2001:1234:4::4) from 2001:1234:1::1, 30 hops max, 16 byte packets
1 2001:1234:1::a (2001:1234:1::a) 0.056 ms 0.04 ms 0.016 ms
2 2001:13::c (2001:13::c) 0.025 ms 0.025 ms 0.022 ms
3 2001:23::b (2001:23::b) 0.023 ms 0.095 ms 0.085 ms
4 2001:24::d (2001:24::d) 0.051 ms 0.033 ms 0.019 ms
5 2001:1234:4::4 (2001:1234:4::4) 0.025 ms 0.028 ms 0.017 ms
To understand the reason for this change of route, let us look at the content of the BGP routing table onas1ra:
as1ra> sh ipv6 bgp
BGP table version is 0, local router ID is 0.0.0.2
Status codes: s suppressed, d damped, h history, * valid, > best, = multipath,
i internal, r RIB-failure, S Stale, R Removed
Origin codes: i - IGP, e - EGP, ? - incomplete
Network Next Hop Metric LocPrf Weight Path
*> 2001:1234:1::/64 :: 0 32768 i
*> 2001:1234:2::/64 2001:13::c 0 3 2 i
2001:12::b 0 0 2 i
2001:1234:3::/64 2001:12::b 0 2 3 i
*> 2001:13::c 0 0 3 i
*> 2001:1234:4::/64 2001:13::c 0 3 2 4 i
2001:12::b 0 2 4 i
Displayed 4 out of 7 total prefixes
The routes that were learned formas2rbare still part of the BGP routing table, but they are now marked as invalid (the corresponding line does not contain the*character anymore. BGP considers these routes to be invalid becauseas1radoes not know how to reach the corresponding BGP nexthop (2001:12::b). If we look at the FIB on routeras1ra, we clearly observe that the router does not have a route to reach prefis2001:12::b/64.
mininet> as1ra ip -6 route
2001:12::/64 dev as1ra-eth0 proto kernel metric 256 linkdown pref medium
2001:13::/64 dev as1ra-eth1 proto kernel metric 256 pref medium
2001:1234:1::/64 dev as1ra-eth2 proto kernel metric 256 pref medium
2001:1234:2::/64 via fe80::e827:73ff:fe2d:d497 dev as1ra-eth1 proto zebra metric 20 pref medium
2001:1234:3::/64 via fe80::e827:73ff:fe2d:d497 dev as1ra-eth1 proto zebra metric 20 pref medium
2001:1234:4::/64 via fe80::e827:73ff:fe2d:d497 dev as1ra-eth1 proto zebra metric 20 pref medium
fe80::/64 dev as1ra-eth0 proto kernel metric 256 linkdown pref medium
fe80::/64 dev as1ra-eth1 proto kernel metric 256 pref medium
fe80::/64 dev as1ra-eth2 proto kernel metric 256 pref medium
If we look at the packets on theas1ra-as2rblink, we observe the last packets excahnged on this link:
14:19:40.125910 IP6 (class 0xc0, flowlabel 0x67094, hlim 255, next-header TCP (6) payload length: 51) 2001:12::b.54230 > 2001:12::a.179: Flags [P.], cksum 0x4074 (incorrect -> 0xe912), seq 2604:2623, ack 2622, win 59, options [nop,nop,TS val 7294331 ecr 7294330], length 19: BGP
Keepalive Message (4), length: 19
14:19:40.166495 IP6 (class 0xc0, flowlabel 0x147fb, hlim 255, next-header TCP (6) payload length: 32) 2001:12::a.179 > 2001:12::b.54230: Flags [.], cksum 0x4061 (incorrect -> 0xed21), seq 2622, ack 2623, win 60, options [nop,nop,TS val 7294341 ecr 7294331], length 0
A close look at the logs ofas1ra1shows that the BGP session betweenas1raandas2rbwas considered to be down 3 minutes after the link became down. This is by default BGP sends BGP Keepalive messages every 60 seconds and considers the session to be down when it has not received any response to three successive BGP Keepalive messages.
mininet> as1ra tail /tmp/bgpd_as1ra.log
2017/12/15 14:22:40 BGP: %NOTIFICATION: sent to neighbor 2001:12::b 4/0 (Hold Timer Expired) 0 bytes
2017/12/15 14:22:40 BGP: Notification sent to neighbor 2001:12::b:0: type 4/0
2017/12/15 14:22:40 BGP: %ADJCHANGE: neighbor 2001:12::b Down BGP Notification send
2017/12/15 14:28:00 BGP: Vty connection from ::1
This is confirmed by the BGP summary that shows that routeras1rais trying to reestablish the session withas2rb.
as1ra> sh bgp summary
IPv6 Unicast Summary:
---------------------
BGP router identifier 0.0.0.2, local AS number 1
RIB entries 7, using 504 bytes of memory
Peers 2, using 9888 bytes of memory
Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd
2001:12::b 4 2 193 200 0 0 0 00:08:35 Connect
2001:13::c 4 3 204 208 0 0 0 03:18:35 3
At this point, we can re-enable the link between the two routers withip link set dev as2rb-eth0 upand reconfigure the static address onas2rbwithip -6 addr add 2001:12::b/64 dev as2rb-eth0. Shortly after, the two routers attempt again to restablish the session between themselves.
A BGP session is always established over a TCP connection. We thus observe first the three-way handshake that establishes the TCP connection on port 179:
14:45:28.402708 IP6 (class 0xc0, flowlabel 0x2900f, hlim 255, next-header TCP (6) payload length: 40) 2001:12::a.38394 > 2001:12::b.179: Flags [S], cksum 0x4069 (incorrect -> 0xb993), seq 3505799218, win 28800, options [mss 1440,sackOK,TS val 7681400 ecr 0,nop,wscale 9], length 0
14:45:28.402766 IP6 (class 0xc0, flowlabel 0x43cfb, hlim 255, next-header TCP (6) payload length: 40) 2001:12::b.179 > 2001:12::a.38394: Flags [S.], cksum 0x4069 (incorrect -> 0xa74c), seq 1701541837, ack 3505799219, win 28560, options [mss 1440,sackOK,TS val 7681400 ecr 7681400,nop,wscale 9], length 0
14:45:28.402780 IP6 (class 0xc0, flowlabel 0x2900f, hlim 255, next-header TCP (6) payload length: 32) 2001:12::a.38394 > 2001:12::b.179: Flags [.], cksum 0x4061 (incorrect -> 0x455e), seq 1, ack 1, win 57, options [nop,nop,TS val 7681400 ecr 7681400], length 0
14:45:28.402951 IP6 (class 0xc0, flowlabel 0x2900f, hlim 255, next-header TCP (6) payload length: 99) 2001:12::a.38394 > 2001:12::b.179: Flags [P.], cksum 0x40a4 (incorrect -> 0x7603), seq 1:68, ack 1, win 57, options [nop,nop,TS val 7681400 ecr 7681400], length 67: BGP
Open Message (1), length: 67
Version 4, my AS 1, Holdtime 180s, ID 0.0.0.2
Optional parameters, length: 38
Option Capabilities Advertisement (2), length: 6
Multiprotocol Extensions (1), length: 4
AFI IPv4 (1), SAFI Unicast (1)
0x0000: 0001 0001
Option Capabilities Advertisement (2), length: 6
Multiprotocol Extensions (1), length: 4
AFI IPv6 (2), SAFI Unicast (1)
0x0000: 0002 0001
Option Capabilities Advertisement (2), length: 2
Route Refresh (Cisco) (128), length: 0
Option Capabilities Advertisement (2), length: 2
Route Refresh (2), length: 0
Option Capabilities Advertisement (2), length: 6
32-Bit AS Number (65), length: 4
4 Byte AS 1
0x0000: 0000 0001
Option Capabilities Advertisement (2), length: 4
Graceful Restart (64), length: 2
Restart Flags: [none], Restart Time 120s
0x0000: 0078
The first BGP message that we observe is theOpenmessage which provides basic information about the session and allows to advertise that are supported by the initiating router (in this case,as1ra). The router provides the BGP version (4), its AS number (1), the timer to consider a session to be down (180s) and its router id. It also advertises optional capabilities, including the ability to support IPv6.
as2rbreplies with a BGP message that contains its AS number (2), hold timer, router-id and optional capabilities. This message also contains a keepalive.
14:45:28.403426 IP6 (class 0xc0, flowlabel 0x43cfb, hlim 255, next-header TCP (6) payload length: 118) 2001:12::b.179 > 2001:12::a.38394: Flags [P.], cksum 0x40b7 (incorrect -> 0x61a8), seq 1:87, ack 68, win 56, options [nop,nop,TS val 7681400 ecr 7681400], length 86: BGP
Open Message (1), length: 67
Version 4, my AS 2, Holdtime 180s, ID 0.0.0.3
Optional parameters, length: 38
Option Capabilities Advertisement (2), length: 6
Multiprotocol Extensions (1), length: 4
AFI IPv4 (1), SAFI Unicast (1)
0x0000: 0001 0001
Option Capabilities Advertisement (2), length: 6
Multiprotocol Extensions (1), length: 4
AFI IPv6 (2), SAFI Unicast (1)
0x0000: 0002 0001
Option Capabilities Advertisement (2), length: 2
Route Refresh (Cisco) (128), length: 0
Option Capabilities Advertisement (2), length: 2
Route Refresh (2), length: 0
Option Capabilities Advertisement (2), length: 6
32-Bit AS Number (65), length: 4
4 Byte AS 2
0x0000: 0000 0002
Option Capabilities Advertisement (2), length: 4
Graceful Restart (64), length: 2
Restart Flags: [none], Restart Time 120s
0x0000: 0078
Keepalive Message (4), length: 19
We can then observe the BGP Update messages that advertise routes.as1rais the first to send its update:
14:45:29.406999 IP6 (class 0xc0, flowlabel 0x2900f, hlim 255, next-header TCP (6) payload length: 463) 2001:12::a.38394 > 2001:12::b.179: Flags [P.], cksum 0x4210 (incorrect -> 0xe1af), seq 106:537, ack 106, win 57, options [nop,nop,TS val 7681651 ecr 7681400], length 431: BGP
Update Message (2), length: 23
End-of-Rib Marker (empty NLRI)
Update Message (2), length: 94
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::a, fe80::20ca:6fff:fe8a:98a4, nh-length: 32, no SNPA
2001:1234:1::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000a fe80 0000 0000 0000 20ca 6fff
0x0020: fe8a 98a4 0040 2001 1234 0001 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 6, Flags [TE]: 1
0x0000: 0201 0000 0001
Multi Exit Discriminator (4), length: 4, Flags [O]: 0
0x0000: 0000 0000
Update Message (2), length: 95
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::a, fe80::20ca:6fff:fe8a:98a4, nh-length: 32, no SNPA
2001:1234:2::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000a fe80 0000 0000 0000 20ca 6fff
0x0020: fe8a 98a4 0040 2001 1234 0002 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 14, Flags [TE]: 1 3 2
0x0000: 0203 0000 0001 0000 0003 0000 0002
Update Message (2), length: 91
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::a, fe80::20ca:6fff:fe8a:98a4, nh-length: 32, no SNPA
2001:1234:3::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000a fe80 0000 0000 0000 20ca 6fff
0x0020: fe8a 98a4 0040 2001 1234 0003 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 10, Flags [TE]: 1 3
0x0000: 0202 0000 0001 0000 0003
Update Message (2), length: 99
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::a, fe80::20ca:6fff:fe8a:98a4, nh-length: 32, no SNPA
2001:1234:4::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000a fe80 0000 0000 0000 20ca 6fff
0x0020: fe8a 98a4 0040 2001 1234 0004 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 18, Flags [TE]: 1 3 2 4
0x0000: 0204 0000 0001 0000 0003 0000 0002 0000
0x0010: 0004
Update Message (2), length: 29
Multi-Protocol Unreach NLRI (15), length: 3, Flags [O]:
AFI: IPv6 (2), SAFI: Unicast (1)
End-of-Rib Marker (empty NLRI)
0x0000: 0002 01
For efficiency reasons, BGP packs multiple update messages in the same TCP segment. Here, we can obser the different prefixes, their nexthops and the associated AS Path. Quagga advertises both the IPv6 address and the link-local address for each nexthop.as2rbsends its BGP Update shortly after:
14:45:29.407455 IP6 (class 0xc0, flowlabel 0x43cfb, hlim 255, next-header TCP (6) payload length: 455) 2001:12::b.179 > 2001:12::a.38394: Flags [P.], cksum 0x4208 (incorrect -> 0xeab5), seq 106:529, ack 537, win 58, options [nop,nop,TS val 7681651 ecr 7681651], length 423: BGP
Update Message (2), length: 23
End-of-Rib Marker (empty NLRI)
Update Message (2), length: 95
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::b, fe80::5c58:b0ff:fe0d:ef98, nh-length: 32, no SNPA
2001:1234:1::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000b fe80 0000 0000 0000 5c58 b0ff
0x0020: fe0d ef98 0040 2001 1234 0001 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 14, Flags [TE]: 2 3 1
0x0000: 0203 0000 0002 0000 0003 0000 0001
Update Message (2), length: 94
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::b, fe80::5c58:b0ff:fe0d:ef98, nh-length: 32, no SNPA
2001:1234:2::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000b fe80 0000 0000 0000 5c58 b0ff
0x0020: fe0d ef98 0040 2001 1234 0002 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 6, Flags [TE]: 2
0x0000: 0201 0000 0002
Multi Exit Discriminator (4), length: 4, Flags [O]: 0
0x0000: 0000 0000
Update Message (2), length: 91
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::b, fe80::5c58:b0ff:fe0d:ef98, nh-length: 32, no SNPA
2001:1234:3::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000b fe80 0000 0000 0000 5c58 b0ff
0x0020: fe0d ef98 0040 2001 1234 0003 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 10, Flags [TE]: 2 3
0x0000: 0202 0000 0002 0000 0003
Update Message (2), length: 91
Multi-Protocol Reach NLRI (14), length: 46, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
nexthop: 2001:12::b, fe80::5c58:b0ff:fe0d:ef98, nh-length: 32, no SNPA
2001:1234:4::/64
0x0000: 0002 0120 2001 0012 0000 0000 0000 0000
0x0010: 0000 000b fe80 0000 0000 0000 5c58 b0ff
0x0020: fe0d ef98 0040 2001 1234 0004 0000
Origin (1), length: 1, Flags [T]: IGP
0x0000: 00
AS Path (2), length: 10, Flags [TE]: 2 4
0x0000: 0202 0000 0002 0000 0004
Update Message (2), length: 29
Multi-Protocol Unreach NLRI (15), length: 3, Flags [O]:
AFI: IPv6 (2), SAFI: Unicast (1)
End-of-Rib Marker (empty NLRI)
0x0000: 0002 01
Shortly after, we observe two additional messages whereas1rawithdraws the routes for2001:1234:2::/64and2001:1234:4::/64because its best path to reach this destination is now the path viaas2rb.
14:45:29.458127 IP6 (class 0xc0, flowlabel 0x2900f, hlim 255, next-header TCP (6) payload length: 80) 2001:12::a.38394 > 2001:12::b.179: Flags [P.], cksum 0x4091 (incorrect -> 0x538c), seq 537:585, ack 529, win 59, options [nop,nop,TS val 7681664 ecr 7681651], length 48: BGP
Update Message (2), length: 48
Multi-Protocol Unreach NLRI (15), length: 21, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
2001:1234:2::/64
2001:1234:4::/64
0x0000: 0002 0140 2001 1234 0002 0000 4020 0112
0x0010: 3400 0400 00
We can confirm this by looking at the BGP routing table ofas1ra:
as1ra> sh bgp ipv6
BGP table version is 0, local router ID is 0.0.0.2
Status codes: s suppressed, d damped, h history, * valid, > best, = multipath,
i internal, r RIB-failure, S Stale, R Removed
Origin codes: i - IGP, e - EGP, ? - incomplete
Network Next Hop Metric LocPrf Weight Path
*> 2001:1234:1::/64 :: 0 32768 i
*> 2001:1234:2::/64 2001:12::b 0 0 2 i
* 2001:13::c 0 3 2 i
* 2001:1234:3::/64 2001:12::b 0 2 3 i
*> 2001:13::c 0 0 3 i
*> 2001:1234:4::/64 2001:12::b 0 2 4 i
* 2001:13::c 0 3 2 4 i
Displayed 4 out of 7 total prefixes
as2rbalso sends a withdraw for prefix2001:1234:1::/64for similar reasons.
14:45:29.458229 IP6 (class 0xc0, flowlabel 0x43cfb, hlim 255, next-header TCP (6) payload length: 71) 2001:12::b.179 > 2001:12::a.38394: Flags [P.], cksum 0x4088 (incorrect -> 0x98db), seq 529:568, ack 585, win 58, options [nop,nop,TS val 7681664 ecr 7681664], length 39: BGP
Update Message (2), length: 39
Multi-Protocol Unreach NLRI (15), length: 12, Flags [OE]:
AFI: IPv6 (2), SAFI: Unicast (1)
2001:1234:1::/64
0x0000: 0002 0140 2001 1234 0001 0000
At this point, the BGP session between the two routers is stable and only Keepalive messages are exchanged.