GoBGP is an open source tool developed in Go language and running on Linux systems that provides control plane functionality for the BGP protocol. Compared to Quagga/FRRouting, GoBGP has better performance and shorter convergence time, and can be used for larger networks, such as acting as an IXP router. GoBGP can be configured via the gRPC API using multiple languages such as Python, C++, and of course the CLI. goBGP also supports OpenConfig, and its YANG model conforms to draft-ietf-idr-bgp-model-03. Because GoBGP can easily interfere with routing manually and is more involved, it is a good tool for experimentation. In this paper, we present the main features and practices of gobgp.

Background

Installation and Composition

Installation of GoBGP is very simple, just download the tar.gz file from https://github.com/osrg/gobgp/releases and unzip it. The one selected here is v2.27.0.

1
$ tar -xzf gobgp_2.27.0_linux_amd64.tar.gz
  • gobgpd
    • daemon for Gobgp, complete implementation of BGP protocol
    • Can interact with gobgpd via the gRPC API
    • You can also configure bgp through configuration files
  • gobgp
    • Full-featured CLI
    • You can view BGP-related information and configure BGP
  • Configuration file: supports multiple formats toml/yaml/json, etc.

Support Features

  • Full-featured CLI
  • Multiprotocol Support
    • IPv4/Pv6
    • Labeled IPv4/IPv6
    • Labeled IPv4/IPv6
    • EVPN
    • Flowspec IPv4/IPv6/L2
  • Flexible Policy
  • Graceful Restart
    • Both restarting/helper speak role
  • Route Reflector
  • Route Server
  • MRT Dumping
  • BMP
  • RPKI Validation
  • FIB manipulation
  • gRPC API
  • Standard configuration format

Performance Test

Compared with Quagga/FRRouting, GoBGP has better performance and shorter convergence time, and can be applied to larger networks, such as acting as an IXP router. For more performance tests on BGP, see Comparing Open Source BGP stacks with internet routes.

GoBGP performance

Integration with Quagga/Zebra, etc

GoBGP supports only one routing protocol, BGP, but it can be integrated with Zebra to work with Quagga/FRR by way of an API to support multiple routing protocols.

Integration with Quagga/Zebra

GoBGP can be integrated into the Quagga/Zebra system.

GoBGP can be integrated into the Quagga/Zebra system

Using GoBGP

Basic operation

We can start gobgpd as a bgp server to establish a BGP connection with the switch.

Here is the network topology diagram

For example, for

1
2
3
4
5
6
7
8
9
[global.config]
  as = 1002
  router-id = "172.25.0.136"

[[neighbors]]
  [neighbors.config]
    peer-as = 1002
    neighbor-address = "172.25.0.129"
    auth-password: "xxxxx"

Start gobgpd.

1
2
3
4
5
6
/ # ./gobgpd -t yaml -f gobgpd.yaml
{"level":"info","msg":"gobgpd started","time":"2022-03-03T07:28:56Z"}
{"Topic":"Config","level":"info","msg":"Finished reading the config file","time":"2022-03-03T07:28:56Z"}
{"level":"info","msg":"Peer 172.25.0.129 is added","time":"2022-03-03T07:28:56Z"}
{"Topic":"Peer","level":"info","msg":"Add a peer configuration for:172.25.0.129","time":"2022-03-03T07:28:56Z"}
{"Key":"172.25.0.129","State":"BGP_FSM_OPENCONFIRM","Topic":"Peer","level":"info","msg":"Peer Up","time":"2022-03-03T07:29:01Z"}

Check the peer information via gobgp, here the Establ of the State means the connection has been established, if the State is Active then you need to check if the switch configuration is correct.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
/ # ./gobgp neighbor
Peer           AS  Up/Down State       |#Received  Accepted
172.25.0.129 1002 00:01:29 Establ      |        8         8
/ # ./gobgp neighbor 172.25.0.129
BGP neighbor is 172.25.0.129, remote AS 1002
  BGP version 4, remote router ID 172.25.100.4
  BGP state = ESTABLISHED, up for 00:01:34
  BGP OutQ = 0, Flops = 0
  Hold time is 90, keepalive interval is 30 seconds
  Configured hold time is 90, keepalive interval is 30 seconds

  Neighbor capabilities:
    multiprotocol:
        ipv4-unicast: advertised and received
    route-refresh: advertised and received
    extended-nexthop: advertised
        Local:  nlri: ipv4-unicast, nexthop: ipv6
    4-octet-as: advertised and received
  Message statistics:
                         Sent       Rcvd
    Opens:                  1          1
    Notifications:          0          0
    Updates:                0          7
    Keepalives:             4          4
    Route Refresh:          0          0
    Discarded:              0          0
    Total:                  5         12
  Route statistics:
    Advertised:             0
    Received:               8
    Accepted:               8

View global table.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
/ # ./gobgp global rib
   Network              Next Hop             AS_PATH              Age        Attrs
*> 10.0.0.0/24          172.25.0.129         801 45090 45090      00:04:16   [{Origin: ?} {LocalPref: 100}]
*> 10.0.2.0/24          172.25.0.129         801 45090 45090      00:04:16   [{Origin: ?} {LocalPref: 100}]
*> 172.25.0.0/25        172.25.0.129         801 1001             00:04:16   [{Origin: i} {LocalPref: 100}]
*> 172.25.0.128/25      172.25.0.129                              00:04:16   [{Origin: i} {Med: 0} {LocalPref: 100}]
*> 172.25.100.1/32      172.25.0.129         801                  00:04:16   [{Origin: i} {Med: 0} {LocalPref: 100}]
*> 172.25.100.2/32      172.25.0.129         801                  00:04:16   [{Origin: i} {Med: 0} {LocalPref: 100}]
*> 172.25.100.3/32      172.25.0.129         801 1001             00:04:16   [{Origin: i} {LocalPref: 100}]
*> 172.25.100.4/32      172.25.0.129                              00:04:16   [{Origin: i} {Med: 0} {LocalPref: 100}]

View adjacent rib-in and rib-out.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
/ # ./gobgp neighbor 172.25.0.129 adj-in
   ID  Network              Next Hop             AS_PATH              Age        Attrs
   0   10.0.0.0/24          172.25.0.129         801 45090 45090      00:07:18   [{Origin: ?} {LocalPref: 100}]
   0   10.0.2.0/24          172.25.0.129         801 45090 45090      00:07:18   [{Origin: ?} {LocalPref: 100}]
   0   172.25.0.0/25        172.25.0.129         801 1001             00:07:18   [{Origin: i} {LocalPref: 100}]
   0   172.25.0.128/25      172.25.0.129                              00:07:18   [{Origin: i} {Med: 0} {LocalPref: 100}]
   0   172.25.100.1/32      172.25.0.129         801                  00:07:18   [{Origin: i} {Med: 0} {LocalPref: 100}]
   0   172.25.100.2/32      172.25.0.129         801                  00:07:18   [{Origin: i} {Med: 0} {LocalPref: 100}]
   0   172.25.100.3/32      172.25.0.129         801 1001             00:07:18   [{Origin: i} {LocalPref: 100}]
   0   172.25.100.4/32      172.25.0.129                              00:07:18   [{Origin: i} {Med: 0} {LocalPref: 100}]
/ # ./gobgp neighbor 172.25.0.129 adj-out
Network not in table

You can declare the route with the following command.

1
gobgp global rib -a ipv4 add 192.168.1.0/24

Route Reflector

To ensure connectivity between iBGP peers, full connectivity relationships between IBGP peers are required . The Route Reflection mode is a mature alternative to the Full Mesh mode, which decreases dramatically as the cluster size grows, allowing a BGP Speaker (i.e., Route Reflector) to broadcast learned routing information to other BGP Peers, greatly reducing the number of BGP Peer This greatly reduces the number of BGP Peer connections.

For gobgpd, the BGP Server can be supported as a Route Reflector by modifying the configuration file by adding the RouteReflector.RouteReflectorConfig configuration as follows.

  • Node 172.25.0.7 as an RR node to establish a bgp peer with switch 172.25.0.1
  • Node 172.25.0.6 as RR client node, establish bgp peer with RR node 172.25.0.7
  • Node 172.25.0.8 as RR client node, establish bgp peer with RR node 172.25.0.7
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
[global.config]
  as = 1001
  router-id = "172.25.0.7"

[[neighbors]]
  [neighbors.config]
    neighbor-address = "172.25.0.1"
    peer-as = 1001
    auth-password = "xxxxxx"

[[neighbors]]
  [neighbors.config]
    neighbor-address = "172.25.0.6"
    peer-as = 1001
    auth-password = "xxxxxx"

  [neighbors.route-reflector.config]
    route-reflector-client = true
    route-reflector-cluster-id = "172.25.0.137"

[[neighbors]]
  [neighbors.config]
    neighbor-address = "172.25.0.8"
    peer-as = 1001
    auth-password = "xxxxxx"

  [neighbors.route-reflector.config]
    route-reflector-client = true
    route-reflector-cluster-id = "172.25.0.137"

Route Server

There are some scenarios in the existing network, In order to achieve network traffic interoperability, it is usually necessary to perform full connectivity through eBGP . Full connectivity between border devices is more demanding in terms of cost consumption and device performance, and is not conducive to network topology and device expansion. Route Server is similar to IBGP full connectivity using route reflector , which is a device (or multiple devices) used to perform routing services, and its main function is to propagate routes to individual clients (border devices), and the routes published to clients do not modify path attributes such as AS_PATH, Nexthop, MED, etc., thus reducing the border router full connectivity consumption.

As shown in the figure below, an IX (Internet eXchange) contains multiple independent SPs (service providers) that want to interoperate traffic. Each SP has a border router connected to the common switching network. Each SP has its own AS number, and the BGP Router addresses range from 10.0.0.1 to 10.0.0.8.

Route Server

In this case, the 8 BGP Peers are required to establish a full connection. Like iBGP, this full mesh connection is more demanding in terms of cost consumption and device performance, and is not conducive to the expansion of network topology and number of devices.

IX ebgp full mesh

BGP Route Server can simplify SP connectivity as follows.

IX ebgp Route server

The following figure shows the transparent route propagation implemented by route server.

transparent route propagation implemented by route server

For more information about route server, see Route Server.

Route Server is also supported for GoBGP.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
[global.config]
  as = 64512
  router-id = "192.168.255.1"

[[neighbors]]
  [neighbors.config]
    neighbor-address = "10.0.255.1"
    peer-as = 65001
    auth-password = "hoge1"
  [neighbors.transport.config]
    passive-mode = true
  [neighbors.route-server.config]
    route-server-client = true

[[neighbors]]
  [neighbors.config]
    neighbor-address = "10.0.255.2"
    peer-as = 65002
    auth-password = "hoge2"
  [neighbors.transport.config]
    passive-mode = true
  [neighbors.route-server.config]
    route-server-client = true

BGP Policy

Policy is a way to control how a BGP route is inserted into a RIB or broadcast to a BGP Peer, and is divided into two parts Condition and Action. When the Policy is configured and the Condition condition is triggered, an Action action is performed to modify the route.

  • Condition includes prefix, neighbor (source/destination of the route) and aspath, etc.
  • Action includes accept , reject and MED/aspath/community manipulation, etc.

Policy Model

Policy model includes Import Policy and Export Policy

  • Import policy is invoked before best path calculation and pushing routes to RIB.

  • Export policy is invoked after that.

    Policy Model

You can view the policy with the following command.

1
2
$ gobgp global policy import
$ gobgp global policy export

Route Server Policy Model

For the Route Server model, Import and Export policies are specific to a Peer: Import and Export policies are specific to a Peer.

  • The Import policy defines what routes will be imported into the master RIB.

  • The Export policy defines what routes will be exported from the master RIB.

    Route Server Policy Model

1
2
$ gobgp neighbor <neighbor-addr> policy import
$ gobgp neighbor <neighbor-addr> policy export

Policy Structure

A Policy contains multiple Statements, each with its own Condtions and Actions.

Policy Structure

Conditions includes.

  • prefix
  • neighbor
  • aspath
  • aspath length
  • community
  • extended community
  • rpki validation result
  • route type (internal/external/local)
  • large community
  • afi-safi in

Actions includes.

  • accept or reject
  • add/replace/remove community or remove all communities
  • add/subtract or replace MED value
  • set next-hop (specific address/own local address/don’t modify)
  • set local-pref
  • prepend AS number in the AS_PATH attribute

You can view the Policy configuration with the following command.

1
2
3
4
5
6
7
8
$ gobgp policy
$ gobgp policy statement
$ gobgp policy prefix
$ gobgp policy neighbor
$ gobgp policy as-path
$ gobgp policy community
$ gobgp policy ext-community
$ gobgp policy large-community

Policy Configuration

Policy configuration is a bit more complicated, the following are the steps to configure it, which can be found here.

  1. define defined-sets
    1. define prefix-sets
    2. define neighbor-sets
  2. define bgp-defined-sets
    1. define community-sets
    2. define ext-community-sets
    3. define as-path-setList
    4. define large-community-sets
  3. define policy-definitions
  4. attach policies to global rib (or neighbor local rib when neighbor is route-server-client).

Graceful Restart

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
[global.config]
  as = 64512
  router-id = "192.168.255.1"

[[neighbors]]
  [neighbors.config]
    neighbor-address = "10.0.255.1"
    peer-as = 65001
  [neighbors.graceful-restart.config]
    enabled = true

BMP

GoBGP supports the BGP Monitoring Protocol (RFC 7854) for real-time monitoring of the operational status of BGP sessions, including the establishment and closure of peer relationships, routing information, etc.

1
2
3
4
5
6
7
8
[global.config]
  as = 64512
  router-id = "192.168.255.1"

[[bmp-servers]]
  [bmp-servers.config]
    address = "127.0.0.1"
    port=11019

Dynamic Neighbors

In a BGP network, when multiple peers change frequently, if the peers are configured statically, it is necessary to add or remove peers frequently at the local end, which is a great maintenance workload. In this case, you can configure BGP dynamic peer function to make BGP listen for BGP connection requests from specified network segments and dynamically establish BGP peers, and add these peers to the same peer group. This reduces the workload of network maintenance by eliminating the need to add or remove BGP peers at the local end when peer changes occur.

Switches generally support the configuration of Dynamic Neighbors, for example, here is the method of configuring Dynamic Neighbors for Huawei switches, and Dynamic Neighbors are also supported for gobgp.

As shown below, there are two main parts to configure.

  • Create a peer group and describe the basic information of the peer group.
  • Configure the peer group to listen on the 172.40.0.0/16 network segment
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
[global.config]
  as = 65001
  router-id = "172.40.1.2"

[[peer-groups]]
  [peer-groups.config]
    peer-group-name = "sample-group"
    peer-as = 65002
  [[peer-groups.afi-safis]]
    [peer-groups.afi-safis.config]
      afi-safi-name = "ipv4-unicast"
  [[peer-groups.afi-safis]]
    [peer-groups.afi-safis.config]
      afi-safi-name = "ipv4-flowspec"

[[dynamic-neighbors]]
  [dynamic-neighbors.config]
    prefix = "172.40.0.0/16"
    peer-group = "sample-group"

Others

There are a lot of other features on GitHub about MRT/BMP/EVPN, so I won’t go over them here, but you can check the documentation if you need to.

GoBGP Programming

Basic Server

Referring to the documentation provided by the gobgp library, we can implement a simple go bgp server as shown below.

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
package main

import (
 "context"
 "time"

 "github.com/sirupsen/logrus"
 apb "google.golang.org/protobuf/types/known/anypb"

 api "github.com/osrg/gobgp/v3/api"
 "github.com/osrg/gobgp/v3/pkg/log"
 "github.com/osrg/gobgp/v3/pkg/server"
)

func main() {
 log := logrus.New()

  // 创建 BGP Server 实例
 s := server.NewBgpServer(server.LoggerOption(&myLogger{logger: log}))
 go s.Serve()

 // global configuration
 if err := s.StartBgp(context.Background(), &api.StartBgpRequest{
  Global: &api.Global{
   Asn:         65003,
   RouterId:   "10.0.255.254",
   ListenPort: -1, // gobgp won't listen on tcp:179
  },
 }); err != nil {
  log.Fatal(err)
 }

 // monitor the change of the peer state
 if err := s.MonitorPeer(ctx, &api.MonitorPeerRequest{}, func(p *api.Peer) { log.Print(p) }); err != nil {
  log.Fatal(err)
 }

 // neighbor configuration
 n := &api.Peer{
  Conf: &api.PeerConf{
   NeighborAddress: "172.17.0.2",
   PeerAsn:          65002,
  },
 }

 if err := s.AddPeer(context.Background(), &api.AddPeerRequest{
  Peer: n,
 }); err != nil {
  log.Fatal(err)
 }

 // add routes
 nlri, _ := apb.New(&api.IPAddressPrefix{
  Prefix:    "10.0.0.0",
  PrefixLen: 24,
 })

 a1, _ := apb.New(&api.OriginAttribute{
  Origin: 0,
 })
 a2, _ := apb.New(&api.NextHopAttribute{
  NextHop: "10.0.0.1",
 })
 a3, _ := apb.New(&api.AsPathAttribute{
  Segments: []*api.AsSegment{
   {
    Type:    2,
    Numbers: []uint32{6762, 39919, 65000, 35753, 65000},
   },
  },
 })
 attrs := []*apb.Any{a1, a2, a3}

 _, err := s.AddPath(context.Background(), &api.AddPathRequest{
  Path: &api.Path{
   Family: &api.Family{Afi: api.Family_AFI_IP, Safi: api.Family_SAFI_UNICAST},
   Nlri:   nlri,
   Pattrs: attrs,
  },
 })
 if err != nil {
  log.Fatal(err)
 }

 v6Family := &api.Family{
  Afi:  api.Family_AFI_IP6,
  Safi: api.Family_SAFI_UNICAST,
 }

 // add v6 route
 nlri, _ = apb.New(&api.IPAddressPrefix{
  PrefixLen: 64,
  Prefix:    "2001:db8:1::",
 })
 v6Attrs, _ := apb.New(&api.MpReachNLRIAttribute{
  Family:   v6Family,
  NextHops: []string{"2001:db8::1"},
  Nlris:    []*apb.Any{nlri},
 })

 c, _ := apb.New(&api.CommunitiesAttribute{
  Communities: []uint32{100, 200},
 })

 _, err = s.AddPath(context.Background(), &api.AddPathRequest{
  Path: &api.Path{
   Family: v6Family,
   Nlri:   nlri,
   Pattrs: []*apb.Any{a1, v6Attrs, c},
  },
 })
 if err != nil {
  log.Fatal(err)
 }

 s.ListPath(context.Background(), &api.ListPathRequest{Family: v6Family}, func(p *api.Destination) {
  log.Info(p)
 })

 // do something useful here instead of exiting
 time.Sleep(time.Minute * 3)
}

// ...

As you can see, the sample code is relatively simple and mainly uses the following APIs.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
// 创建 BGP Server 实例
func NewBgpServer(opt ...ServerOption) *BgpServer

// 启动 BGP Server
func (s *BgpServer) Serve()

// global 配置
// BGP Server 的 AS 是 65003,RouterId 是 10.0.255.254
api.Global{
  Asn:         65003,
  RouterId:   "10.0.255.254",
  ListenPort: -1, // gobgp won't listen on tcp:179
}

// 根据传入 global 配置,开启 BGP Server 的 BGP 协商
func (s *BgpServer) StartBgp(ctx context.Context, r *api.StartBgpRequest) error

// 观察 BGP Peer 状态变化
func (s *BgpServer) MonitorPeer(ctx context.Context, r *api.MonitorPeerRequest, fn func(*api.Peer)) error

// Peer 信息
api.Peer{
  Conf: &api.PeerConf{
   NeighborAddress: "172.17.0.2",
   PeerAsn:          65002,
  },
 }

// 建立 BGP Peer 连接
func (s *BgpServer) AddPeer(ctx context.Context, r *api.AddPeerRequest) error

// 传递 BGP 路由信息
func (s *BgpServer) AddPath(ctx context.Context, r *api.AddPathRequest) (*api.AddPathResponse, error)

Route Reflector

The api.Peer structure can be configured in more detail so that the joined BGP Peer is acting as an RR client.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
n := &api.Peer{
 Conf: &api.PeerConf{
  NeighborAddress: "172.25.0.6",
  PeerAsn:          1001,
 },
 RouteReflector: &api.RouteReflector{
  RouteReflectorClient: true,
  RouteReflectorClusterId: "172.25.0.7",
 }
}

BMP

Here is a list of several common BMP messages.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
type BMPMessage struct {
 Header     BMPHeader
 PeerHeader BMPPeerHeader
 Body       BMPBody
}

type BMPRouteMonitoring struct {
 BGPUpdate        *bgp.BGPMessage
 BGPUpdatePayload []byte
}

type BMPPeerDownNotification struct {
 Reason          uint8
 BGPNotification *bgp.BGPMessage
 Data            []byte
}

type BMPPeerUpNotification struct {
 LocalAddress    net.IP
 LocalPort       uint16
 RemotePort      uint16
 SentOpenMsg     *bgp.BGPMessage
 ReceivedOpenMsg *bgp.BGPMessage
}

// ...