BGP: Introduce new cli to get bgp state

This change introduces new BGP status CLI : 'cilium bgp peers'.
Status of BGP peers is returned from all or specific node in json or summary formats.

Signed-off-by: harsimran pabla <[email protected]>

CODEOWNERS

@@ -4,6 +4,7 @@
 # @cilium/cli                Commandline interfaces
 # @cilium/contributing       Developer documentation & tools
 # @cilium/github-sec         GitHub security (handling of secrets, consequences of pull_request_target, etc.)
+# @cilium/sig-bgp            BGP integration
 # @cilium/sig-clustermesh    Clustermesh and external workloads
 # @cilium/sig-hubble         Hubble integration
 # @cilium/sig-k8s            K8s integration, K8s CNI plugin
@@ -22,6 +23,7 @@
 /.github/kind-config*.yaml @cilium/ci-structure
 /.github/tools/ @cilium/ci-structure
 /.github/workflows/ @cilium/github-sec @cilium/ci-structure
+/bgp/ @cilium/sig-bgp
 /cmd/ @cilium/cli
 /clustermesh/ @cilium/sig-clustermesh
 /connectivity/ @cilium/cli

bgp/bgp.go

@@ -0,0 +1,61 @@
+// SPDX-License-Identifier: Apache-2.0
+// Copyright Authors of Cilium
+
+package bgp
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"time"
+
+	corev1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
+	"github.com/cilium/cilium-cli/k8s"
+)
+
+// Parameters contains options for CLI
+type Parameters struct {
+	CiliumNamespace  string
+	AgentPodSelector string
+	NodeName         string
+	Writer           io.Writer
+	WaitDuration     time.Duration
+	Output           string
+}
+
+// Status is used to get bgp state from cilium agents
+type Status struct {
+	client     *k8s.Client
+	params     Parameters
+	ciliumPods []*corev1.Pod
+}
+
+// NewStatus returns new bgp.Status struct
+func NewStatus(client *k8s.Client, p Parameters) *Status {
+	return &Status{
+		client: client,
+		params: p,
+	}
+}
+
+// initTargetCiliumPods stores cilium agent pods in the status.ciliumPods.
+// If node selector option is specified then only that nodes' cilium-agent
+// pod is stored else all cilium-agents in the cluster are stored.
+func (s *Status) initTargetCiliumPods(ctx context.Context) error {
+	opts := metav1.ListOptions{LabelSelector: s.params.AgentPodSelector}
+	if s.params.NodeName != "" {
+		opts.FieldSelector = fmt.Sprintf("spec.nodeName=%s", s.params.NodeName)
+	}
+
+	ciliumPods, err := s.client.ListPods(ctx, s.params.CiliumNamespace, opts)
+	if err != nil {
+		return fmt.Errorf("unable to list Cilium pods: %w", err)
+	}
+
+	for _, ciliumPod := range ciliumPods.Items {
+		s.ciliumPods = append(s.ciliumPods, ciliumPod.DeepCopy())
+	}
+	return nil
+}

bgp/peers.go

@@ -0,0 +1,183 @@
+// SPDX-License-Identifier: Apache-2.0
+// Copyright Authors of Cilium
+
+package bgp
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+	"sort"
+	"strings"
+	"sync"
+	"text/tabwriter"
+	"time"
+
+	"github.com/cilium/cilium/api/v1/models"
+	corev1 "k8s.io/api/core/v1"
+
+	"github.com/cilium/cilium-cli/defaults"
+	"github.com/cilium/cilium-cli/status"
+)
+
+const (
+	padding     = 3
+	minWidth    = 5
+	paddingChar = ' '
+)
+
+// GetPeeringState gets peering state from all/specific cilium agent pods.
+func (s *Status) GetPeeringState(ctx context.Context) error {
+	ctx, cancelFn := context.WithTimeout(ctx, s.params.WaitDuration)
+	defer cancelFn()
+
+	err := s.initTargetCiliumPods(ctx)
+	if err != nil {
+		return err
+	}
+
+	res, err := s.fetchPeeringStateConcurrently(ctx)
+	if err != nil {
+		return err
+	}
+
+	return s.writeStatus(res)
+}
+
+func (s *Status) fetchPeeringStateConcurrently(ctx context.Context) (map[string][]*models.BgpPeer, error) {
+	allFetchedData := make(map[string][]*models.BgpPeer)
+
+	// res contains data returned from cilium pod
+	type res struct {
+		nodeName string
+		data     []*models.BgpPeer
+		err      error
+	}
+	resCh := make(chan res)
+
+	var wg sync.WaitGroup
+
+	// max number of concurrent go routines will be number of cilium agent pods
+	wg.Add(len(s.ciliumPods))
+
+	// concurrently fetch state from each cilium pod
+	for _, pod := range s.ciliumPods {
+		go func(ctx context.Context, pod *corev1.Pod) {
+			defer wg.Done()
+
+			peers, err := s.fetchPeeringStateFromPod(ctx, pod)
+			resCh <- res{
+				nodeName: pod.Spec.NodeName,
+				data:     peers,
+				err:      err,
+			}
+		}(ctx, pod)
+	}
+
+	// close resCh when data from all nodes is collected
+	go func() {
+		wg.Wait()
+		close(resCh)
+	}()
+
+	// read from the channel till it is closed.
+	// on error, store error and continue to next node.
+	var err error
+	for fetchedData := range resCh {
+		if fetchedData.err != nil {
+			err = errors.Join(err, fetchedData.err)
+		} else {
+			allFetchedData[fetchedData.nodeName] = fetchedData.data
+		}
+	}
+
+	return allFetchedData, err
+}
+
+func (s *Status) fetchPeeringStateFromPod(ctx context.Context, pod *corev1.Pod) ([]*models.BgpPeer, error) {
+	cmd := []string{"cilium", "bgp", "peers", "-o", "json"}
+	output, err := s.client.ExecInPod(ctx, pod.Namespace, pod.Name, defaults.AgentContainerName, cmd)
+	if err != nil {
+		return nil, fmt.Errorf("failed to fetch bgp state from %s: %v", pod.Name, err)
+	}
+
+	bgpPeers := make([]*models.BgpPeer, 0)
+
+	err = json.Unmarshal(output.Bytes(), &bgpPeers)
+	if err != nil {
+		return nil, fmt.Errorf("failed to unmarshal bgp state from %s: %v", pod.Name, err)
+	}
+
+	return bgpPeers, nil
+}
+
+func (s *Status) writeStatus(res map[string][]*models.BgpPeer) error {
+	if s.params.Output == status.OutputJSON {
+		jsonStatus, err := json.MarshalIndent(res, "", " ")
+		if err != nil {
+			return err
+		}
+		fmt.Println(string(jsonStatus))
+	} else {
+		printSummary(os.Stdout, res)
+	}
+
+	return nil
+}
+
+func printSummary(out io.Writer, peersPerNode map[string][]*models.BgpPeer) {
+	// sort by node names
+	var nodes []string
+	for node := range peersPerNode {
+		nodes = append(nodes, node)
+	}
+	sort.Strings(nodes)
+
+	// sort peers per node
+	for _, peers := range peersPerNode {
+		// sort by local AS, if peers from same AS then sort by peer address.
+		sort.Slice(peers, func(i, j int) bool {
+			return peers[i].LocalAsn < peers[j].LocalAsn || peers[i].PeerAddress < peers[j].PeerAddress
+		})
+	}
+
+	// tab writer with min width 5 and padding 3
+	w := tabwriter.NewWriter(out, minWidth, 0, padding, paddingChar, 0)
+	fmt.Fprintln(w, "Node\tLocal AS\tPeer AS\tPeer Address\tSession State\tUptime\tFamily\tReceived\tAdvertised")
+
+	for _, node := range nodes {
+		peers := peersPerNode[node]
+
+		for i, peer := range peers {
+			if i == 0 {
+				// print name for first row of peers
+				fmt.Fprintf(w, "%s\t", node)
+			} else {
+				// skip name for all rest of the peers
+				fmt.Fprint(w, "\t")
+			}
+
+			fmt.Fprintf(w, "%d\t", peer.LocalAsn)
+			fmt.Fprintf(w, "%d\t", peer.PeerAsn)
+			fmt.Fprintf(w, "%s\t", peer.PeerAddress)
+			fmt.Fprintf(w, "%s\t", peer.SessionState)
+			fmt.Fprintf(w, "%s\t", time.Duration(peer.UptimeNanoseconds).Round(time.Second).String())
+
+			for j, afisafi := range peer.Families {
+				if j > 0 {
+					// skip space for session info
+					fmt.Fprint(w, strings.Repeat("\t", 6))
+				}
+
+				fmt.Fprintf(w, "%s/%s\t", afisafi.Afi, afisafi.Safi)
+				fmt.Fprintf(w, "%d\t", afisafi.Received)
+				fmt.Fprintf(w, "%d\t", afisafi.Advertised)
+				fmt.Fprintf(w, "\n")
+			}
+		}
+	}
+	w.Flush()
+}