New asynchronous, caching DNS resolver for reverse resolutions

Add nodes for the remote side of connections iff we have a DNS reverse resolution for the IP.
Unit test for the resolver

probe/endpoint/reporter.go

@@ -26,6 +26,7 @@ type Reporter struct {
 	includeNAT       bool
 	conntracker      *Conntracker
 	natmapper        *natmapper
+	revResolver      *reverseResolver
 }
 
 // SpyDuration is an exported prometheus metric
@@ -64,12 +65,16 @@ func NewReporter(hostID, hostName string, includeProcesses bool, useConntrack bo
 			log.Printf("Failed to start natMapper: %v", err)
 		}
 	}
+
+	revRes := newReverseResolver()
+
 	return &Reporter{
 		hostID:           hostID,
 		hostName:         hostName,
 		includeProcesses: includeProcesses,
 		conntracker:      conntracker,
 		natmapper:        natmapper,
+		revResolver:      revRes,
 	}
 }
 
@@ -81,6 +86,7 @@ func (r *Reporter) Stop() {
 	if r.natmapper != nil {
 		r.natmapper.Stop()
 	}
+	r.revResolver.Stop()
 }
 
 // Report implements Reporter.
@@ -145,6 +151,13 @@ func (r *Reporter) addConnection(rpt *report.Report, localAddr, remoteAddr strin
 			})
 		)
 
+		// in case we have a reverse resolution for the IP, we can use it for the name...
+		if revRemoteName, err := r.revResolver.Get(remoteAddr, false); err == nil {
+			remoteNode = remoteNode.AddMetadata(map[string]string{
+				"name": revRemoteName,
+			})
+		}
+
 		if localIsClient {
 			// New nodes are merged into the report so we don't need to do any counting here; the merge does it for us.
 			localNode = localNode.WithEdge(remoteAddressNodeID, report.EdgeMetadata{
@@ -177,6 +190,13 @@ func (r *Reporter) addConnection(rpt *report.Report, localAddr, remoteAddr strin
 			})
 		)
 
+		// in case we have a reverse resolution for the IP, we can use it for the name...
+		if revRemoteName, err := r.revResolver.Get(remoteAddr, false); err == nil {
+			remoteNode = remoteNode.AddMetadata(map[string]string{
+				"name": revRemoteName,
+			})
+		}
+
 		if localIsClient {
 			// New nodes are merged into the report so we don't need to do any counting here; the merge does it for us.
 			localNode = localNode.WithEdge(remoteEndpointNodeID, report.EdgeMetadata{

probe/endpoint/reporter_test.go

@@ -54,7 +54,7 @@ var (
 			LocalAddress:  fixLocalAddress,
 			LocalPort:     fixLocalPort,
 			RemoteAddress: fixRemoteAddress,
-			RemotePort:    fixRemotePortB,
+			RemotePort:    fixRemotePort,
 			Proc: procspy.Proc{
 				PID:  fixProcessPID,
 				Name: fixProcessName,

probe/endpoint/resolver.go

@@ -0,0 +1,85 @@
+package endpoint
+
+import (
+	"net"
+	"strings"
+	"time"
+
+	"github.com/bluele/gcache"
+)
+
+const (
+	rAddrCacheLen        = 500 // Default cache length
+	rAddrBacklog         = 1000
+	rAddrCacheExpiration = 30 * time.Minute
+)
+
+type revResFunc func(addr string) (names []string, err error)
+
+type revResRequest struct {
+	address string
+	done    chan struct{}
+}
+
+// ReverseResolver is a caching, reverse resolver
+type reverseResolver struct {
+	addresses chan revResRequest
+	cache     gcache.Cache
+	resolver  revResFunc
+}
+
+// NewReverseResolver starts a new reverse resolver that
+// performs reverse resolutions and caches the result.
+func newReverseResolver() *reverseResolver {
+	r := reverseResolver{
+		addresses: make(chan revResRequest, rAddrBacklog),
+		cache:     gcache.New(rAddrCacheLen).LRU().Expiration(rAddrCacheExpiration).Build(),
+		resolver:  net.LookupAddr,
+	}
+	go r.loop()
+	return &r
+}
+
+// Get the reverse resolution for an IP address if already in the cache,
+// a gcache.NotFoundKeyError error otherwise.
+// Note: it returns one of the possible names that can be obtained for that IP.
+func (r *reverseResolver) Get(address string, wait bool) (string, error) {
+	val, err := r.cache.Get(address)
+	if err == nil {
+		return val.(string), nil
+	}
+	if err == gcache.NotFoundKeyError {
+		request := revResRequest{address: address, done: make(chan struct{})}
+		// we trigger a asynchronous reverse resolution when not cached
+		select {
+		case r.addresses <- request:
+			if wait {
+				<-request.done
+			}
+		default:
+		}
+	}
+	return "", err
+}
+
+func (r *reverseResolver) loop() {
+	throttle := time.Tick(time.Second / 10)
+	for request := range r.addresses {
+		<-throttle // rate limit our DNS resolutions
+		// and check if the answer is already in the cache
+		if _, err := r.cache.Get(request.address); err == nil {
+			continue
+		}
+		names, err := r.resolver(request.address)
+		if err == nil && len(names) > 0 {
+			name := strings.TrimRight(names[0], ".")
+			r.cache.Set(request.address, name)
+		}
+		close(request.done)
+	}
+}
+
+// Stop the async reverse resolver
+func (r *reverseResolver) Stop() {
+	close(r.addresses)
+}

probe/endpoint/resolver_internal_test.go

@@ -0,0 +1,41 @@
+package endpoint
+
+import (
+	"errors"
+	"testing"
+)
+
+func TestReverseResolver(t *testing.T) {
+	tests := map[string]string{
+		"8.8.8.8": "google-public-dns-a.google.com",
+		"8.8.4.4": "google-public-dns-b.google.com",
+	}
+
+	revRes := newReverseResolver()
+
+	// use a mocked resolver function
+	revRes.resolver = func(addr string) (names []string, err error) {
+		if name, ok := tests[addr]; ok {
+			return []string{name}, nil
+		}
+		return []string{}, errors.New("invalid IP")
+	}
+
+	// first time: no names are returned for our reverse resolutions
+	for ip := range tests {
+		if have, err := revRes.Get(ip, true); have != "" || err == nil {
+			t.Errorf("we didn't get an error, or the cache was not empty, when trying to resolve '%q'", ip)
+		}
+	}
+
+	// so, if we check again these IPs, we should have the names now
+	for ip, want := range tests {
+		have, err := revRes.Get(ip, true)
+		if err != nil {
+			t.Errorf("%s: %v", ip, err)
+		}
+		if want != have {
+			t.Errorf("%s: want %q, have %q", ip, want, have)
+		}
+	}
+}

render/detailed_node.go

@@ -205,8 +205,11 @@ func OriginTable(r report.Report, originID string, addHostTags bool, addContaine
 
 func connectionDetailsRows(topology report.Topology, originID string) []Row {
 	rows := []Row{}
-	labeler := func(nodeID string) (string, bool) {
+	labeler := func(nodeID string, meta map[string]string) (string, bool) {
 		if _, addr, port, ok := report.ParseEndpointNodeID(nodeID); ok {
+			if name, ok := meta["name"]; ok {
+				return fmt.Sprintf("%s:%s", name, port), true
+			}
 			return fmt.Sprintf("%s:%s", addr, port), true
 		}
 		if _, addr, ok := report.ParseAddressNodeID(nodeID); ok {
@@ -214,13 +217,13 @@ func connectionDetailsRows(topology report.Topology, originID string) []Row {
 		}
 		return "", false
 	}
-	local, ok := labeler(originID)
+	local, ok := labeler(originID, topology.Nodes[originID].Metadata)
 	if !ok {
 		return rows
 	}
 	// Firstly, collection outgoing connections from this node.
 	for _, serverNodeID := range topology.Nodes[originID].Adjacency {
-		remote, ok := labeler(serverNodeID)
+		remote, ok := labeler(serverNodeID, topology.Nodes[serverNodeID].Metadata)
 		if !ok {
 			continue
 		}
@@ -239,7 +242,7 @@ func connectionDetailsRows(topology report.Topology, originID string) []Row {
 		if !serverNodeIDs.Contains(originID) {
 			continue
 		}
-		remote, ok := labeler(clientNodeID)
+		remote, ok := labeler(clientNodeID, clientNode.Metadata)
 		if !ok {
 			continue
 		}

report/topology.go

@@ -102,6 +102,12 @@ func (n Node) WithMetadata(m map[string]string) Node {
 	return result
 }
 
+// AddMetadata returns a fresh copy of n, with Metadata set to the merge of n and the metadata provided
+func (n Node) AddMetadata(m map[string]string) Node {
+	additional := MakeNodeWith(m)
+	return n.Merge(additional)
+}
+
 // WithCounters returns a fresh copy of n, with Counters set to c
 func (n Node) WithCounters(c map[string]int) Node {
 	result := n.Copy()