[management] use peer.Status.LastSeen field to detect stale peer.Status.Connected values (#7268)

management/internals/network_map_db/pgsql/peer.go

@@ -11,7 +11,7 @@ import (
 const (
 	GetPeersQuery = `
 	select id, key, ssh_key, dns_label, extra_dns_labels, user_id, ssh_enabled, login_expiration_enabled, last_login, ip, ipv6,
-	peer_status_requires_approval, peer_status_connected, proxy_meta_embedded, proxy_meta_cluster,
+	peer_status_requires_approval, peer_status_connected, peer_status_last_seen, proxy_meta_embedded, proxy_meta_cluster,
 	meta_wt_version, meta_go_os, meta_os_version, meta_kernel_version, meta_network_addresses, meta_files, meta_certificates, meta_capabilities, meta_flags, meta_sync_message_version,
 	location_country_code, location_city_name, location_connection_ip
 	from peers

management/internals/network_map_db/shared_types.go

@@ -11,6 +11,7 @@ import (
 
 	"github.com/miekg/dns"
 	"github.com/netbirdio/netbird/management/server/integrations/integrated_validator"
+	nbpeer "github.com/netbirdio/netbird/management/server/peer"
 	"github.com/netbirdio/netbird/management/server/settings"
 	"github.com/netbirdio/netbird/shared/management/networkmap"
 	"github.com/netbirdio/netbird/shared/management/networkmap/nmdata"
@@ -186,6 +187,7 @@ type Peer struct {
 	IPv6                       []byte         `nmap:"json"`
 	PeerStatusRequiresApproval sql.NullBool   `nmap:"map_to:RequiresApproval"`
 	PeerStatusConnected        sql.NullBool   `nmap:"skip"`
+	PeerStatusLastSeen         sql.NullTime   `nmap:"skip"`
 	ProxyMetaEmbedded          sql.NullBool   `nmap:"skip"`
 	ProxyMetaCluster           sql.NullString `nmap:"skip"`
 	MetaWtVersion              sql.NullString `nmap:"skip"`
@@ -318,8 +320,8 @@ func ConvertToNmdataPeers(peers []Peer) ([]nmdata.Peer, map[string][]*nmdata.Pee
 		}
 		dp.ProxyMeta.Cluster = p.ProxyMetaCluster.String
 		// This is only used to build private service candidates, not connected peers are skipped
-		dp.Connected = p.PeerStatusConnected.Bool
-		if dp.ProxyMeta.Embedded && p.PeerStatusConnected.Bool {
+		dp.Connected = nbpeer.IsConnected(p.PeerStatusConnected.Bool, p.PeerStatusLastSeen.Time)
+		if dp.ProxyMeta.Embedded && dp.Connected {
 			clusterToPeerIdx[p.ProxyMetaCluster.String] = append(clusterToPeerIdx[p.ProxyMetaCluster.String], &dp)
 		}
 		if p.MetaWtVersion.Valid {

management/internals/network_map_db/shared_types_test.go

@@ -1,10 +1,14 @@
 package networkmapdb
 
 import (
+	"database/sql"
 	"testing"
+	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+
+	nbpeer "github.com/netbirdio/netbird/management/server/peer"
 )
 
 func TestRecordTypeAndRdata(t *testing.T) {
@@ -58,3 +62,31 @@ func TestConvertToNmdataPeers_RejectsMalformedCertificates(t *testing.T) {
 
 	assert.Error(t, err, "a stored value that is not a list of chains must not load as no certificates")
 }
+
+// TestConvertToNmdataPeers_StaleEmbeddedProxyIsNotACandidate: a management instance that
+// stops without marking its peers disconnected leaves the connected flag set. An embedded
+// proxy peer in that state must not become a private service DNS target, or clients
+// resolve the service to a tunnel IP that no longer answers.
+func TestConvertToNmdataPeers_StaleEmbeddedProxyIsNotACandidate(t *testing.T) {
+	embedded := func(id string, lastSeen time.Time) Peer {
+		return Peer{
+			ID:                  id,
+			PeerStatusConnected: sql.NullBool{Bool: true, Valid: true},
+			PeerStatusLastSeen:  sql.NullTime{Time: lastSeen, Valid: true},
+			ProxyMetaEmbedded:   sql.NullBool{Bool: true, Valid: true},
+			ProxyMetaCluster:    sql.NullString{String: "cluster", Valid: true},
+		}
+	}
+
+	peers, clusterToPeers, err := ConvertToNmdataPeers([]Peer{
+		embedded("live", time.Now().Add(-time.Minute)),
+		embedded("stale", time.Now().Add(-nbpeer.StaleAfter-time.Minute)),
+	})
+
+	require.NoError(t, err)
+	require.Len(t, peers, 2)
+	assert.True(t, peers[0].Connected)
+	assert.False(t, peers[1].Connected, "a peer not seen for longer than StaleAfter must read as disconnected")
+	require.Len(t, clusterToPeers["cluster"], 1)
+	assert.Equal(t, "live", clusterToPeers["cluster"][0].ID)
+}

management/internals/network_map_db/sqlite/peer.go

@@ -10,7 +10,7 @@ import (
 const (
 	GetPeersQuery = `
 	select id, key, ssh_key, dns_label, extra_dns_labels, user_id, ssh_enabled, login_expiration_enabled, last_login, ip, ipv6,
-	peer_status_requires_approval, peer_status_connected, proxy_meta_embedded, proxy_meta_cluster,
+	peer_status_requires_approval, peer_status_connected, peer_status_last_seen, proxy_meta_embedded, proxy_meta_cluster,
 	meta_wt_version, meta_go_os, meta_os_version, meta_kernel_version, meta_network_addresses, meta_files, meta_certificates, meta_capabilities, meta_flags, meta_sync_message_version,
 	location_country_code, location_city_name, location_connection_ip
 	from peers