itest: make use of session ttl in failover test

lntest/harness_net.go

@@ -275,7 +275,9 @@ func (n *NetworkHarness) Stop() {
 
 // extraArgsEtcd returns extra args for configuring LND to use an external etcd
 // database (for remote channel DB and wallet DB).
-func extraArgsEtcd(etcdCfg *etcd.Config, name string, cluster bool) []string {
+func extraArgsEtcd(etcdCfg *etcd.Config, name string, cluster bool,
+	leaderSessionTTL int) []string {
+
 	extraArgs := []string{
 		"--db.backend=etcd",
 		fmt.Sprintf("--db.etcd.host=%v", etcdCfg.Host),
@@ -289,10 +291,13 @@ func extraArgsEtcd(etcdCfg *etcd.Config, name string, cluster bool) []string {
 	}
 
 	if cluster {
-		extraArgs = append(extraArgs, "--cluster.enable-leader-election")
+		clusterArgs := []string{
-		extraArgs = append(
+			"--cluster.enable-leader-election",
-			extraArgs, fmt.Sprintf("--cluster.id=%v", name),
+			fmt.Sprintf("--cluster.id=%v", name),
-		)
+			fmt.Sprintf("--cluster.leader-session-ttl=%v",
+				leaderSessionTTL),
+		}
+		extraArgs = append(extraArgs, clusterArgs...)
 	}
 
 	return extraArgs
@@ -302,13 +307,13 @@ func extraArgsEtcd(etcdCfg *etcd.Config, name string, cluster bool) []string {
 // etcd database as its (remote) channel and wallet DB. The passsed cluster
 // flag indicates that we'd like the node to join the cluster leader election.
 func (n *NetworkHarness) NewNodeWithSeedEtcd(name string, etcdCfg *etcd.Config,
-	password []byte, entropy []byte, statelessInit, cluster bool) (
+	password []byte, entropy []byte, statelessInit, cluster bool,
-	*HarnessNode, []string, []byte, error) {
+	leaderSessionTTL int) (*HarnessNode, []string, []byte, error) {
 
 	// We don't want to use the embedded etcd instance.
 	const dbBackend = BackendBbolt
 
-	extraArgs := extraArgsEtcd(etcdCfg, name, cluster)
+	extraArgs := extraArgsEtcd(etcdCfg, name, cluster, leaderSessionTTL)
 	return n.newNodeWithSeed(
 		name, extraArgs, password, entropy, statelessInit, dbBackend,
 	)
@@ -320,12 +325,13 @@ func (n *NetworkHarness) NewNodeWithSeedEtcd(name string, etcdCfg *etcd.Config,
 // If the wait flag is false then we won't wait until RPC is available (this is
 // useful when the node is not expected to become the leader right away).
 func (n *NetworkHarness) NewNodeEtcd(name string, etcdCfg *etcd.Config,
-	password []byte, cluster, wait bool) (*HarnessNode, error) {
+	password []byte, cluster, wait bool, leaderSessionTTL int) (
+	*HarnessNode, error) {
 
 	// We don't want to use the embedded etcd instance.
 	const dbBackend = BackendBbolt
 
-	extraArgs := extraArgsEtcd(etcdCfg, name, cluster)
+	extraArgs := extraArgsEtcd(etcdCfg, name, cluster, leaderSessionTTL)
 	return n.newNode(name, extraArgs, true, password, dbBackend, wait)
 }
 

lntest/itest/lnd_etcd_failover_test.go

@@ -81,12 +81,15 @@ func testEtcdFailoverCase(net *lntest.NetworkHarness, ht *harnessTest,
 	}
 	defer cleanup()
 
+	// Make leader election session TTL 5 sec to make the test run fast.
+	const leaderSessionTTL = 5
+
 	observer, err := cluster.MakeLeaderElector(
 		ctxb, cluster.EtcdLeaderElector, "observer",
-		lncfg.DefaultEtcdElectionPrefix, etcdCfg,
+		lncfg.DefaultEtcdElectionPrefix, leaderSessionTTL, etcdCfg,
 	)
 	if err != nil {
-		ht.Fatalf("Cannot start election observer")
+		ht.Fatalf("Cannot start election observer: %v", err)
 	}
 
 	password := []byte("the quick brown fox jumps the lazy dog")
@@ -96,6 +99,7 @@ func testEtcdFailoverCase(net *lntest.NetworkHarness, ht *harnessTest,
 
 	carol1, _, _, err := net.NewNodeWithSeedEtcd(
 		"Carol-1", etcdCfg, password, entropy[:], stateless, cluster,
+		leaderSessionTTL,
 	)
 	if err != nil {
 		ht.Fatalf("unable to start Carol-1: %v", err)
@@ -122,7 +126,7 @@ func testEtcdFailoverCase(net *lntest.NetworkHarness, ht *harnessTest,
 	// At this point Carol-1 is the elected leader, while Carol-2 will wait
 	// to become the leader when Carol-1 stops.
 	carol2, err := net.NewNodeEtcd(
-		"Carol-2", etcdCfg, password, cluster, false,
+		"Carol-2", etcdCfg, password, cluster, false, leaderSessionTTL,
 	)
 	if err != nil {
 		ht.Fatalf("Unable to start Carol-2: %v", err)
@@ -145,17 +149,14 @@ func testEtcdFailoverCase(net *lntest.NetworkHarness, ht *harnessTest,
 	)
 
 	// Shut down or kill Carol-1 and wait for Carol-2 to become the leader.
-	var failoverTimeout time.Duration
+	failoverTimeout := time.Duration(2*leaderSessionTTL) * time.Second
 	if kill {
 		err = net.KillNode(carol1)
 		if err != nil {
 			ht.Fatalf("Can't kill Carol-1: %v", err)
 		}
-
-		failoverTimeout = 2 * time.Minute
 	} else {
 		shutdownAndAssert(net, ht, carol1)
-		failoverTimeout = 30 * time.Second
 	}
 
 	err = carol2.WaitUntilLeader(failoverTimeout)