Fix #339: Reduce Sleeps in tests

License: MIT
Signed-off-by: Hector Sanjuan <code@hector.link>

Makefile

@@ -68,7 +68,7 @@ check:
 	golint -set_exit_status -min_confidence 0.3 ./...
 
 test: deps
-	go test -timeout 20m -loglevel "CRITICAL" -v ./...
+	go test -loglevel "CRITICAL" -v ./...
 
 test_sharness: $(sharness)
 	@sh sharness/run-sharness-tests.sh

ci/Jenkinsfile

@@ -1,2 +1,2 @@
-golang([test: "go test -v -timeout 20m ./..."])
+golang([test: "go test -v -loglevel ERROR ./..."])
 

cluster.go

@@ -272,7 +272,6 @@ func (c *Cluster) pushInformerMetrics() {
 	// The following control how often to make and log
 	// a retry
 	retries := 0
-	retryDelay := 500 * time.Millisecond
 	retryWarnMod := 60
 	for {
 		select {
@@ -293,7 +292,7 @@ func (c *Cluster) pushInformerMetrics() {
 				retries++
 			}
 			// retry in retryDelay
-			timer.Reset(retryDelay)
+			timer.Reset(metric.GetTTL() / 4)
 			continue
 		}
 

cluster_config.go

@@ -105,12 +105,12 @@ type Config struct {
 	// possible.
 	ReplicationFactorMin int
 
-	// MonitorPingInterval is the frequency by which a cluster peer pings
+	// MonitorPingInterval is the frequency with which a cluster peer pings
 	// the monitoring component. The ping metric has a TTL set to the double
 	// of this value.
 	MonitorPingInterval time.Duration
 
-	// PeerWatchInterval is the frequency that we watch for changes
+	// PeerWatchInterval is the frequency that we use to watch for changes
 	// in the consensus peerset and save new peers to the configuration
 	// file. This also affects how soon we realize that we have
 	// been removed from a cluster.

cluster_test.go

@@ -315,7 +315,7 @@ func TestClusterRecoverAllLocal(t *testing.T) {
 		t.Fatal("pin should have worked:", err)
 	}
 
-	time.Sleep(time.Second)
+	pinDelay()
 
 	recov, err := cl.RecoverAllLocal()
 	if err != nil {

config/config.go

@@ -15,7 +15,8 @@ import (
 
 var logger = logging.Logger("config")
 
-// How often to save the configuration file if it needs saving.
+// ConfigSaveInterval specifies how often to save the configuration file if
+// it needs saving.
 var ConfigSaveInterval = time.Second
 
 // The ComponentConfig interface allows components to define configurations

config_test.go

@@ -17,29 +17,30 @@ var testingClusterCfg = []byte(`{
     "secret": "2588b80d5cb05374fa142aed6cbb047d1f4ef8ef15e37eba68c65b9d30df67ed",
     "peers": [],
     "bootstrap": [],
-    "leave_on_shutdown": true,
+    "leave_on_shutdown": false,
     "listen_multiaddress": "/ip4/127.0.0.1/tcp/10000",
     "state_sync_interval": "1m0s",
     "ipfs_sync_interval": "2m10s",
     "replication_factor": -1,
-    "monitor_ping_interval": "1s"
+    "monitor_ping_interval": "150ms",
+    "peer_watch_interval": "100ms"
 }
 `)
 
 var testingRaftCfg = []byte(`{
     "data_folder": "raftFolderFromTests",
-    "wait_for_leader_timeout": "30s",
+    "wait_for_leader_timeout": "10s",
-    "commit_retries": 1,
+    "commit_retries": 2,
-    "commit_retry_delay": "1s",
+    "commit_retry_delay": "50ms",
-    "network_timeout": "20s",
+    "network_timeout": "5s",
-    "heartbeat_timeout": "1s",
+    "heartbeat_timeout": "100ms",
-    "election_timeout": "1s",
+    "election_timeout": "100ms",
     "commit_timeout": "50ms",
-    "max_append_entries": 64,
+    "max_append_entries": 256,
     "trailing_logs": 10240,
     "snapshot_interval": "2m0s",
     "snapshot_threshold": 8192,
-    "leader_lease_timeout": "500ms"
+    "leader_lease_timeout": "80ms"
 }`)
 
 var testingAPICfg = []byte(`{
@@ -71,11 +72,11 @@ var testingTrackerCfg = []byte(`
 `)
 
 var testingMonCfg = []byte(`{
-    "check_interval": "1s"
+    "check_interval": "400ms"
 }`)
 
 var testingDiskInfCfg = []byte(`{
-    "metric_ttl": "1s",
+    "metric_ttl": "150ms",
     "metric_type": "freespace"
 }`)
 

consensus/raft/consensus_test.go

@@ -50,6 +50,7 @@ func makeTestingHost(t *testing.T) host.Host {
 }
 
 func testingConsensus(t *testing.T, idn int) *Consensus {
+	cleanRaft(idn)
 	h := makeTestingHost(t)
 	st := mapstate.NewMapState()
 
@@ -72,6 +73,7 @@ func TestShutdownConsensus(t *testing.T) {
 	// Bring it up twice to make sure shutdown cleans up properly
 	// but also to make sure raft comes up ok when re-initialized
 	cc := testingConsensus(t, 1)
+	defer cleanRaft(1)
 	err := cc.Shutdown()
 	if err != nil {
 		t.Fatal("Consensus cannot shutdown:", err)

coverage.sh

@@ -7,9 +7,9 @@ for dir in $dirs;
 do
         if ls "$dir"/*.go &> /dev/null;
         then
-            cmdflags="-timeout 20m -v -coverprofile=profile.out -covermode=count $dir"
+            cmdflags="-v -coverprofile=profile.out -covermode=count $dir"
             if [ "$dir" == "." ]; then
-                cmdflags="-timeout 20m -v -coverprofile=profile.out -covermode=count -loglevel CRITICAL ."
+                cmdflags="-v -coverprofile=profile.out -covermode=count -loglevel CRITICAL ."
             fi
             echo go test $cmdflags
             go test $cmdflags

ipfscluster_test.go

@@ -35,7 +35,7 @@ var (
 	nClusters = 6
 
 	// number of pins to pin/unpin/check
-	nPins = 500
+	nPins = 100
 
 	logLevel = "CRITICAL"
 
@@ -205,17 +205,29 @@ func createClusters(t *testing.T) ([]*Cluster, []*test.IpfsMock) {
 	for i := 1; i < nClusters; i++ {
 		cfgs[i].Bootstrap = []ma.Multiaddr{bootstrapAddr}
 	}
+	time.Sleep(200 * time.Millisecond)
 
 	// Start the rest
-	var wg sync.WaitGroup
+	// We don't do this in parallel because it causes libp2p dial backoffs
 	for i := 1; i < nClusters; i++ {
-		wg.Add(1)
+		clusters[i] = createCluster(t, cfgs[i], concfgs[i], apis[i], ipfss[i], states[i], trackers[i], mons[i], allocs[i], infs[i])
-		go func(i int) {
+		time.Sleep(200 * time.Millisecond)
-			clusters[i] = createCluster(t, cfgs[i], concfgs[i], apis[i], ipfss[i], states[i], trackers[i], mons[i], allocs[i], infs[i])
+	}
-			wg.Done()
+
-		}(i)
+	// open connections among all peers. This ensures smoother operations.
+	// Best effort. Some errors do happen.
+	for _, c := range clusters {
+		peers, err := c.consensus.Peers()
+		if err != nil {
+			shutdownClusters(t, clusters, ipfsMocks)
+			t.Fatal(err)
+		}
+		for _, p := range peers {
+			if p != c.id {
+				c.host.Network().DialPeer(c.ctx, p)
+			}
+		}
 	}
-	wg.Wait()
 
 	// ---------------------------------------------
 
@@ -223,8 +235,7 @@ func createClusters(t *testing.T) ([]*Cluster, []*test.IpfsMock) {
 	// for i := 1; i < nClusters; i++ {
 	// 	clusters[0].PeerAdd(clusterAddr(clusters[i]))
 	// }
-	delay()
+
-	delay()
 	return clusters, ipfsMocks
 }
 
@@ -255,26 +266,31 @@ func runF(t *testing.T, clusters []*Cluster, f func(*testing.T, *Cluster)) {
 func delay() {
 	var d int
 	if nClusters > 10 {
-		d = 8
+		d = 2000
-
-	} else if nClusters > 5 {
-		d = 5
 	} else {
-		d = nClusters
+		d = 1000
 	}
-	time.Sleep(time.Duration(d) * time.Second)
+	time.Sleep(time.Duration(d) * time.Millisecond)
 }
 
-func waitForLeader(t *testing.T, clusters []*Cluster) {
+func pinDelay() {
-	timer := time.NewTimer(time.Minute)
+	time.Sleep(400 * time.Millisecond)
-	ticker := time.NewTicker(time.Second)
+}
-	// Wait for consensus to pick a new leader in case we shut it down
 
-	// Make sure we don't check on a shutdown cluster
+func ttlDelay() {
-	j := rand.Intn(len(clusters))
+	diskInfCfg := &disk.Config{}
-	for clusters[j].shutdownB {
+	diskInfCfg.LoadJSON(testingDiskInfCfg)
-		j = rand.Intn(len(clusters))
+	time.Sleep(diskInfCfg.MetricTTL * 3)
-	}
+}
+
+// Waits for consensus to pick a new leader in case we shut it down
+// Makes sure all peers know about it.
+// Makes sure new metrics have come in for the new leader.
+func waitForLeader(t *testing.T, clusters []*Cluster) {
+	ttlDelay()
+
+	timer := time.NewTimer(time.Minute)
+	ticker := time.NewTicker(time.Second / 4)
 
 loop:
 	for {
@@ -282,12 +298,20 @@ loop:
 		case <-timer.C:
 			t.Fatal("timed out waiting for a leader")
 		case <-ticker.C:
-			_, err := clusters[j].consensus.Leader()
+			for _, cl := range clusters {
-			if err == nil {
+				if cl.shutdownB {
-				break loop
+					continue // skip shutdown clusters
+				}
+				_, err := cl.consensus.Leader()
+				if err != nil {
+					continue loop
+				}
 			}
+			break loop
 		}
 	}
+
+	ttlDelay()
 }
 
 func TestClustersVersion(t *testing.T) {
@@ -305,7 +329,6 @@ func TestClustersVersion(t *testing.T) {
 func TestClustersPeers(t *testing.T) {
 	clusters, mock := createClusters(t)
 	defer shutdownClusters(t, clusters, mock)
-	delay()
 
 	j := rand.Intn(nClusters) // choose a random cluster peer
 	peers := clusters[j].Peers()
@@ -345,6 +368,9 @@ func TestClustersPin(t *testing.T) {
 	defer shutdownClusters(t, clusters, mock)
 	exampleCid, _ := cid.Decode(test.TestCid1)
 	prefix := exampleCid.Prefix()
+
+	ttlDelay()
+
 	for i := 0; i < nPins; i++ {
 		j := rand.Intn(nClusters)           // choose a random cluster peer
 		h, err := prefix.Sum(randomBytes()) // create random cid
@@ -360,6 +386,7 @@ func TestClustersPin(t *testing.T) {
 		}
 	}
 	delay()
+	delay()
 	fpinned := func(t *testing.T, c *Cluster) {
 		status := c.tracker.StatusAll()
 		for _, v := range status {
@@ -378,7 +405,7 @@ func TestClustersPin(t *testing.T) {
 	// Unpin everything
 	pinList := clusters[0].Pins()
 
-	for i := 0; i < nPins; i++ {
+	for i := 0; i < len(pinList); i++ {
 		j := rand.Intn(nClusters) // choose a random cluster peer
 		err := clusters[j].Unpin(pinList[i].Cid)
 		if err != nil {
@@ -392,6 +419,7 @@ func TestClustersPin(t *testing.T) {
 
 	}
 	delay()
+	delay()
 
 	funpinned := func(t *testing.T, c *Cluster) {
 		status := c.tracker.StatusAll()
@@ -408,7 +436,7 @@ func TestClustersStatusAll(t *testing.T) {
 	defer shutdownClusters(t, clusters, mock)
 	h, _ := cid.Decode(test.TestCid1)
 	clusters[0].Pin(api.PinCid(h))
-	delay()
+	pinDelay()
 	// Global status
 	f := func(t *testing.T, c *Cluster) {
 		statuses, err := c.StatusAll()
@@ -452,10 +480,11 @@ func TestClustersStatusAllWithErrors(t *testing.T) {
 	defer shutdownClusters(t, clusters, mock)
 	h, _ := cid.Decode(test.TestCid1)
 	clusters[0].Pin(api.PinCid(h))
-	delay()
+	pinDelay()
 
 	// shutdown 1 cluster peer
 	clusters[1].Shutdown()
+	delay()
 
 	f := func(t *testing.T, c *Cluster) {
 		// skip if it's the shutdown peer
@@ -513,7 +542,9 @@ func TestClustersSyncAllLocal(t *testing.T) {
 	h2, _ := cid.Decode(test.TestCid2)
 	clusters[0].Pin(api.PinCid(h))
 	clusters[0].Pin(api.PinCid(h2))
-	delay()
+	pinDelay()
+	pinDelay()
+
 	f := func(t *testing.T, c *Cluster) {
 		// Sync bad ID
 		infos, err := c.SyncAllLocal()
@@ -541,7 +572,8 @@ func TestClustersSyncLocal(t *testing.T) {
 	h2, _ := cid.Decode(test.TestCid2)
 	clusters[0].Pin(api.PinCid(h))
 	clusters[0].Pin(api.PinCid(h2))
-	delay()
+	pinDelay()
+	pinDelay()
 
 	f := func(t *testing.T, c *Cluster) {
 		info, err := c.SyncLocal(h)
@@ -572,7 +604,8 @@ func TestClustersSyncAll(t *testing.T) {
 	h2, _ := cid.Decode(test.TestCid2)
 	clusters[0].Pin(api.PinCid(h))
 	clusters[0].Pin(api.PinCid(h2))
-	delay()
+	pinDelay()
+	pinDelay()
 
 	j := rand.Intn(nClusters) // choose a random cluster peer
 	ginfos, err := clusters[j].SyncAll()
@@ -603,7 +636,8 @@ func TestClustersSync(t *testing.T) {
 	h2, _ := cid.Decode(test.TestCid2)
 	clusters[0].Pin(api.PinCid(h))
 	clusters[0].Pin(api.PinCid(h2))
-	delay()
+	pinDelay()
+	pinDelay()
 
 	j := rand.Intn(nClusters)
 	ginfo, err := clusters[j].Sync(h)
@@ -662,10 +696,13 @@ func TestClustersRecoverLocal(t *testing.T) {
 	defer shutdownClusters(t, clusters, mock)
 	h, _ := cid.Decode(test.ErrorCid) // This cid always fails
 	h2, _ := cid.Decode(test.TestCid2)
+
+	ttlDelay()
+
 	clusters[0].Pin(api.PinCid(h))
 	clusters[0].Pin(api.PinCid(h2))
-
+	pinDelay()
-	delay()
+	pinDelay()
 
 	f := func(t *testing.T, c *Cluster) {
 		info, err := c.RecoverLocal(h)
@@ -694,10 +731,14 @@ func TestClustersRecover(t *testing.T) {
 	defer shutdownClusters(t, clusters, mock)
 	h, _ := cid.Decode(test.ErrorCid) // This cid always fails
 	h2, _ := cid.Decode(test.TestCid2)
+
+	ttlDelay()
+
 	clusters[0].Pin(api.PinCid(h))
 	clusters[0].Pin(api.PinCid(h2))
 
-	delay()
+	pinDelay()
+	pinDelay()
 
 	j := rand.Intn(nClusters)
 	ginfo, err := clusters[j].Recover(h)
@@ -771,6 +812,8 @@ func TestClustersReplication(t *testing.T) {
 		c.config.ReplicationFactorMax = nClusters - 1
 	}
 
+	ttlDelay()
+
 	// Why is replication factor nClusters - 1?
 	// Because that way we know that pinning nCluster
 	// pins with an strategy like numpins/disk
@@ -789,7 +832,7 @@ func TestClustersReplication(t *testing.T) {
 		if err != nil {
 			t.Error(err)
 		}
-		time.Sleep(time.Second)
+		pinDelay()
 
 		// check that it is held by exactly nClusters -1 peers
 		gpi, err := clusters[j].Status(h)
@@ -814,7 +857,7 @@ func TestClustersReplication(t *testing.T) {
 		if numRemote != 1 {
 			t.Errorf("We wanted 1 peer track as remote but %d do", numRemote)
 		}
-		time.Sleep(time.Second) // this is for metric to be up to date
+		ttlDelay()
 	}
 
 	f := func(t *testing.T, c *Cluster) {
@@ -875,13 +918,15 @@ func TestClustersReplicationFactorMax(t *testing.T) {
 		c.config.ReplicationFactorMax = nClusters - 1
 	}
 
+	ttlDelay()
+
 	h, _ := cid.Decode(test.TestCid1)
 	err := clusters[0].Pin(api.PinCid(h))
 	if err != nil {
 		t.Fatal(err)
 	}
 
-	delay()
+	pinDelay()
 
 	f := func(t *testing.T, c *Cluster) {
 		p, err := c.PinGet(h)
@@ -918,13 +963,15 @@ func TestClustersReplicationFactorMaxLower(t *testing.T) {
 		c.config.ReplicationFactorMax = nClusters
 	}
 
+	ttlDelay() // make sure we have places to pin
+
 	h, _ := cid.Decode(test.TestCid1)
 	err := clusters[0].Pin(api.PinCid(h))
 	if err != nil {
 		t.Fatal(err)
 	}
 
-	delay()
+	pinDelay()
 
 	p1, err := clusters[0].PinGet(h)
 	if err != nil {
@@ -944,7 +991,7 @@ func TestClustersReplicationFactorMaxLower(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	delay()
+	pinDelay()
 
 	p2, err := clusters[0].PinGet(h)
 	if err != nil {
@@ -970,24 +1017,21 @@ func TestClustersReplicationFactorInBetween(t *testing.T) {
 		c.config.ReplicationFactorMax = nClusters
 	}
 
+	ttlDelay()
+
 	// Shutdown two peers
 	clusters[nClusters-1].Shutdown()
 	clusters[nClusters-2].Shutdown()
 
-	time.Sleep(time.Second) // let metric expire
-
 	waitForLeader(t, clusters)
 
-	// allow metrics to arrive to new leader
-	delay()
-
 	h, _ := cid.Decode(test.TestCid1)
 	err := clusters[0].Pin(api.PinCid(h))
 	if err != nil {
 		t.Fatal(err)
 	}
 
-	delay()
+	pinDelay()
 
 	f := func(t *testing.T, c *Cluster) {
 		if c == clusters[nClusters-1] || c == clusters[nClusters-2] {
@@ -1029,14 +1073,9 @@ func TestClustersReplicationFactorMin(t *testing.T) {
 
 	// Shutdown two peers
 	clusters[nClusters-1].Shutdown()
-	clusters[nClusters-2].Shutdown()
-
-	time.Sleep(time.Second) // let metric expire
-
 	waitForLeader(t, clusters)
-
+	clusters[nClusters-2].Shutdown()
-	// allow metrics to arrive to new leader
+	waitForLeader(t, clusters)
-	delay()
 
 	h, _ := cid.Decode(test.TestCid1)
 	err := clusters[0].Pin(api.PinCid(h))
@@ -1063,28 +1102,29 @@ func TestClustersReplicationMinMaxNoRealloc(t *testing.T) {
 		c.config.ReplicationFactorMax = nClusters
 	}
 
+	ttlDelay()
+
 	h, _ := cid.Decode(test.TestCid1)
 	err := clusters[0].Pin(api.PinCid(h))
 	if err != nil {
 		t.Fatal(err)
 	}
 
+	pinDelay()
+
 	// Shutdown two peers
 	clusters[nClusters-1].Shutdown()
-	clusters[nClusters-2].Shutdown()
-
-	time.Sleep(time.Second) // let metric expire
-
 	waitForLeader(t, clusters)
-
+	clusters[nClusters-2].Shutdown()
-	// allow metrics to arrive to new leader
+	waitForLeader(t, clusters)
-	delay()
 
 	err = clusters[0].Pin(api.PinCid(h))
 	if err != nil {
 		t.Fatal(err)
 	}
 
+	pinDelay()
+
 	p, err := clusters[0].PinGet(h)
 	if err != nil {
 		t.Fatal(err)
@@ -1114,13 +1154,15 @@ func TestClustersReplicationMinMaxRealloc(t *testing.T) {
 		c.config.ReplicationFactorMax = 4
 	}
 
+	ttlDelay() // make sure metrics are in
+
 	h, _ := cid.Decode(test.TestCid1)
 	err := clusters[0].Pin(api.PinCid(h))
 	if err != nil {
 		t.Fatal(err)
 	}
 
-	delay()
+	pinDelay()
 
 	p, err := clusters[0].PinGet(h)
 	if err != nil {
@@ -1142,19 +1184,16 @@ func TestClustersReplicationMinMaxRealloc(t *testing.T) {
 	alloc1.Shutdown()
 	alloc2.Shutdown()
 
-	time.Sleep(time.Second) // let metric expire
-
 	waitForLeader(t, clusters)
 
-	// allow metrics to arrive to new leader
-	delay()
-
 	// Repin - (although this might have been taken of if there was an alert
 	err = safePeer.Pin(api.PinCid(h))
 	if err != nil {
 		t.Fatal(err)
 	}
 
+	pinDelay()
+
 	p, err = safePeer.PinGet(h)
 	if err != nil {
 		t.Fatal(err)
@@ -1176,7 +1215,7 @@ func TestClustersReplicationMinMaxRealloc(t *testing.T) {
 	lenSA := len(secondAllocations)
 	expected := minInt(nClusters-2, 4)
 	if lenSA != expected {
-		t.Errorf("Inssufficient reallocation, could have allocated to %d peers but instead only allocated to %d peers", expected, lenSA)
+		t.Errorf("Insufficient reallocation, could have allocated to %d peers but instead only allocated to %d peers", expected, lenSA)
 	}
 
 	if lenSA < 3 {
@@ -1194,6 +1233,8 @@ func TestClustersReplicationRealloc(t *testing.T) {
 		c.config.ReplicationFactorMax = nClusters - 1
 	}
 
+	ttlDelay()
+
 	j := rand.Intn(nClusters)
 	h, _ := cid.Decode(test.TestCid1)
 	err := clusters[j].Pin(api.PinCid(h))
@@ -1202,7 +1243,7 @@ func TestClustersReplicationRealloc(t *testing.T) {
 	}
 
 	// Let the pin arrive
-	time.Sleep(time.Second / 2)
+	pinDelay()
 
 	pin := clusters[j].Pins()[0]
 	pinSerial := pin.ToSerial()
@@ -1217,7 +1258,7 @@ func TestClustersReplicationRealloc(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	time.Sleep(time.Second / 2)
+	pinDelay()
 
 	pin2 := clusters[j].Pins()[0]
 	pinSerial2 := pin2.ToSerial()
@@ -1245,10 +1286,7 @@ func TestClustersReplicationRealloc(t *testing.T) {
 
 	// let metrics expire and give time for the cluster to
 	// see if they have lost the leader
-	time.Sleep(4 * time.Second)
 	waitForLeader(t, clusters)
-	// wait for new metrics to arrive
-	time.Sleep(2 * time.Second)
 
 	// Make sure we haven't killed our randomly
 	// selected cluster
@@ -1262,7 +1300,7 @@ func TestClustersReplicationRealloc(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	time.Sleep(time.Second / 2)
+	pinDelay()
 
 	numPinned := 0
 	for i, c := range clusters {
@@ -1303,12 +1341,11 @@ func TestClustersReplicationNotEnoughPeers(t *testing.T) {
 	}
 
 	// Let the pin arrive
-	time.Sleep(time.Second / 2)
+	pinDelay()
 
 	clusters[0].Shutdown()
 	clusters[1].Shutdown()
 
-	delay()
 	waitForLeader(t, clusters)
 
 	err = clusters[2].Pin(api.PinCid(h))
@@ -1337,7 +1374,7 @@ func TestClustersRebalanceOnPeerDown(t *testing.T) {
 	// pin something
 	h, _ := cid.Decode(test.TestCid1)
 	clusters[0].Pin(api.PinCid(h))
-	time.Sleep(time.Second * 2) // let the pin arrive
+	pinDelay()
 	pinLocal := 0
 	pinRemote := 0
 	var localPinner peer.ID
@@ -1361,7 +1398,7 @@ func TestClustersRebalanceOnPeerDown(t *testing.T) {
 		t.Fatal("Not pinned as expected")
 	}
 
-	// find a kill the local pinner
+	// kill the local pinner
 	for _, c := range clusters {
 		if c.id == localPinner {
 			c.Shutdown()
@@ -1370,8 +1407,8 @@ func TestClustersRebalanceOnPeerDown(t *testing.T) {
 		}
 	}
 
-	// Sleep a monitoring interval
+	delay()
-	time.Sleep(6 * time.Second)
+	waitForLeader(t, clusters) // in case we killed the leader
 
 	// It should be now pinned in the remote pinner
 	if s := remotePinnerCluster.tracker.Status(h).Status; s != api.TrackerStatusPinned {
@@ -1452,8 +1489,6 @@ func validateClusterGraph(t *testing.T, graph api.ConnectGraph, clusterIDs map[p
 func TestClustersGraphConnected(t *testing.T) {
 	clusters, mock := createClusters(t)
 	defer shutdownClusters(t, clusters, mock)
-	delay()
-	delay()
 
 	j := rand.Intn(nClusters) // choose a random cluster peer to query
 	graph, err := clusters[j].ConnectGraph()
@@ -1496,9 +1531,8 @@ func TestClustersGraphUnhealthy(t *testing.T) {
 
 	clusters[discon1].Shutdown()
 	clusters[discon2].Shutdown()
-	delay()
+
 	waitForLeader(t, clusters)
-	delay()
 
 	graph, err := clusters[j].ConnectGraph()
 	if err != nil {

monitor/basic/peer_monitor.go

@@ -22,7 +22,7 @@ var logger = logging.Logger("monitor")
 var AlertChannelCap = 256
 
 // WindowCap specifies how many metrics to keep for given host and metric type
-var WindowCap = 10
+var WindowCap = 100
 
 // peerMetrics is just a circular queue
 type peerMetrics struct {
@@ -55,6 +55,7 @@ func (pmets *peerMetrics) latest() (api.Metric, error) {
 	//	pmets.mux.RLock()
 	//	defer pmets.mux.RUnlock()
 	if len(pmets.window) == 0 {
+		logger.Warning("no metrics")
 		return api.Metric{}, errors.New("no metrics")
 	}
 	return pmets.window[pmets.last], nil

peer_manager_test.go

@@ -29,7 +29,6 @@ func peerManagerClusters(t *testing.T) ([]*Cluster, []*test.IpfsMock) {
 		}(i)
 	}
 	wg.Wait()
-	delay()
 	return cls, mocks
 }
 
@@ -65,7 +64,7 @@ func TestClustersPeerAdd(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	delay()
+	pinDelay()
 
 	f := func(t *testing.T, c *Cluster) {
 		ids := c.Peers()
@@ -88,8 +87,6 @@ func TestClustersPeerAdd(t *testing.T) {
 			t.Error("By now cluster peers should reflect all peers")
 		}
 
-		time.Sleep(2 * time.Second)
-
 		// check that they are part of the configuration
 		// This only works because each peer only has one multiaddress
 		// (localhost)
@@ -214,6 +211,7 @@ func TestClustersPeerRemoveSelf(t *testing.T) {
 	defer shutdownClusters(t, clusters, mocks)
 
 	for i := 0; i < len(clusters); i++ {
+		waitForLeader(t, clusters)
 		peers := clusters[i].Peers()
 		t.Logf("Current cluster size: %d", len(peers))
 		if len(peers) != (len(clusters) - i) {
@@ -286,7 +284,7 @@ func TestClustersPeerRemoveLeader(t *testing.T) {
 		if more {
 			t.Error("should be done")
 		}
-		time.Sleep(time.Second)
+		time.Sleep(time.Second / 2)
 	}
 }
 
@@ -341,10 +339,10 @@ func TestClustersPeerRemoveReallocsPins(t *testing.T) {
 		checkErr(t, err)
 		err = leader.Pin(api.PinCid(h))
 		checkErr(t, err)
-		time.Sleep(time.Second) // time to update the metrics
+		ttlDelay()
 	}
 
-	delay()
+	pinDelay()
 
 	// At this point, all peers must have 1 pin associated to them.
 	// Find out which pin is associated to leader.
@@ -373,9 +371,7 @@ func TestClustersPeerRemoveReallocsPins(t *testing.T) {
 		t.Fatal("error removing peer:", err)
 	}
 
-	time.Sleep(2 * time.Second)
 	waitForLeader(t, clusters)
-	delay()
 
 	for _, icid := range interestingCids {
 		// Now check that the allocations are new.
@@ -405,7 +401,7 @@ func TestClustersPeerJoin(t *testing.T) {
 	}
 	hash, _ := cid.Decode(test.TestCid1)
 	clusters[0].Pin(api.PinCid(hash))
-	delay()
+	pinDelay()
 
 	f := func(t *testing.T, c *Cluster) {
 		peers := c.Peers()
@@ -438,7 +434,7 @@ func TestClustersPeerJoinAllAtOnce(t *testing.T) {
 
 	hash, _ := cid.Decode(test.TestCid1)
 	clusters[0].Pin(api.PinCid(hash))
-	delay()
+	pinDelay()
 
 	f2 := func(t *testing.T, c *Cluster) {
 		peers := c.Peers()
@@ -555,7 +551,7 @@ func TestClustersPeerRejoin(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	delay()
+	pinDelay()
 
 	// Rejoin c0
 	c0, m0 := createOnePeerCluster(t, 0, testingClusterSecret)

pnet_test.go

@@ -1,10 +1,6 @@
 package ipfscluster
 
-import (
+import "testing"
-	"testing"
-
-	pnet "github.com/libp2p/go-libp2p-pnet"
-)
 
 func TestClusterSecretFormat(t *testing.T) {
 	goodSecret := "0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef"
@@ -57,28 +53,29 @@ func TestSimplePNet(t *testing.T) {
 	}
 }
 
-func TestClusterSecretRequired(t *testing.T) {
+// // Adds one minute to tests. Disabled for the moment.
-	cl1Secret, err := pnet.GenerateV1Bytes()
+// func TestClusterSecretRequired(t *testing.T) {
-	if err != nil {
+// 	cl1Secret, err := pnet.GenerateV1Bytes()
-		t.Fatal("Unable to generate cluster secret.")
+// 	if err != nil {
-	}
+// 		t.Fatal("Unable to generate cluster secret.")
-	cl1, _ := createOnePeerCluster(t, 1, (*cl1Secret)[:])
+// 	}
-	cl2, _ := createOnePeerCluster(t, 2, testingClusterSecret)
+// 	cl1, _ := createOnePeerCluster(t, 1, (*cl1Secret)[:])
-	defer cleanRaft()
+// 	cl2, _ := createOnePeerCluster(t, 2, testingClusterSecret)
-	defer cl1.Shutdown()
+// 	defer cleanRaft()
-	defer cl2.Shutdown()
+// 	defer cl1.Shutdown()
-	peers1 := cl1.Peers()
+// 	defer cl2.Shutdown()
-	peers2 := cl2.Peers()
+// 	peers1 := cl1.Peers()
+// 	peers2 := cl2.Peers()
+//
+// 	_, err = cl1.PeerAdd(clusterAddr(cl2))
+// 	if err == nil {
+// 		t.Fatal("Peer entered private cluster without key.")
+// 	}
 
-	_, err = cl1.PeerAdd(clusterAddr(cl2))
+// 	if len(peers1) != len(peers2) {
-	if err == nil {
+// 		t.Fatal("Expected same number of peers")
-		t.Fatal("Peer entered private cluster without key.")
+// 	}
-	}
+// 	if len(peers1) != 1 {
-
+// 		t.Fatal("Expected no peers other than self")
-	if len(peers1) != len(peers2) {
+// 	}
-		t.Fatal("Expected same number of peers")
+// }
-	}
-	if len(peers1) != 1 {
-		t.Fatal("Expected no peers other than self")
-	}
-}