// TestDiscoverySecondPeerUp ensures that a second peer joining a discovery
// cluster works.
func TestDiscoverySecondPeerUp(t *testing.T) {
etcdtest.RunServer(func(s *server.Server) {
v := url.Values{}
v.Set("value", "started")
resp, err := etcdtest.PutForm(fmt.Sprintf("%s%s", s.URL(), "/v2/keys/_etcd/registry/3/_state"), v)
assert.Equal(t, resp.StatusCode, http.StatusCreated)
u, ok := s.PeerURL("ETCDTEST")
if !ok {
t.Fatalf("Couldn't find the URL")
}
wc := goetcd.NewClient([]string{s.URL()})
testResp, err := wc.Set("test", "0", 0)
if err != nil {
t.Fatalf("Couldn't set a test key on the leader %v", err)
}
v = url.Values{}
v.Set("value", u)
resp, err = etcdtest.PutForm(fmt.Sprintf("%s%s", s.URL(), "/v2/keys/_etcd/registry/3/ETCDTEST"), v)
assert.Equal(t, resp.StatusCode, http.StatusCreated)
proc, err := startServer([]string{"-discovery", s.URL() + "/v2/keys/_etcd/registry/3"})
if err != nil {
t.Fatal(err.Error())
}
defer stopServer(proc)
watch := fmt.Sprintf("%s%s%d", s.URL(), "/v2/keys/_etcd/registry/3/node1?wait=true&waitIndex=", testResp.EtcdIndex)
resp, err = http.Get(watch)
if err != nil {
t.Fatal(err.Error())
}
// TODO(bp): need to have a better way of knowing a machine is up
for i := 0; i < 10; i++ {
time.Sleep(1 * time.Second)
etcdc := goetcd.NewClient(nil)
_, err = etcdc.Set("foobar", "baz", 0)
if err == nil {
break
}
}
if err != nil {
t.Fatal(err.Error())
}
})
}
// Sending set commands
func Set(stop chan bool) {
stopSet := false
i := 0
c := etcd.NewClient(nil)
for {
key := fmt.Sprintf("%s_%v", "foo", i)
result, err := c.Set(key, "bar", 0)
if err != nil || result.Node.Key != "/"+key || result.Node.Value != "bar" {
select {
case <-stop:
stopSet = true
default:
}
}
select {
case <-stop:
stopSet = true
default:
}
if stopSet {
break
}
i++
}
stop <- true
}
// Create a five nodes
// Kill all the nodes and restart
func TestMultiNodeKillAllAndRecovery(t *testing.T) {
procAttr := new(os.ProcAttr)
procAttr.Files = []*os.File{nil, os.Stdout, os.Stderr}
clusterSize := 5
argGroup, etcds, err := CreateCluster(clusterSize, procAttr, false)
defer DestroyCluster(etcds)
if err != nil {
t.Fatal("cannot create cluster")
}
c := etcd.NewClient(nil)
c.SyncCluster()
time.Sleep(time.Second)
// send 10 commands
for i := 0; i < 10; i++ {
// Test Set
_, err := c.Set("foo", "bar", 0)
if err != nil {
panic(err)
}
}
time.Sleep(time.Second)
// kill all
DestroyCluster(etcds)
time.Sleep(time.Second)
stop := make(chan bool)
leaderChan := make(chan string, 1)
all := make(chan bool, 1)
time.Sleep(time.Second)
for i := 0; i < clusterSize; i++ {
etcds[i], err = os.StartProcess(EtcdBinPath, argGroup[i], procAttr)
}
go Monitor(clusterSize, 1, leaderChan, all, stop)
<-all
<-leaderChan
result, err := c.Set("foo", "bar", 0)
if err != nil {
t.Fatalf("Recovery error: %s", err)
}
if result.Node.ModifiedIndex != 16 {
t.Fatalf("recovery failed! [%d/16]", result.Node.ModifiedIndex)
}
}
func (d *Discoverer) Do(discoveryURL string, name string, peer string) (peers []string, err error) {
d.name = name
d.peer = peer
d.discoveryURL = discoveryURL
u, err := url.Parse(discoveryURL)
if err != nil {
return
}
// prefix is prepended to all keys for this discovery
d.prefix = strings.TrimPrefix(u.Path, "/v2/keys/")
// keep the old path in case we need to set the KeyPrefix below
oldPath := u.Path
u.Path = ""
// Connect to a scheme://host not a full URL with path
log.Infof("Discovery via %s using prefix %s.", u.String(), d.prefix)
d.client = etcd.NewClient([]string{u.String()})
if !strings.HasPrefix(oldPath, "/v2/keys") {
d.client.SetKeyPrefix("")
}
// Register this machine first and announce that we are a member of
// this cluster
err = d.heartbeat()
if err != nil {
return
}
// Start the very slow heartbeat to the cluster now in anticipation
// that everything is going to go alright now
go d.startHeartbeat()
// Attempt to take the leadership role, if there is no error we are it!
resp, err := d.client.Create(path.Join(d.prefix, stateKey), startedState, 0)
// Bail out on unexpected errors
if err != nil {
if clientErr, ok := err.(*etcd.EtcdError); !ok || clientErr.ErrorCode != etcdErr.EcodeNodeExist {
return nil, err
}
}
// If we got a response then the CAS was successful, we are leader
if resp != nil && resp.Node.Value == startedState {
// We are the leader, we have no peers
log.Infof("Discovery _state was empty, so this machine is the initial leader.")
return nil, nil
}
// Fall through to finding the other discovery peers
return d.findPeers()
}
// This test creates a single node and then set a value to it.
// Then this test kills the node and restart it and tries to get the value again.
func TestSingleNodeRecovery(t *testing.T) {
procAttr := new(os.ProcAttr)
procAttr.Files = []*os.File{nil, os.Stdout, os.Stderr}
args := []string{"etcd", "-name=node1", "-data-dir=/tmp/node1"}
process, err := os.StartProcess(EtcdBinPath, append(args, "-f"), procAttr)
if err != nil {
t.Fatal("start process failed:" + err.Error())
return
}
time.Sleep(time.Second)
c := etcd.NewClient(nil)
c.SyncCluster()
// Test Set
result, err := c.Set("foo", "bar", 100)
node := result.Node
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL < 95 {
if err != nil {
t.Fatal(err)
}
t.Fatalf("Set 1 failed with %s %s %v", node.Key, node.Value, node.TTL)
}
time.Sleep(time.Second)
process.Kill()
process, err = os.StartProcess(EtcdBinPath, args, procAttr)
defer process.Kill()
if err != nil {
t.Fatal("start process failed:" + err.Error())
return
}
time.Sleep(time.Second)
result, err = c.Get("foo", false, false)
node = result.Node
if err != nil {
t.Fatal("get fail: " + err.Error())
return
}
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL > 99 {
if err != nil {
t.Fatal(err)
}
t.Fatalf("Recovery Get failed with %s %s %v", node.Key, node.Value, node.TTL)
}
}
func write(endpoint string, requests int, end chan int) {
client := etcd.NewClient([]string{endpoint})
for i := 0; i < requests; i++ {
key := strconv.Itoa(i)
_, err := client.Set(key, key, 0)
if err != nil {
println(err.Error())
}
}
end <- 1
}
// NewHandler creates an HTTP handler that can be registered on a router.
func NewHandler(addr string) http.Handler {
h := &handler{
Router: mux.NewRouter(),
client: etcd.NewClient([]string{addr}),
}
h.StrictSlash(false)
h.handleFunc("/{key:.*}", h.getIndexHandler).Methods("GET")
h.handleFunc("/{key:.*}", h.acquireHandler).Methods("POST")
h.handleFunc("/{key:.*}", h.renewLockHandler).Methods("PUT")
h.handleFunc("/{key:.*}", h.releaseLockHandler).Methods("DELETE")
return h
}
func watch(endpoint string, key string) {
client := etcd.NewClient([]string{endpoint})
receiver := make(chan *etcd.Response)
go client.Watch(key, 0, true, receiver, nil)
log.Printf("watching: %s", key)
received := 0
for {
<-receiver
received++
}
}
// Create a three nodes and try to set value
func templateTestSimpleMultiNode(t *testing.T, tls bool) {
procAttr := new(os.ProcAttr)
procAttr.Files = []*os.File{nil, os.Stdout, os.Stderr}
clusterSize := 3
_, etcds, err := CreateCluster(clusterSize, procAttr, tls)
if err != nil {
t.Fatal("cannot create cluster")
}
defer DestroyCluster(etcds)
time.Sleep(time.Second)
c := etcd.NewClient(nil)
c.SyncCluster()
// Test Set
result, err := c.Set("foo", "bar", 100)
node := result.Node
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL < 95 {
if err != nil {
t.Fatal(err)
}
t.Fatalf("Set 1 failed with %s %s %v", node.Key, node.Value, node.TTL)
}
time.Sleep(time.Second)
result, err = c.Set("foo", "bar", 100)
node = result.Node
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL < 95 {
if err != nil {
t.Fatal(err)
}
t.Fatalf("Set 2 failed with %s %s %v", node.Key, node.Value, node.TTL)
}
}
// Create a five nodes
// Randomly kill one of the node and keep on sending set command to the cluster
func TestMultiNodeKillOne(t *testing.T) {
procAttr := new(os.ProcAttr)
procAttr.Files = []*os.File{nil, os.Stdout, os.Stderr}
clusterSize := 5
argGroup, etcds, err := CreateCluster(clusterSize, procAttr, false)
if err != nil {
t.Fatal("cannot create cluster")
}
defer DestroyCluster(etcds)
time.Sleep(2 * time.Second)
c := etcd.NewClient(nil)
c.SyncCluster()
stop := make(chan bool)
// Test Set
go Set(stop)
for i := 0; i < 10; i++ {
num := rand.Int() % clusterSize
fmt.Println("kill node", num+1)
// kill
etcds[num].Kill()
etcds[num].Release()
time.Sleep(time.Second)
// restart
etcds[num], err = os.StartProcess(EtcdBinPath, argGroup[num], procAttr)
if err != nil {
panic(err)
}
time.Sleep(time.Second)
}
fmt.Println("stop")
stop <- true
<-stop
}
// remove the node and node rejoin with previous log
func TestRemoveNode(t *testing.T) {
procAttr := new(os.ProcAttr)
procAttr.Files = []*os.File{nil, os.Stdout, os.Stderr}
clusterSize := 3
argGroup, etcds, _ := CreateCluster(clusterSize, procAttr, false)
defer DestroyCluster(etcds)
time.Sleep(time.Second)
c := etcd.NewClient(nil)
c.SyncCluster()
rmReq, _ := http.NewRequest("DELETE", "http://127.0.0.1:7001/remove/node3", nil)
client := &http.Client{}
for i := 0; i < 2; i++ {
for i := 0; i < 2; i++ {
client.Do(rmReq)
fmt.Println("send remove to node3 and wait for its exiting")
etcds[2].Wait()
resp, err := c.Get("_etcd/machines", false, false)
if err != nil {
panic(err)
}
if len(resp.Node.Nodes) != 2 {
t.Fatal("cannot remove peer")
}
if i == 1 {
// rejoin with log
etcds[2], err = os.StartProcess(EtcdBinPath, argGroup[2], procAttr)
} else {
// rejoin without log
etcds[2], err = os.StartProcess(EtcdBinPath, append(argGroup[2], "-f"), procAttr)
}
if err != nil {
panic(err)
}
time.Sleep(time.Second)
resp, err = c.Get("_etcd/machines", false, false)
if err != nil {
panic(err)
}
if len(resp.Node.Nodes) != 3 {
t.Fatalf("add peer fails #1 (%d != 3)", len(resp.Node.Nodes))
}
}
// first kill the node, then remove it, then add it back
for i := 0; i < 2; i++ {
etcds[2].Kill()
fmt.Println("kill node3 and wait for its exiting")
etcds[2].Wait()
client.Do(rmReq)
resp, err := c.Get("_etcd/machines", false, false)
if err != nil {
panic(err)
}
if len(resp.Node.Nodes) != 2 {
t.Fatal("cannot remove peer")
}
if i == 1 {
// rejoin with log
etcds[2], err = os.StartProcess(EtcdBinPath, append(argGroup[2]), procAttr)
} else {
// rejoin without log
etcds[2], err = os.StartProcess(EtcdBinPath, append(argGroup[2], "-f"), procAttr)
}
if err != nil {
panic(err)
}
time.Sleep(time.Second)
resp, err = c.Get("_etcd/machines", false, false)
if err != nil {
panic(err)
}
if len(resp.Node.Nodes) != 3 {
t.Fatalf("add peer fails #2 (%d != 3)", len(resp.Node.Nodes))
}
}
}
}
// Create a single node and try to set value
func TestSingleNode(t *testing.T) {
procAttr := new(os.ProcAttr)
procAttr.Files = []*os.File{nil, os.Stdout, os.Stderr}
args := []string{"etcd", "-name=node1", "-f", "-data-dir=/tmp/node1"}
process, err := os.StartProcess(EtcdBinPath, args, procAttr)
if err != nil {
t.Fatal("start process failed:" + err.Error())
return
}
defer process.Kill()
time.Sleep(time.Second)
c := etcd.NewClient(nil)
c.SyncCluster()
// Test Set
result, err := c.Set("foo", "bar", 100)
node := result.Node
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL < 95 {
if err != nil {
t.Fatal("Set 1: ", err)
}
t.Fatalf("Set 1 failed with %s %s %v", node.Key, node.Value, node.TTL)
}
time.Sleep(time.Second)
result, err = c.Set("foo", "bar", 100)
node = result.Node
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL != 100 {
if err != nil {
t.Fatal("Set 2: ", err)
}
t.Fatalf("Set 2 failed with %s %s %v", node.Key, node.Value, node.TTL)
}
// Add a test-and-set test
// First, we'll test we can change the value if we get it write
result, err = c.CompareAndSwap("foo", "foobar", 100, "bar", 0)
node = result.Node
if err != nil || node.Key != "/foo" || node.Value != "foobar" || node.TTL != 100 {
if err != nil {
t.Fatal(err)
}
t.Fatalf("Set 3 failed with %s %s %v", node.Key, node.Value, node.TTL)
}
// Next, we'll make sure we can't set it without the correct prior value
_, err = c.CompareAndSwap("foo", "foofoo", 100, "bar", 0)
if err == nil {
t.Fatalf("Set 4 expecting error when setting key with incorrect previous value")
}
}
// This test creates a single node and then set a value to it to trigger snapshot
func TestSimpleSnapshot(t *testing.T) {
procAttr := new(os.ProcAttr)
procAttr.Files = []*os.File{nil, os.Stdout, os.Stderr}
args := []string{"etcd", "-name=node1", "-data-dir=/tmp/node1", "-snapshot=true", "-snapshot-count=500"}
process, err := os.StartProcess(EtcdBinPath, append(args, "-f"), procAttr)
if err != nil {
t.Fatal("start process failed:" + err.Error())
}
defer process.Kill()
time.Sleep(time.Second)
c := etcd.NewClient(nil)
c.SyncCluster()
// issue first 501 commands
for i := 0; i < 501; i++ {
result, err := c.Set("foo", "bar", 100)
node := result.Node
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL < 95 {
if err != nil {
t.Fatal(err)
}
t.Fatalf("Set failed with %s %s %v", node.Key, node.Value, node.TTL)
}
}
// wait for a snapshot interval
time.Sleep(3 * time.Second)
snapshots, err := ioutil.ReadDir("/tmp/node1/snapshot")
if err != nil {
t.Fatal("list snapshot failed:" + err.Error())
}
if len(snapshots) != 1 {
t.Fatal("wrong number of snapshot :[1/", len(snapshots), "]")
}
index, _ := strconv.Atoi(snapshots[0].Name()[2:5])
if index < 507 || index > 510 {
t.Fatal("wrong name of snapshot :", snapshots[0].Name())
}
// issue second 501 commands
for i := 0; i < 501; i++ {
result, err := c.Set("foo", "bar", 100)
node := result.Node
if err != nil || node.Key != "/foo" || node.Value != "bar" || node.TTL < 95 {
if err != nil {
t.Fatal(err)
}
t.Fatalf("Set failed with %s %s %v", node.Key, node.Value, node.TTL)
}
}
// wait for a snapshot interval
time.Sleep(3 * time.Second)
snapshots, err = ioutil.ReadDir("/tmp/node1/snapshot")
if err != nil {
t.Fatal("list snapshot failed:" + err.Error())
}
if len(snapshots) != 1 {
t.Fatal("wrong number of snapshot :[1/", len(snapshots), "]")
}
index, _ = strconv.Atoi(snapshots[0].Name()[2:6])
if index < 1014 || index > 1017 {
t.Fatal("wrong name of snapshot :", snapshots[0].Name())
}
}