diff --git a/snapshot/member_test.go b/snapshot/member_test.go
new file mode 100644
index 000000000000..a7a74b7d0366
--- /dev/null
+++ b/snapshot/member_test.go
@@ -0,0 +1,116 @@
+// Copyright 2018 The etcd Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package snapshot
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/coreos/etcd/clientv3"
+	"github.com/coreos/etcd/embed"
+	"github.com/coreos/etcd/etcdserver"
+)
+
+// TestSnapshotV3RestoreMultiMemberAdd ensures that multiple members
+// can boot into the same cluster after being restored from a same
+// snapshot file, and also be able to add another member to the cluster.
+func TestSnapshotV3RestoreMultiMemberAdd(t *testing.T) {
+	kvs := []kv{{"foo1", "bar1"}, {"foo2", "bar2"}, {"foo3", "bar3"}}
+	dbPath := createSnapshot(t, kvs)
+
+	clusterN := 3
+	cURLs, pURLs, srvs := restoreCluster(t, clusterN, dbPath)
+	defer func() {
+		for i := 0; i < clusterN; i++ {
+			os.RemoveAll(srvs[i].Config().Dir)
+			srvs[i].Close()
+		}
+	}()
+
+	// wait for health interval + leader election
+	time.Sleep(etcdserver.HealthInterval + 2*time.Second)
+
+	cli, err := clientv3.New(clientv3.Config{Endpoints: []string{cURLs[0].String()}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer cli.Close()
+
+	urls := newEmbedURLs(2)
+	newCURLs, newPURLs := urls[:1], urls[1:]
+	if _, err = cli.MemberAdd(context.Background(), []string{newPURLs[0].String()}); err != nil {
+		t.Fatal(err)
+	}
+
+	cfg := embed.NewConfig()
+	cfg.Name = "3"
+	cfg.InitialClusterToken = testClusterTkn
+	cfg.ClusterState = "existing"
+	cfg.LCUrls, cfg.ACUrls = newCURLs, newCURLs
+	cfg.LPUrls, cfg.APUrls = newPURLs, newPURLs
+	cfg.InitialCluster = ""
+	for i := 0; i < clusterN; i++ {
+		cfg.InitialCluster += fmt.Sprintf(",%d=%s", i, pURLs[i].String())
+	}
+	cfg.InitialCluster = cfg.InitialCluster[1:]
+	cfg.InitialCluster += fmt.Sprintf(",%s=%s", cfg.Name, newPURLs[0].String())
+	cfg.Dir = filepath.Join(os.TempDir(), fmt.Sprint(time.Now().Nanosecond()))
+
+	srv, err := embed.StartEtcd(cfg)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer func() {
+		os.RemoveAll(cfg.Dir)
+		srv.Close()
+	}()
+	select {
+	case <-srv.Server.ReadyNotify():
+	case <-time.After(10 * time.Second):
+		t.Fatalf("failed to start embed.Etcd")
+	}
+
+	cli2, err := clientv3.New(clientv3.Config{Endpoints: []string{newCURLs[0].String()}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer cli2.Close()
+
+	ctx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
+	mresp, err := cli2.MemberList(ctx)
+	cancel()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(mresp.Members) != 4 {
+		t.Fatalf("expected 4 members, got %+v", mresp)
+	}
+	for i := range kvs {
+		var gresp *clientv3.GetResponse
+		ctx, cancel = context.WithTimeout(context.Background(), 3*time.Second)
+		gresp, err = cli2.Get(ctx, kvs[i].k)
+		cancel()
+		if err != nil {
+			t.Fatal(err)
+		}
+		if string(gresp.Kvs[0].Value) != kvs[i].v {
+			t.Fatalf("#%d: value expected %s, got %s", i, kvs[i].v, string(gresp.Kvs[0].Value))
+		}
+	}
+}
diff --git a/snapshot/v3_snapshot_test.go b/snapshot/v3_snapshot_test.go
new file mode 100644
index 000000000000..3574818f839a
--- /dev/null
+++ b/snapshot/v3_snapshot_test.go
@@ -0,0 +1,276 @@
+// Copyright 2018 The etcd Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package snapshot
+
+import (
+	"context"
+	"fmt"
+	"math/rand"
+	"net/url"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/coreos/etcd/clientv3"
+	"github.com/coreos/etcd/embed"
+	"github.com/coreos/etcd/pkg/logger"
+	"github.com/coreos/etcd/pkg/types"
+
+	"github.com/coreos/pkg/capnslog"
+)
+
+// TestSnapshotV3RestoreSingle tests single node cluster restoring
+// from a snapshot file.
+func TestSnapshotV3RestoreSingle(t *testing.T) {
+	kvs := []kv{{"foo1", "bar1"}, {"foo2", "bar2"}, {"foo3", "bar3"}}
+	dbPath := createSnapshot(t, kvs)
+	defer os.RemoveAll(dbPath)
+
+	clusterN := 1
+	urls := newEmbedURLs(clusterN * 2)
+	cURLs, pURLs := urls[:clusterN], urls[clusterN:]
+
+	cfg := embed.NewConfig()
+	cfg.Name = "s1"
+	cfg.InitialClusterToken = testClusterTkn
+	cfg.ClusterState = "existing"
+	cfg.LCUrls, cfg.ACUrls = cURLs, cURLs
+	cfg.LPUrls, cfg.APUrls = pURLs, pURLs
+	cfg.InitialCluster = fmt.Sprintf("%s=%s", cfg.Name, pURLs[0].String())
+	cfg.Dir = filepath.Join(os.TempDir(), fmt.Sprint(time.Now().Nanosecond()))
+
+	sp := NewV3(nil, logger.NewPackageLogger(capnslog.NewPackageLogger("github.com/coreos/etcd", "snapshot")))
+
+	err := sp.Restore(dbPath, RestoreConfig{})
+	if err.Error() != `couldn't find local name "" in the initial cluster configuration` {
+		t.Fatalf("expected restore error, got %v", err)
+	}
+	var iURLs types.URLsMap
+	iURLs, err = types.NewURLsMap(cfg.InitialCluster)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err = sp.Restore(dbPath, RestoreConfig{
+		Name:                cfg.Name,
+		OutputDataDir:       cfg.Dir,
+		InitialCluster:      iURLs,
+		InitialClusterToken: cfg.InitialClusterToken,
+		PeerURLs:            pURLs,
+	}); err != nil {
+		t.Fatal(err)
+	}
+
+	var srv *embed.Etcd
+	srv, err = embed.StartEtcd(cfg)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer func() {
+		os.RemoveAll(cfg.Dir)
+		srv.Close()
+	}()
+	select {
+	case <-srv.Server.ReadyNotify():
+	case <-time.After(3 * time.Second):
+		t.Fatalf("failed to start embed.Etcd")
+	}
+
+	var cli *clientv3.Client
+	cli, err = clientv3.New(clientv3.Config{Endpoints: []string{cfg.ACUrls[0].String()}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer cli.Close()
+	for i := range kvs {
+		var gresp *clientv3.GetResponse
+		gresp, err = cli.Get(context.Background(), kvs[i].k)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if string(gresp.Kvs[0].Value) != kvs[i].v {
+			t.Fatalf("#%d: value expected %s, got %s", i, kvs[i].v, string(gresp.Kvs[0].Value))
+		}
+	}
+}
+
+// TestSnapshotV3RestoreMulti ensures that multiple members
+// can boot into the same cluster after being restored from a same
+// snapshot file.
+func TestSnapshotV3RestoreMulti(t *testing.T) {
+	kvs := []kv{{"foo1", "bar1"}, {"foo2", "bar2"}, {"foo3", "bar3"}}
+	dbPath := createSnapshot(t, kvs)
+	defer os.RemoveAll(dbPath)
+
+	clusterN := 3
+	cURLs, _, srvs := restoreCluster(t, clusterN, dbPath)
+	defer func() {
+		for i := 0; i < clusterN; i++ {
+			os.RemoveAll(srvs[i].Config().Dir)
+			srvs[i].Close()
+		}
+	}()
+
+	// wait for leader election
+	time.Sleep(time.Second)
+
+	for i := 0; i < clusterN; i++ {
+		cli, err := clientv3.New(clientv3.Config{Endpoints: []string{cURLs[i].String()}})
+		if err != nil {
+			t.Fatal(err)
+		}
+		defer cli.Close()
+		for i := range kvs {
+			var gresp *clientv3.GetResponse
+			gresp, err = cli.Get(context.Background(), kvs[i].k)
+			if err != nil {
+				t.Fatal(err)
+			}
+			if string(gresp.Kvs[0].Value) != kvs[i].v {
+				t.Fatalf("#%d: value expected %s, got %s", i, kvs[i].v, string(gresp.Kvs[0].Value))
+			}
+		}
+	}
+}
+
+type kv struct {
+	k, v string
+}
+
+func createSnapshot(t *testing.T, kvs []kv) string {
+	clusterN := 1
+	urls := newEmbedURLs(clusterN * 2)
+	cURLs, pURLs := urls[:clusterN], urls[clusterN:]
+
+	cfg := embed.NewConfig()
+	cfg.Name = "default"
+	cfg.ClusterState = "new"
+	cfg.LCUrls, cfg.ACUrls = cURLs, cURLs
+	cfg.LPUrls, cfg.APUrls = pURLs, pURLs
+	cfg.InitialCluster = fmt.Sprintf("%s=%s", cfg.Name, pURLs[0].String())
+	cfg.Dir = filepath.Join(os.TempDir(), fmt.Sprint(time.Now().Nanosecond()))
+	srv, err := embed.StartEtcd(cfg)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer func() {
+		os.RemoveAll(cfg.Dir)
+		srv.Close()
+	}()
+	select {
+	case <-srv.Server.ReadyNotify():
+	case <-time.After(3 * time.Second):
+		t.Fatalf("failed to start embed.Etcd")
+	}
+
+	cli, err := clientv3.New(clientv3.Config{Endpoints: []string{cfg.ACUrls[0].String()}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	for i := range kvs {
+		if _, err = cli.Put(context.Background(), kvs[i].k, kvs[i].v); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	sp := NewV3(cli, logger.NewPackageLogger(capnslog.NewPackageLogger("github.com/coreos/etcd", "snapshot")))
+	dpPath := filepath.Join(os.TempDir(), fmt.Sprintf("snapshot%d.db", time.Now().Nanosecond()))
+	if err = sp.Save(context.Background(), dpPath); err != nil {
+		t.Fatal(err)
+	}
+
+	os.RemoveAll(cfg.Dir)
+	srv.Close()
+	return dpPath
+}
+
+const testClusterTkn = "tkn"
+
+func restoreCluster(t *testing.T, clusterN int, dbPath string) (
+	cURLs []url.URL,
+	pURLs []url.URL,
+	srvs []*embed.Etcd) {
+	urls := newEmbedURLs(clusterN * 2)
+	cURLs, pURLs = urls[:clusterN], urls[clusterN:]
+
+	ics := ""
+	for i := 0; i < clusterN; i++ {
+		ics += fmt.Sprintf(",%d=%s", i, pURLs[i].String())
+	}
+	ics = ics[1:]
+	iURLs, err := types.NewURLsMap(ics)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	cfgs := make([]*embed.Config, clusterN)
+	for i := 0; i < clusterN; i++ {
+		cfg := embed.NewConfig()
+		cfg.Name = fmt.Sprintf("%d", i)
+		cfg.InitialClusterToken = testClusterTkn
+		cfg.ClusterState = "existing"
+		cfg.LCUrls, cfg.ACUrls = []url.URL{cURLs[i]}, []url.URL{cURLs[i]}
+		cfg.LPUrls, cfg.APUrls = []url.URL{pURLs[i]}, []url.URL{pURLs[i]}
+		cfg.InitialCluster = ics
+		cfg.Dir = filepath.Join(os.TempDir(), fmt.Sprint(time.Now().Nanosecond()+i))
+
+		sp := NewV3(nil, logger.NewPackageLogger(capnslog.NewPackageLogger("github.com/coreos/etcd", "snapshot")))
+		if err := sp.Restore(dbPath, RestoreConfig{
+			Name:                cfg.Name,
+			OutputDataDir:       cfg.Dir,
+			InitialCluster:      iURLs,
+			InitialClusterToken: cfg.InitialClusterToken,
+			PeerURLs:            types.URLs{pURLs[i]},
+		}); err != nil {
+			t.Fatal(err)
+		}
+		cfgs[i] = cfg
+	}
+
+	sch := make(chan *embed.Etcd)
+	for i := range cfgs {
+		go func(idx int) {
+			srv, err := embed.StartEtcd(cfgs[idx])
+			if err != nil {
+				t.Fatal(err)
+			}
+
+			<-srv.Server.ReadyNotify()
+			sch <- srv
+		}(i)
+	}
+
+	srvs = make([]*embed.Etcd, clusterN)
+	for i := 0; i < clusterN; i++ {
+		select {
+		case srv := <-sch:
+			srvs[i] = srv
+		case <-time.After(5 * time.Second):
+			t.Fatalf("#%d: failed to start embed.Etcd", i)
+		}
+	}
+	return cURLs, pURLs, srvs
+}
+
+// TODO: TLS
+func newEmbedURLs(n int) (urls []url.URL) {
+	urls = make([]url.URL, n)
+	for i := 0; i < n; i++ {
+		rand.Seed(int64(time.Now().Nanosecond()))
+		u, _ := url.Parse(fmt.Sprintf("unix://localhost:%d", rand.Intn(45000)))
+		urls[i] = *u
+	}
+	return urls
+}