diff --git a/cluster.go b/cluster.go
index 647a25be..72bace76 100644
--- a/cluster.go
+++ b/cluster.go
@@ -14,8 +14,8 @@ import (
 	"github.com/go-redis/redis/internal/proto"
 )
 
-var errClusterNoNodes = internal.RedisError("redis: cluster has no nodes")
-var errNilClusterState = internal.RedisError("redis: cannot load cluster slots")
+var errClusterNoNodes = fmt.Errorf("redis: cluster has no nodes")
+var errNilClusterState = fmt.Errorf("redis: cannot load cluster slots")
 
 // ClusterOptions are used to configure a cluster client and should be
 // passed to NewClusterClient.
@@ -64,6 +64,19 @@ func (opt *ClusterOptions) init() {
 		opt.ReadOnly = true
 	}
 
+	switch opt.ReadTimeout {
+	case -1:
+		opt.ReadTimeout = 0
+	case 0:
+		opt.ReadTimeout = 3 * time.Second
+	}
+	switch opt.WriteTimeout {
+	case -1:
+		opt.WriteTimeout = 0
+	case 0:
+		opt.WriteTimeout = opt.ReadTimeout
+	}
+
 	switch opt.MinRetryBackoff {
 	case -1:
 		opt.MinRetryBackoff = 0
@@ -192,6 +205,19 @@ func (c *clusterNodes) Close() error {
 	return firstErr
 }
 
+func (c *clusterNodes) Err() error {
+	c.mu.RLock()
+	defer c.mu.RUnlock()
+
+	if c.closed {
+		return pool.ErrClosed
+	}
+	if len(c.addrs) == 0 {
+		return errClusterNoNodes
+	}
+	return nil
+}
+
 func (c *clusterNodes) NextGeneration() uint32 {
 	c.generation++
 	return c.generation
@@ -468,13 +494,22 @@ func (c *ClusterClient) Options() *ClusterOptions {
 	return c.opt
 }
 
-func (c *ClusterClient) state() *clusterState {
+func (c *ClusterClient) retryBackoff(attempt int) time.Duration {
+	return internal.RetryBackoff(attempt, c.opt.MinRetryBackoff, c.opt.MaxRetryBackoff)
+}
+
+func (c *ClusterClient) state() (*clusterState, error) {
 	v := c._state.Load()
 	if v != nil {
-		return v.(*clusterState)
+		return v.(*clusterState), nil
 	}
+
+	if err := c.nodes.Err(); err != nil {
+		return nil, err
+	}
+
 	c.lazyReloadState()
-	return nil
+	return nil, errNilClusterState
 }
 
 func (c *ClusterClient) cmdInfo(name string) *CommandInfo {
@@ -495,15 +530,20 @@ func (c *ClusterClient) cmdInfo(name string) *CommandInfo {
 	if err != nil {
 		return nil
 	}
-	return c.cmdsInfo[name]
+	info := c.cmdsInfo[name]
+	if info == nil {
+		internal.Logf("info for cmd=%s not found", name)
+	}
+	return info
+}
+
+func (c *ClusterClient) cmdSlot(cmd Cmder) int {
+	cmdInfo := c.cmdInfo(cmd.Name())
+	firstKey := cmd.arg(cmdFirstKeyPos(cmd, cmdInfo))
+	return hashtag.Slot(firstKey)
 }
 
 func (c *ClusterClient) cmdSlotAndNode(state *clusterState, cmd Cmder) (int, *clusterNode, error) {
-	if state == nil {
-		node, err := c.nodes.Random()
-		return 0, node, err
-	}
-
 	cmdInfo := c.cmdInfo(cmd.Name())
 	firstKey := cmd.arg(cmdFirstKeyPos(cmd, cmdInfo))
 	slot := hashtag.Slot(firstKey)
@@ -523,19 +563,51 @@ func (c *ClusterClient) cmdSlotAndNode(state *clusterState, cmd Cmder) (int, *cl
 }
 
 func (c *ClusterClient) Watch(fn func(*Tx) error, keys ...string) error {
-	state := c.state()
-
-	var node *clusterNode
-	var err error
-	if state != nil && len(keys) > 0 {
-		node, err = state.slotMasterNode(hashtag.Slot(keys[0]))
-	} else {
-		node, err = c.nodes.Random()
+	if len(keys) == 0 {
+		return fmt.Errorf("redis: keys don't hash to the same slot")
 	}
+
+	state, err := c.state()
 	if err != nil {
 		return err
 	}
-	return node.Client.Watch(fn, keys...)
+
+	slot := hashtag.Slot(keys[0])
+	for _, key := range keys[1:] {
+		if hashtag.Slot(key) != slot {
+			return fmt.Errorf("redis: Watch requires all keys to be in the same slot")
+		}
+	}
+
+	node, err := state.slotMasterNode(slot)
+	if err != nil {
+		return err
+	}
+
+	for attempt := 0; attempt <= c.opt.MaxRedirects; attempt++ {
+		if attempt > 0 {
+			time.Sleep(c.retryBackoff(attempt))
+		}
+
+		err = node.Client.Watch(fn, keys...)
+		if err == nil {
+			break
+		}
+
+		moved, ask, addr := internal.IsMovedError(err)
+		if moved || ask {
+			c.lazyReloadState()
+			node, err = c.nodes.GetOrCreate(addr)
+			if err != nil {
+				return err
+			}
+			continue
+		}
+
+		return err
+	}
+
+	return err
 }
 
 // Close closes the cluster client, releasing any open resources.
@@ -547,7 +619,13 @@ func (c *ClusterClient) Close() error {
 }
 
 func (c *ClusterClient) Process(cmd Cmder) error {
-	slot, node, err := c.cmdSlotAndNode(c.state(), cmd)
+	state, err := c.state()
+	if err != nil {
+		cmd.setErr(err)
+		return err
+	}
+
+	_, node, err := c.cmdSlotAndNode(state, cmd)
 	if err != nil {
 		cmd.setErr(err)
 		return err
@@ -556,7 +634,7 @@ func (c *ClusterClient) Process(cmd Cmder) error {
 	var ask bool
 	for attempt := 0; attempt <= c.opt.MaxRedirects; attempt++ {
 		if attempt > 0 {
-			time.Sleep(node.Client.retryBackoff(attempt))
+			time.Sleep(c.retryBackoff(attempt))
 		}
 
 		if ask {
@@ -572,7 +650,7 @@ func (c *ClusterClient) Process(cmd Cmder) error {
 
 		// If there is no error - we are done.
 		if err == nil {
-			return nil
+			break
 		}
 
 		// If slave is loading - read from master.
@@ -582,12 +660,11 @@ func (c *ClusterClient) Process(cmd Cmder) error {
 			continue
 		}
 
-		// On network errors try random node.
-		if internal.IsRetryableError(err) || internal.IsClusterDownError(err) {
-			node, err = c.nodes.Random()
-			if err != nil {
-				cmd.setErr(err)
-				return err
+		if internal.IsRetryableError(err) {
+			var nodeErr error
+			node, nodeErr = c.nodes.Random()
+			if nodeErr != nil {
+				break
 			}
 			continue
 		}
@@ -596,20 +673,13 @@ func (c *ClusterClient) Process(cmd Cmder) error {
 		var addr string
 		moved, ask, addr = internal.IsMovedError(err)
 		if moved || ask {
-			state := c.state()
-			if state != nil && slot >= 0 {
-				master, _ := state.slotMasterNode(slot)
-				if moved && (master == nil || master.Client.getAddr() != addr) {
-					c.lazyReloadState()
-				}
-			}
+			c.lazyReloadState()
 
-			node, err = c.nodes.GetOrCreate(addr)
-			if err != nil {
-				cmd.setErr(err)
-				return err
+			var nodeErr error
+			node, nodeErr = c.nodes.GetOrCreate(addr)
+			if nodeErr != nil {
+				break
 			}
-
 			continue
 		}
 
@@ -622,9 +692,9 @@ func (c *ClusterClient) Process(cmd Cmder) error {
 // ForEachMaster concurrently calls the fn on each master node in the cluster.
 // It returns the first error if any.
 func (c *ClusterClient) ForEachMaster(fn func(client *Client) error) error {
-	state := c.state()
-	if state == nil {
-		return errNilClusterState
+	state, err := c.state()
+	if err != nil {
+		return err
 	}
 
 	var wg sync.WaitGroup
@@ -655,9 +725,9 @@ func (c *ClusterClient) ForEachMaster(fn func(client *Client) error) error {
 // ForEachSlave concurrently calls the fn on each slave node in the cluster.
 // It returns the first error if any.
 func (c *ClusterClient) ForEachSlave(fn func(client *Client) error) error {
-	state := c.state()
-	if state == nil {
-		return errNilClusterState
+	state, err := c.state()
+	if err != nil {
+		return err
 	}
 
 	var wg sync.WaitGroup
@@ -688,9 +758,9 @@ func (c *ClusterClient) ForEachSlave(fn func(client *Client) error) error {
 // ForEachNode concurrently calls the fn on each known node in the cluster.
 // It returns the first error if any.
 func (c *ClusterClient) ForEachNode(fn func(client *Client) error) error {
-	state := c.state()
-	if state == nil {
-		return errNilClusterState
+	state, err := c.state()
+	if err != nil {
+		return err
 	}
 
 	var wg sync.WaitGroup
@@ -728,7 +798,7 @@ func (c *ClusterClient) ForEachNode(fn func(client *Client) error) error {
 func (c *ClusterClient) PoolStats() *PoolStats {
 	var acc PoolStats
 
-	state := c.state()
+	state, _ := c.state()
 	if state == nil {
 		return &acc
 	}
@@ -762,10 +832,8 @@ func (c *ClusterClient) lazyReloadState() {
 	go func() {
 		defer atomic.StoreUint32(&c.reloading, 0)
 
-		var state *clusterState
 		for {
-			var err error
-			state, err = c.reloadState()
+			state, err := c.reloadState()
 			if err == pool.ErrClosed {
 				return
 			}
@@ -776,11 +844,10 @@ func (c *ClusterClient) lazyReloadState() {
 			}
 
 			c._state.Store(state)
+			time.Sleep(5 * time.Second)
+			c.nodes.GC(state.generation)
 			break
 		}
-
-		time.Sleep(3 * time.Second)
-		c.nodes.GC(state.generation)
 	}()
 }
 
@@ -843,10 +910,15 @@ func (c *ClusterClient) Pipelined(fn func(Pipeliner) error) ([]Cmder, error) {
 func (c *ClusterClient) pipelineExec(cmds []Cmder) error {
 	cmdsMap, err := c.mapCmdsByNode(cmds)
 	if err != nil {
+		setCmdsErr(cmds, err)
 		return err
 	}
 
-	for i := 0; i <= c.opt.MaxRedirects; i++ {
+	for attempt := 0; attempt <= c.opt.MaxRedirects; attempt++ {
+		if attempt > 0 {
+			time.Sleep(c.retryBackoff(attempt))
+		}
+
 		failedCmds := make(map[*clusterNode][]Cmder)
 
 		for node, cmds := range cmdsMap {
@@ -856,8 +928,12 @@ func (c *ClusterClient) pipelineExec(cmds []Cmder) error {
 				continue
 			}
 
-			err = c.pipelineProcessCmds(cn, cmds, failedCmds)
-			node.Client.releaseConn(cn, err)
+			err = c.pipelineProcessCmds(node, cn, cmds, failedCmds)
+			if err == nil || internal.IsRedisError(err) {
+				_ = node.Client.connPool.Put(cn)
+			} else {
+				_ = node.Client.connPool.Remove(cn)
+			}
 		}
 
 		if len(failedCmds) == 0 {
@@ -866,21 +942,20 @@ func (c *ClusterClient) pipelineExec(cmds []Cmder) error {
 		cmdsMap = failedCmds
 	}
 
-	var firstErr error
-	for _, cmd := range cmds {
-		if err := cmd.Err(); err != nil {
-			firstErr = err
-			break
-		}
-	}
-	return firstErr
+	return firstCmdsErr(cmds)
 }
 
 func (c *ClusterClient) mapCmdsByNode(cmds []Cmder) (map[*clusterNode][]Cmder, error) {
-	state := c.state()
+	state, err := c.state()
+	if err != nil {
+		setCmdsErr(cmds, err)
+		return nil, err
+	}
+
 	cmdsMap := make(map[*clusterNode][]Cmder)
 	for _, cmd := range cmds {
-		_, node, err := c.cmdSlotAndNode(state, cmd)
+		slot := c.cmdSlot(cmd)
+		node, err := state.slotMasterNode(slot)
 		if err != nil {
 			return nil, err
 		}
@@ -890,11 +965,12 @@ func (c *ClusterClient) mapCmdsByNode(cmds []Cmder) (map[*clusterNode][]Cmder, e
 }
 
 func (c *ClusterClient) pipelineProcessCmds(
-	cn *pool.Conn, cmds []Cmder, failedCmds map[*clusterNode][]Cmder,
+	node *clusterNode, cn *pool.Conn, cmds []Cmder, failedCmds map[*clusterNode][]Cmder,
 ) error {
 	cn.SetWriteTimeout(c.opt.WriteTimeout)
 	if err := writeCmd(cn, cmds...); err != nil {
 		setCmdsErr(cmds, err)
+		failedCmds[node] = cmds
 		return err
 	}
 
@@ -907,46 +983,53 @@ func (c *ClusterClient) pipelineProcessCmds(
 func (c *ClusterClient) pipelineReadCmds(
 	cn *pool.Conn, cmds []Cmder, failedCmds map[*clusterNode][]Cmder,
 ) error {
-	var firstErr error
 	for _, cmd := range cmds {
 		err := cmd.readReply(cn)
 		if err == nil {
 			continue
 		}
 
-		if firstErr == nil {
-			firstErr = err
+		if c.checkMovedErr(cmd, err, failedCmds) {
+			continue
 		}
 
-		err = c.checkMovedErr(cmd, failedCmds)
-		if err != nil && firstErr == nil {
-			firstErr = err
+		if internal.IsRedisError(err) {
+			continue
 		}
+
+		return err
 	}
-	return firstErr
+	return nil
 }
 
-func (c *ClusterClient) checkMovedErr(cmd Cmder, failedCmds map[*clusterNode][]Cmder) error {
-	moved, ask, addr := internal.IsMovedError(cmd.Err())
+func (c *ClusterClient) checkMovedErr(
+	cmd Cmder, err error, failedCmds map[*clusterNode][]Cmder,
+) bool {
+	moved, ask, addr := internal.IsMovedError(err)
+
 	if moved {
 		c.lazyReloadState()
 
 		node, err := c.nodes.GetOrCreate(addr)
 		if err != nil {
-			return err
+			return false
 		}
 
 		failedCmds[node] = append(failedCmds[node], cmd)
+		return true
 	}
+
 	if ask {
 		node, err := c.nodes.GetOrCreate(addr)
 		if err != nil {
-			return err
+			return false
 		}
 
 		failedCmds[node] = append(failedCmds[node], NewCmd("ASKING"), cmd)
+		return true
 	}
-	return nil
+
+	return false
 }
 
 // TxPipeline acts like Pipeline, but wraps queued commands with MULTI/EXEC.
@@ -963,25 +1046,25 @@ func (c *ClusterClient) TxPipelined(fn func(Pipeliner) error) ([]Cmder, error) {
 }
 
 func (c *ClusterClient) txPipelineExec(cmds []Cmder) error {
-	cmdsMap, err := c.mapCmdsBySlot(cmds)
+	state, err := c.state()
 	if err != nil {
 		return err
 	}
 
-	state := c.state()
-	if state == nil {
-		return errNilClusterState
-	}
-
+	cmdsMap := c.mapCmdsBySlot(cmds)
 	for slot, cmds := range cmdsMap {
 		node, err := state.slotMasterNode(slot)
 		if err != nil {
 			setCmdsErr(cmds, err)
 			continue
 		}
-
 		cmdsMap := map[*clusterNode][]Cmder{node: cmds}
-		for i := 0; i <= c.opt.MaxRedirects; i++ {
+
+		for attempt := 0; attempt <= c.opt.MaxRedirects; attempt++ {
+			if attempt > 0 {
+				time.Sleep(c.retryBackoff(attempt))
+			}
+
 			failedCmds := make(map[*clusterNode][]Cmder)
 
 			for node, cmds := range cmdsMap {
@@ -992,7 +1075,11 @@ func (c *ClusterClient) txPipelineExec(cmds []Cmder) error {
 				}
 
 				err = c.txPipelineProcessCmds(node, cn, cmds, failedCmds)
-				node.Client.releaseConn(cn, err)
+				if err == nil || internal.IsRedisError(err) {
+					_ = node.Client.connPool.Put(cn)
+				} else {
+					_ = node.Client.connPool.Remove(cn)
+				}
 			}
 
 			if len(failedCmds) == 0 {
@@ -1002,27 +1089,16 @@ func (c *ClusterClient) txPipelineExec(cmds []Cmder) error {
 		}
 	}
 
-	var firstErr error
-	for _, cmd := range cmds {
-		if err := cmd.Err(); err != nil {
-			firstErr = err
-			break
-		}
-	}
-	return firstErr
+	return firstCmdsErr(cmds)
 }
 
-func (c *ClusterClient) mapCmdsBySlot(cmds []Cmder) (map[int][]Cmder, error) {
-	state := c.state()
+func (c *ClusterClient) mapCmdsBySlot(cmds []Cmder) map[int][]Cmder {
 	cmdsMap := make(map[int][]Cmder)
 	for _, cmd := range cmds {
-		slot, _, err := c.cmdSlotAndNode(state, cmd)
-		if err != nil {
-			return nil, err
-		}
+		slot := c.cmdSlot(cmd)
 		cmdsMap[slot] = append(cmdsMap[slot], cmd)
 	}
-	return cmdsMap, nil
+	return cmdsMap
 }
 
 func (c *ClusterClient) txPipelineProcessCmds(
@@ -1039,22 +1115,20 @@ func (c *ClusterClient) txPipelineProcessCmds(
 	cn.SetReadTimeout(c.opt.ReadTimeout)
 
 	if err := c.txPipelineReadQueued(cn, cmds, failedCmds); err != nil {
+		setCmdsErr(cmds, err)
 		return err
 	}
 
-	_, err := pipelineReadCmds(cn, cmds)
-	return err
+	return pipelineReadCmds(cn, cmds)
 }
 
 func (c *ClusterClient) txPipelineReadQueued(
 	cn *pool.Conn, cmds []Cmder, failedCmds map[*clusterNode][]Cmder,
 ) error {
-	var firstErr error
-
 	// Parse queued replies.
 	var statusCmd StatusCmd
-	if err := statusCmd.readReply(cn); err != nil && firstErr == nil {
-		firstErr = err
+	if err := statusCmd.readReply(cn); err != nil {
+		return err
 	}
 
 	for _, cmd := range cmds {
@@ -1063,15 +1137,11 @@ func (c *ClusterClient) txPipelineReadQueued(
 			continue
 		}
 
-		cmd.setErr(err)
-		if firstErr == nil {
-			firstErr = err
+		if c.checkMovedErr(cmd, err, failedCmds) || internal.IsRedisError(err) {
+			continue
 		}
 
-		err = c.checkMovedErr(cmd, failedCmds)
-		if err != nil && firstErr == nil {
-			firstErr = err
-		}
+		return err
 	}
 
 	// Parse number of replies.
@@ -1085,7 +1155,13 @@ func (c *ClusterClient) txPipelineReadQueued(
 
 	switch line[0] {
 	case proto.ErrorReply:
-		return proto.ParseErrorReply(line)
+		err := proto.ParseErrorReply(line)
+		for _, cmd := range cmds {
+			if !c.checkMovedErr(cmd, err, failedCmds) {
+				break
+			}
+		}
+		return err
 	case proto.ArrayReply:
 		// ok
 	default:
@@ -1093,7 +1169,7 @@ func (c *ClusterClient) txPipelineReadQueued(
 		return err
 	}
 
-	return firstErr
+	return nil
 }
 
 func (c *ClusterClient) pubSub(channels []string) *PubSub {
@@ -1112,7 +1188,12 @@ func (c *ClusterClient) pubSub(channels []string) *PubSub {
 					slot = -1
 				}
 
-				masterNode, err := c.state().slotMasterNode(slot)
+				state, err := c.state()
+				if err != nil {
+					return nil, err
+				}
+
+				masterNode, err := state.slotMasterNode(slot)
 				if err != nil {
 					return nil, err
 				}
diff --git a/cluster_test.go b/cluster_test.go
index 324bd1ce..b2106da9 100644
--- a/cluster_test.go
+++ b/cluster_test.go
@@ -200,7 +200,7 @@ var _ = Describe("ClusterClient", func() {
 
 			Eventually(func() string {
 				return client.Get("A").Val()
-			}).Should(Equal("VALUE"))
+			}, 30*time.Second).Should(Equal("VALUE"))
 
 			cnt, err := client.Del("A").Result()
 			Expect(err).NotTo(HaveOccurred())
@@ -215,7 +215,7 @@ var _ = Describe("ClusterClient", func() {
 
 			Eventually(func() string {
 				return client.Get("A").Val()
-			}).Should(Equal("VALUE"))
+			}, 30*time.Second).Should(Equal("VALUE"))
 		})
 
 		It("distributes keys", func() {
@@ -227,7 +227,7 @@ var _ = Describe("ClusterClient", func() {
 			for _, master := range cluster.masters() {
 				Eventually(func() string {
 					return master.Info("keyspace").Val()
-				}, 5*time.Second).Should(Or(
+				}, 30*time.Second).Should(Or(
 					ContainSubstring("keys=31"),
 					ContainSubstring("keys=29"),
 					ContainSubstring("keys=40"),
@@ -251,7 +251,7 @@ var _ = Describe("ClusterClient", func() {
 			for _, master := range cluster.masters() {
 				Eventually(func() string {
 					return master.Info("keyspace").Val()
-				}, 5*time.Second).Should(Or(
+				}, 30*time.Second).Should(Or(
 					ContainSubstring("keys=31"),
 					ContainSubstring("keys=29"),
 					ContainSubstring("keys=40"),
@@ -320,10 +320,6 @@ var _ = Describe("ClusterClient", func() {
 					Expect(err).NotTo(HaveOccurred())
 					Expect(cmds).To(HaveLen(14))
 
-					if opt.RouteByLatency {
-						return
-					}
-
 					for _, key := range keys {
 						slot := hashtag.Slot(key)
 						client.SwapSlotNodes(slot)
@@ -432,6 +428,9 @@ var _ = Describe("ClusterClient", func() {
 		})
 
 		AfterEach(func() {
+			_ = client.ForEachMaster(func(master *redis.Client) error {
+				return master.FlushDB().Err()
+			})
 			Expect(client.Close()).NotTo(HaveOccurred())
 		})
 
@@ -560,6 +559,9 @@ var _ = Describe("ClusterClient", func() {
 		})
 
 		AfterEach(func() {
+			_ = client.ForEachMaster(func(master *redis.Client) error {
+				return master.FlushDB().Err()
+			})
 			Expect(client.Close()).NotTo(HaveOccurred())
 		})
 
@@ -575,10 +577,19 @@ var _ = Describe("ClusterClient", func() {
 			_ = client.ForEachMaster(func(master *redis.Client) error {
 				return master.FlushDB().Err()
 			})
+
+			_ = client.ForEachSlave(func(slave *redis.Client) error {
+				Eventually(func() int64 {
+					return client.DBSize().Val()
+				}, 30*time.Second).Should(Equal(int64(0)))
+				return nil
+			})
 		})
 
 		AfterEach(func() {
-			client.FlushDB()
+			_ = client.ForEachMaster(func(master *redis.Client) error {
+				return master.FlushDB().Err()
+			})
 			Expect(client.Close()).NotTo(HaveOccurred())
 		})
 
@@ -597,7 +608,7 @@ var _ = Describe("ClusterClient without nodes", func() {
 		Expect(client.Close()).NotTo(HaveOccurred())
 	})
 
-	It("returns an error", func() {
+	It("Ping returns an error", func() {
 		err := client.Ping().Err()
 		Expect(err).To(MatchError("redis: cluster has no nodes"))
 	})
@@ -626,7 +637,7 @@ var _ = Describe("ClusterClient without valid nodes", func() {
 
 	It("returns an error", func() {
 		err := client.Ping().Err()
-		Expect(err).To(MatchError("ERR This instance has cluster support disabled"))
+		Expect(err).To(MatchError("redis: cannot load cluster slots"))
 	})
 
 	It("pipeline returns an error", func() {
@@ -634,7 +645,7 @@ var _ = Describe("ClusterClient without valid nodes", func() {
 			pipe.Ping()
 			return nil
 		})
-		Expect(err).To(MatchError("ERR This instance has cluster support disabled"))
+		Expect(err).To(MatchError("redis: cannot load cluster slots"))
 	})
 })
 
@@ -664,7 +675,7 @@ var _ = Describe("ClusterClient timeout", func() {
 		It("Tx timeouts", func() {
 			err := client.Watch(func(tx *redis.Tx) error {
 				return tx.Ping().Err()
-			})
+			}, "foo")
 			Expect(err).To(HaveOccurred())
 			Expect(err.(net.Error).Timeout()).To(BeTrue())
 		})
@@ -676,42 +687,20 @@ var _ = Describe("ClusterClient timeout", func() {
 					return nil
 				})
 				return err
-			})
+			}, "foo")
 			Expect(err).To(HaveOccurred())
 			Expect(err.(net.Error).Timeout()).To(BeTrue())
 		})
 	}
 
-	Context("read timeout", func() {
+	const pause = time.Second
+
+	Context("read/write timeout", func() {
 		BeforeEach(func() {
 			opt := redisClusterOptions()
-			opt.ReadTimeout = time.Nanosecond
-			opt.WriteTimeout = -1
-			client = cluster.clusterClient(opt)
-		})
-
-		testTimeout()
-	})
-
-	Context("write timeout", func() {
-		BeforeEach(func() {
-			opt := redisClusterOptions()
-			opt.ReadTimeout = time.Nanosecond
-			opt.WriteTimeout = -1
-			client = cluster.clusterClient(opt)
-		})
-
-		testTimeout()
-	})
-
-	Context("ClientPause timeout", func() {
-		const pause = time.Second
-
-		BeforeEach(func() {
-			opt := redisClusterOptions()
-			opt.ReadTimeout = pause / 10
-			opt.WriteTimeout = pause / 10
-			opt.MaxRedirects = -1
+			opt.ReadTimeout = 100 * time.Millisecond
+			opt.WriteTimeout = 100 * time.Millisecond
+			opt.MaxRedirects = 1
 			client = cluster.clusterClient(opt)
 
 			err := client.ForEachNode(func(client *redis.Client) error {
diff --git a/command.go b/command.go
index 0e5b2016..a796a93f 100644
--- a/command.go
+++ b/command.go
@@ -46,10 +46,21 @@ type Cmder interface {
 
 func setCmdsErr(cmds []Cmder, e error) {
 	for _, cmd := range cmds {
-		cmd.setErr(e)
+		if cmd.Err() == nil {
+			cmd.setErr(e)
+		}
 	}
 }
 
+func firstCmdsErr(cmds []Cmder) error {
+	for _, cmd := range cmds {
+		if err := cmd.Err(); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
 func writeCmd(cn *pool.Conn, cmds ...Cmder) error {
 	cn.Wb.Reset()
 	for _, cmd := range cmds {
@@ -95,7 +106,6 @@ func cmdFirstKeyPos(cmd Cmder, info *CommandInfo) int {
 		return 1
 	}
 	if info == nil {
-		internal.Logf("info for cmd=%s not found", cmd.Name())
 		return -1
 	}
 	return int(info.FirstKeyPos)
diff --git a/commands_test.go b/commands_test.go
index 4298cba6..0811d12a 100644
--- a/commands_test.go
+++ b/commands_test.go
@@ -27,11 +27,21 @@ var _ = Describe("Commands", func() {
 	Describe("server", func() {
 
 		It("should Auth", func() {
-			_, err := client.Pipelined(func(pipe redis.Pipeliner) error {
+			cmds, err := client.Pipelined(func(pipe redis.Pipeliner) error {
 				pipe.Auth("password")
+				pipe.Auth("")
 				return nil
 			})
 			Expect(err).To(MatchError("ERR Client sent AUTH, but no password is set"))
+			Expect(cmds[0].Err()).To(MatchError("ERR Client sent AUTH, but no password is set"))
+			Expect(cmds[1].Err()).To(MatchError("ERR Client sent AUTH, but no password is set"))
+
+			stats := client.Pool().Stats()
+			Expect(stats.Requests).To(Equal(uint32(2)))
+			Expect(stats.Hits).To(Equal(uint32(1)))
+			Expect(stats.Timeouts).To(Equal(uint32(0)))
+			Expect(stats.TotalConns).To(Equal(uint32(1)))
+			Expect(stats.FreeConns).To(Equal(uint32(1)))
 		})
 
 		It("should Echo", func() {
@@ -187,6 +197,29 @@ var _ = Describe("Commands", func() {
 			Expect(tm).To(BeTemporally("~", time.Now(), 3*time.Second))
 		})
 
+		It("Should Command", func() {
+			cmds, err := client.Command().Result()
+			Expect(err).NotTo(HaveOccurred())
+			Expect(len(cmds)).To(BeNumerically("~", 180, 10))
+
+			cmd := cmds["mget"]
+			Expect(cmd.Name).To(Equal("mget"))
+			Expect(cmd.Arity).To(Equal(int8(-2)))
+			Expect(cmd.Flags).To(ContainElement("readonly"))
+			Expect(cmd.FirstKeyPos).To(Equal(int8(1)))
+			Expect(cmd.LastKeyPos).To(Equal(int8(-1)))
+			Expect(cmd.StepCount).To(Equal(int8(1)))
+
+			cmd = cmds["ping"]
+			Expect(cmd.Name).To(Equal("ping"))
+			Expect(cmd.Arity).To(Equal(int8(-1)))
+			Expect(cmd.Flags).To(ContainElement("stale"))
+			Expect(cmd.Flags).To(ContainElement("fast"))
+			Expect(cmd.FirstKeyPos).To(Equal(int8(0)))
+			Expect(cmd.LastKeyPos).To(Equal(int8(0)))
+			Expect(cmd.StepCount).To(Equal(int8(0)))
+		})
+
 	})
 
 	Describe("debugging", func() {
@@ -2887,24 +2920,6 @@ var _ = Describe("Commands", func() {
 
 	})
 
-	Describe("Command", func() {
-
-		It("returns map of commands", func() {
-			cmds, err := client.Command().Result()
-			Expect(err).NotTo(HaveOccurred())
-			Expect(len(cmds)).To(BeNumerically("~", 180, 10))
-
-			cmd := cmds["mget"]
-			Expect(cmd.Name).To(Equal("mget"))
-			Expect(cmd.Arity).To(Equal(int8(-2)))
-			Expect(cmd.Flags).To(ContainElement("readonly"))
-			Expect(cmd.FirstKeyPos).To(Equal(int8(1)))
-			Expect(cmd.LastKeyPos).To(Equal(int8(-1)))
-			Expect(cmd.StepCount).To(Equal(int8(1)))
-		})
-
-	})
-
 	Describe("Eval", func() {
 
 		It("returns keys and values", func() {
diff --git a/export_test.go b/export_test.go
index 3b7965d7..bcc18c45 100644
--- a/export_test.go
+++ b/export_test.go
@@ -20,8 +20,13 @@ func (c *PubSub) ReceiveMessageTimeout(timeout time.Duration) (*Message, error)
 }
 
 func (c *ClusterClient) SlotAddrs(slot int) []string {
+	state, err := c.state()
+	if err != nil {
+		panic(err)
+	}
+
 	var addrs []string
-	for _, n := range c.state().slotNodes(slot) {
+	for _, n := range state.slotNodes(slot) {
 		addrs = append(addrs, n.Client.getAddr())
 	}
 	return addrs
@@ -29,7 +34,12 @@ func (c *ClusterClient) SlotAddrs(slot int) []string {
 
 // SwapSlot swaps a slot's master/slave address for testing MOVED redirects.
 func (c *ClusterClient) SwapSlotNodes(slot int) {
-	nodes := c.state().slots[slot]
+	state, err := c.state()
+	if err != nil {
+		panic(err)
+	}
+
+	nodes := state.slots[slot]
 	if len(nodes) == 2 {
 		nodes[0], nodes[1] = nodes[1], nodes[0]
 	}
diff --git a/internal/error.go b/internal/error.go
index 90f6503a..e1b8be6b 100644
--- a/internal/error.go
+++ b/internal/error.go
@@ -13,10 +13,23 @@ type RedisError string
 func (e RedisError) Error() string { return string(e) }
 
 func IsRetryableError(err error) bool {
-	return IsNetworkError(err) || err.Error() == "ERR max number of clients reached"
+	if IsNetworkError(err) {
+		return true
+	}
+	s := err.Error()
+	if s == "ERR max number of clients reached" {
+		return true
+	}
+	if strings.HasPrefix(s, "LOADING ") {
+		return true
+	}
+	if strings.HasPrefix(s, "CLUSTERDOWN ") {
+		return true
+	}
+	return false
 }
 
-func IsInternalError(err error) bool {
+func IsRedisError(err error) bool {
 	_, ok := err.(RedisError)
 	return ok
 }
@@ -33,7 +46,7 @@ func IsBadConn(err error, allowTimeout bool) bool {
 	if err == nil {
 		return false
 	}
-	if IsInternalError(err) {
+	if IsRedisError(err) {
 		return false
 	}
 	if allowTimeout {
@@ -45,7 +58,7 @@ func IsBadConn(err error, allowTimeout bool) bool {
 }
 
 func IsMovedError(err error) (moved bool, ask bool, addr string) {
-	if !IsInternalError(err) {
+	if !IsRedisError(err) {
 		return
 	}
 
@@ -69,7 +82,3 @@ func IsMovedError(err error) (moved bool, ask bool, addr string) {
 func IsLoadingError(err error) bool {
 	return strings.HasPrefix(err.Error(), "LOADING ")
 }
-
-func IsClusterDownError(err error) bool {
-	return strings.HasPrefix(err.Error(), "CLUSTERDOWN ")
-}
diff --git a/internal/proto/reader.go b/internal/proto/reader.go
index 2159cf63..cd94329d 100644
--- a/internal/proto/reader.go
+++ b/internal/proto/reader.go
@@ -63,7 +63,7 @@ func (p *Reader) ReadLine() ([]byte, error) {
 		return nil, bufio.ErrBufferFull
 	}
 	if len(line) == 0 {
-		return nil, internal.RedisError("redis: reply is empty")
+		return nil, fmt.Errorf("redis: reply is empty")
 	}
 	if isNilReply(line) {
 		return nil, internal.Nil
diff --git a/internal/proto/scan.go b/internal/proto/scan.go
index 3ab40b94..0431a877 100644
--- a/internal/proto/scan.go
+++ b/internal/proto/scan.go
@@ -11,7 +11,7 @@ import (
 func Scan(b []byte, v interface{}) error {
 	switch v := v.(type) {
 	case nil:
-		return internal.RedisError("redis: Scan(nil)")
+		return fmt.Errorf("redis: Scan(nil)")
 	case *string:
 		*v = internal.BytesToString(b)
 		return nil
diff --git a/main_test.go b/main_test.go
index 30f09c61..64f25d99 100644
--- a/main_test.go
+++ b/main_test.go
@@ -3,6 +3,7 @@ package redis_test
 import (
 	"errors"
 	"fmt"
+	"log"
 	"net"
 	"os"
 	"os/exec"
@@ -51,7 +52,7 @@ var cluster = &clusterScenario{
 }
 
 func init() {
-	//redis.SetLogger(log.New(os.Stderr, "redis: ", log.LstdFlags|log.Lshortfile))
+	redis.SetLogger(log.New(os.Stderr, "redis: ", log.LstdFlags|log.Lshortfile))
 }
 
 var _ = BeforeSuite(func() {
diff --git a/pubsub.go b/pubsub.go
index 4a5c65f5..e754a16f 100644
--- a/pubsub.go
+++ b/pubsub.go
@@ -95,7 +95,10 @@ func (c *PubSub) releaseConn(cn *pool.Conn, err error) {
 }
 
 func (c *PubSub) _releaseConn(cn *pool.Conn, err error) {
-	if internal.IsBadConn(err, true) && c.cn == cn {
+	if c.cn != cn {
+		return
+	}
+	if internal.IsBadConn(err, true) {
 		_ = c.closeTheCn()
 	}
 }
diff --git a/redis.go b/redis.go
index b18973cd..db1f39c3 100644
--- a/redis.go
+++ b/redis.go
@@ -197,8 +197,11 @@ type pipelineProcessor func(*pool.Conn, []Cmder) (bool, error)
 
 func (c *baseClient) pipelineExecer(p pipelineProcessor) pipelineExecer {
 	return func(cmds []Cmder) error {
-		var firstErr error
-		for i := 0; i <= c.opt.MaxRetries; i++ {
+		for attempt := 0; attempt <= c.opt.MaxRetries; attempt++ {
+			if attempt > 0 {
+				time.Sleep(c.retryBackoff(attempt))
+			}
+
 			cn, _, err := c.getConn()
 			if err != nil {
 				setCmdsErr(cmds, err)
@@ -206,18 +209,18 @@ func (c *baseClient) pipelineExecer(p pipelineProcessor) pipelineExecer {
 			}
 
 			canRetry, err := p(cn, cmds)
-			c.releaseConn(cn, err)
-			if err == nil {
-				return nil
-			}
-			if firstErr == nil {
-				firstErr = err
+
+			if err == nil || internal.IsRedisError(err) {
+				_ = c.connPool.Put(cn)
+				break
 			}
+			_ = c.connPool.Remove(cn)
+
 			if !canRetry || !internal.IsRetryableError(err) {
 				break
 			}
 		}
-		return firstErr
+		return firstCmdsErr(cmds)
 	}
 }
 
@@ -230,23 +233,17 @@ func (c *baseClient) pipelineProcessCmds(cn *pool.Conn, cmds []Cmder) (bool, err
 
 	// Set read timeout for all commands.
 	cn.SetReadTimeout(c.opt.ReadTimeout)
-	return pipelineReadCmds(cn, cmds)
+	return true, pipelineReadCmds(cn, cmds)
 }
 
-func pipelineReadCmds(cn *pool.Conn, cmds []Cmder) (retry bool, firstErr error) {
-	for i, cmd := range cmds {
+func pipelineReadCmds(cn *pool.Conn, cmds []Cmder) error {
+	for _, cmd := range cmds {
 		err := cmd.readReply(cn)
-		if err == nil {
-			continue
-		}
-		if i == 0 {
-			retry = true
-		}
-		if firstErr == nil {
-			firstErr = err
+		if err != nil && !internal.IsRedisError(err) {
+			return err
 		}
 	}
-	return
+	return nil
 }
 
 func (c *baseClient) txPipelineProcessCmds(cn *pool.Conn, cmds []Cmder) (bool, error) {
@@ -260,11 +257,11 @@ func (c *baseClient) txPipelineProcessCmds(cn *pool.Conn, cmds []Cmder) (bool, e
 	cn.SetReadTimeout(c.opt.ReadTimeout)
 
 	if err := c.txPipelineReadQueued(cn, cmds); err != nil {
+		setCmdsErr(cmds, err)
 		return false, err
 	}
 
-	_, err := pipelineReadCmds(cn, cmds)
-	return false, err
+	return false, pipelineReadCmds(cn, cmds)
 }
 
 func txPipelineWriteMulti(cn *pool.Conn, cmds []Cmder) error {
@@ -276,21 +273,16 @@ func txPipelineWriteMulti(cn *pool.Conn, cmds []Cmder) error {
 }
 
 func (c *baseClient) txPipelineReadQueued(cn *pool.Conn, cmds []Cmder) error {
-	var firstErr error
-
 	// Parse queued replies.
 	var statusCmd StatusCmd
-	if err := statusCmd.readReply(cn); err != nil && firstErr == nil {
-		firstErr = err
+	if err := statusCmd.readReply(cn); err != nil {
+		return err
 	}
 
-	for _, cmd := range cmds {
+	for _ = range cmds {
 		err := statusCmd.readReply(cn)
-		if err != nil {
-			cmd.setErr(err)
-			if firstErr == nil {
-				firstErr = err
-			}
+		if err != nil && !internal.IsRedisError(err) {
+			return err
 		}
 	}
 
diff --git a/ring.go b/ring.go
index 72d52bf7..a9314fb5 100644
--- a/ring.go
+++ b/ring.go
@@ -34,7 +34,9 @@ type RingOptions struct {
 	DB       int
 	Password string
 
-	MaxRetries int
+	MaxRetries      int
+	MinRetryBackoff time.Duration
+	MaxRetryBackoff time.Duration
 
 	DialTimeout  time.Duration
 	ReadTimeout  time.Duration
@@ -50,6 +52,19 @@ func (opt *RingOptions) init() {
 	if opt.HeartbeatFrequency == 0 {
 		opt.HeartbeatFrequency = 500 * time.Millisecond
 	}
+
+	switch opt.MinRetryBackoff {
+	case -1:
+		opt.MinRetryBackoff = 0
+	case 0:
+		opt.MinRetryBackoff = 8 * time.Millisecond
+	}
+	switch opt.MaxRetryBackoff {
+	case -1:
+		opt.MaxRetryBackoff = 0
+	case 0:
+		opt.MaxRetryBackoff = 512 * time.Millisecond
+	}
 }
 
 func (opt *RingOptions) clientOptions() *Options {
@@ -165,6 +180,10 @@ func (c *Ring) Options() *RingOptions {
 	return c.opt
 }
 
+func (c *Ring) retryBackoff(attempt int) time.Duration {
+	return internal.RetryBackoff(attempt, c.opt.MinRetryBackoff, c.opt.MaxRetryBackoff)
+}
+
 // PoolStats returns accumulated connection pool stats.
 func (c *Ring) PoolStats() *PoolStats {
 	var acc PoolStats
@@ -241,6 +260,9 @@ func (c *Ring) ForEachShard(fn func(client *Client) error) error {
 
 func (c *Ring) cmdInfo(name string) *CommandInfo {
 	err := c.cmdsInfoOnce.Do(func() error {
+		c.mu.RLock()
+		defer c.mu.RUnlock()
+
 		var firstErr error
 		for _, shard := range c.shards {
 			cmdsInfo, err := shard.Client.Command().Result()
@@ -257,7 +279,11 @@ func (c *Ring) cmdInfo(name string) *CommandInfo {
 	if err != nil {
 		return nil
 	}
-	return c.cmdsInfo[name]
+	info := c.cmdsInfo[name]
+	if info == nil {
+		internal.Logf("info for cmd=%s not found", name)
+	}
+	return info
 }
 
 func (c *Ring) addClient(name string, cl *Client) {
@@ -399,7 +425,7 @@ func (c *Ring) Pipelined(fn func(Pipeliner) error) ([]Cmder, error) {
 	return c.Pipeline().pipelined(fn)
 }
 
-func (c *Ring) pipelineExec(cmds []Cmder) (firstErr error) {
+func (c *Ring) pipelineExec(cmds []Cmder) error {
 	cmdsMap := make(map[string][]Cmder)
 	for _, cmd := range cmds {
 		cmdInfo := c.cmdInfo(cmd.Name())
@@ -410,36 +436,33 @@ func (c *Ring) pipelineExec(cmds []Cmder) (firstErr error) {
 		cmdsMap[name] = append(cmdsMap[name], cmd)
 	}
 
-	for i := 0; i <= c.opt.MaxRetries; i++ {
+	for attempt := 0; attempt <= c.opt.MaxRetries; attempt++ {
+		if attempt > 0 {
+			time.Sleep(c.retryBackoff(attempt))
+		}
+
 		var failedCmdsMap map[string][]Cmder
 
 		for name, cmds := range cmdsMap {
 			shard, err := c.shardByName(name)
 			if err != nil {
 				setCmdsErr(cmds, err)
-				if firstErr == nil {
-					firstErr = err
-				}
 				continue
 			}
 
 			cn, _, err := shard.Client.getConn()
 			if err != nil {
 				setCmdsErr(cmds, err)
-				if firstErr == nil {
-					firstErr = err
-				}
 				continue
 			}
 
 			canRetry, err := shard.Client.pipelineProcessCmds(cn, cmds)
-			shard.Client.releaseConn(cn, err)
-			if err == nil {
+			if err == nil || internal.IsRedisError(err) {
+				_ = shard.Client.connPool.Put(cn)
 				continue
 			}
-			if firstErr == nil {
-				firstErr = err
-			}
+			_ = shard.Client.connPool.Remove(cn)
+
 			if canRetry && internal.IsRetryableError(err) {
 				if failedCmdsMap == nil {
 					failedCmdsMap = make(map[string][]Cmder)
@@ -454,5 +477,5 @@ func (c *Ring) pipelineExec(cmds []Cmder) (firstErr error) {
 		cmdsMap = failedCmdsMap
 	}
 
-	return firstErr
+	return firstCmdsErr(cmds)
 }
diff --git a/tx.go b/tx.go
index 5ef89619..93c295d9 100644
--- a/tx.go
+++ b/tx.go
@@ -36,11 +36,10 @@ func (c *Client) Watch(fn func(*Tx) error, keys ...string) error {
 			return err
 		}
 	}
-	firstErr := fn(tx)
-	if err := tx.Close(); err != nil && firstErr == nil {
-		firstErr = err
-	}
-	return firstErr
+
+	err := fn(tx)
+	_ = tx.Close()
+	return err
 }
 
 // close closes the transaction, releasing any open resources.
@@ -53,7 +52,7 @@ func (c *Tx) Close() error {
 // of a transaction.
 func (c *Tx) Watch(keys ...string) *StatusCmd {
 	args := make([]interface{}, 1+len(keys))
-	args[0] = "WATCH"
+	args[0] = "watch"
 	for i, key := range keys {
 		args[1+i] = key
 	}
@@ -65,7 +64,7 @@ func (c *Tx) Watch(keys ...string) *StatusCmd {
 // Unwatch flushes all the previously watched keys for a transaction.
 func (c *Tx) Unwatch(keys ...string) *StatusCmd {
 	args := make([]interface{}, 1+len(keys))
-	args[0] = "UNWATCH"
+	args[0] = "unwatch"
 	for i, key := range keys {
 		args[1+i] = key
 	}