tikvclient: fix a bug that double close channels.

metrics/tikvclient.go

@@ -180,13 +180,13 @@ var (
 		})
 
 	// TiKVPendingBatchRequests indicates the number of requests pending in the batch channel.
-	TiKVPendingBatchRequests = prometheus.NewGauge(
+	TiKVPendingBatchRequests = prometheus.NewGaugeVec(
 		prometheus.GaugeOpts{
 			Namespace: "tidb",
 			Subsystem: "tikvclient",
 			Name:      "pending_batch_requests",
 			Help:      "Pending batch requests",
-		})
+		}, []string{"store"})
 
 	TiKVBatchWaitDuration = prometheus.NewHistogram(
 		prometheus.HistogramOpts{

store/tikv/client.go

@@ -25,6 +25,7 @@ import (
 
 	"github.com/grpc-ecosystem/go-grpc-middleware/tracing/opentracing"
 	"github.com/pingcap/errors"
+	"github.com/pingcap/failpoint"
 	"github.com/pingcap/kvproto/pkg/coprocessor"
 	"github.com/pingcap/kvproto/pkg/debugpb"
 	"github.com/pingcap/kvproto/pkg/tikvpb"
@@ -33,6 +34,7 @@ import (
 	"github.com/pingcap/tidb/metrics"
 	"github.com/pingcap/tidb/store/tikv/tikvrpc"
 	"github.com/pingcap/tidb/util/logutil"
+	"github.com/prometheus/client_golang/prometheus"
 	"go.uber.org/zap"
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/credentials"
@@ -83,6 +85,8 @@ type connArray struct {
 	idleNotify *uint32
 	idle       bool
 	idleDetect *time.Timer
+
+	pendingRequests prometheus.Gauge
 }
 
 type batchCommandsClient struct {
@@ -105,7 +109,35 @@ func (c *batchCommandsClient) isStopped() bool {
 	return atomic.LoadInt32(&c.closed) != 0
 }
 
+func (c *batchCommandsClient) send(request *tikvpb.BatchCommandsRequest, entries []*batchCommandsEntry) {
+	// Use the lock to protect the stream client won't be replaced by RecvLoop,
+	// and new added request won't be removed by `failPendingRequests`.
+	c.clientLock.Lock()
+	defer c.clientLock.Unlock()
+	for i, requestID := range request.RequestIds {
+		c.batched.Store(requestID, entries[i])
+	}
+	if err := c.client.Send(request); err != nil {
+		logutil.BgLogger().Error(
+			"batch commands send error",
+			zap.String("target", c.target),
+			zap.Error(err),
+		)
+		c.failPendingRequests(err)
+	}
+}
+
+func (c *batchCommandsClient) recv() (*tikvpb.BatchCommandsResponse, error) {
+	failpoint.Inject("gotErrorInRecvLoop", func(_ failpoint.Value) (*tikvpb.BatchCommandsResponse, error) {
+		return nil, errors.New("injected error in batchRecvLoop")
+	})
+	// When `conn.Close()` is called, `client.Recv()` will return an error.
+	return c.client.Recv()
+}
+
+// `failPendingRequests` must be called in locked contexts in order to avoid double closing channels.
 func (c *batchCommandsClient) failPendingRequests(err error) {
+	failpoint.Inject("panicInFailPendingRequests", nil)
 	c.batched.Range(func(key, value interface{}) bool {
 		id, _ := key.(uint64)
 		entry, _ := value.(*batchCommandsEntry)
@@ -116,6 +148,31 @@ func (c *batchCommandsClient) failPendingRequests(err error) {
 	})
 }
 
+func (c *batchCommandsClient) reCreateStreamingClient(err error) bool {
+	// Hold the lock to forbid batchSendLoop using the old client.
+	c.clientLock.Lock()
+	defer c.clientLock.Unlock()
+	c.failPendingRequests(err) // fail all pending requests.
+
+	// Re-establish a application layer stream. TCP layer is handled by gRPC.
+	tikvClient := tikvpb.NewTikvClient(c.conn)
+	streamClient, err := tikvClient.BatchCommands(context.TODO())
+	if err == nil {
+		logutil.BgLogger().Info(
+			"batchRecvLoop re-create streaming success",
+			zap.String("target", c.target),
+		)
+		c.client = streamClient
+		return true
+	}
+	logutil.BgLogger().Error(
+		"batchRecvLoop re-create streaming fail",
+		zap.String("target", c.target),
+		zap.Error(err),
+	)
+	return false
+}
+
 func (c *batchCommandsClient) batchRecvLoop(cfg config.TiKVClient) {
 	defer func() {
 		if r := recover(); r != nil {
@@ -129,8 +186,7 @@ func (c *batchCommandsClient) batchRecvLoop(cfg config.TiKVClient) {
 	}()
 
 	for {
-		// When `conn.Close()` is called, `client.Recv()` will return an error.
-		resp, err := c.client.Recv()
+		resp, err := c.recv()
 		if err != nil {
 			now := time.Now()
 			for { // try to re-create the streaming in the loop.
@@ -143,28 +199,10 @@ func (c *batchCommandsClient) batchRecvLoop(cfg config.TiKVClient) {
 					zap.Error(err),
 				)
 
-				// Hold the lock to forbid batchSendLoop using the old client.
-				c.clientLock.Lock()
-				c.failPendingRequests(err) // fail all pending requests.
-
-				// Re-establish a application layer stream. TCP layer is handled by gRPC.
-				tikvClient := tikvpb.NewTikvClient(c.conn)
-				streamClient, err := tikvClient.BatchCommands(context.TODO())
-				c.clientLock.Unlock()
-
-				if err == nil {
-					logutil.BgLogger().Info(
-						"batchRecvLoop re-create streaming success",
-						zap.String("target", c.target),
-					)
-					c.client = streamClient
+				if c.reCreateStreamingClient(err) {
 					break
 				}
-				logutil.BgLogger().Error(
-					"batchRecvLoop re-create streaming fail",
-					zap.String("target", c.target),
-					zap.Error(err),
-				)
+
 				// TODO: Use a more smart backoff strategy.
 				time.Sleep(time.Second)
 			}
@@ -220,6 +258,7 @@ func newConnArray(maxSize uint, addr string, security config.Security, idleNotif
 
 func (a *connArray) Init(addr string, security config.Security) error {
 	a.target = addr
+	a.pendingRequests = metrics.TiKVPendingBatchRequests.WithLabelValues(a.target)
 
 	opt := grpc.WithInsecure()
 	if len(security.ClusterSSLCA) != 0 {
@@ -450,7 +489,7 @@ func (a *connArray) batchSendLoop(cfg config.TiKVClient) {
 		requests = requests[:0]
 		requestIDs = requestIDs[:0]
 
-		metrics.TiKVPendingBatchRequests.Set(float64(len(a.batchCommandsCh)))
+		a.pendingRequests.Set(float64(len(a.batchCommandsCh)))
 		a.fetchAllPendingRequests(int(cfg.MaxBatchSize), &entries, &requests)
 
 		if len(entries) < int(cfg.MaxBatchSize) && cfg.MaxBatchWaitTime > 0 {
@@ -484,27 +523,12 @@ func (a *connArray) batchSendLoop(cfg config.TiKVClient) {
 			requestIDs = append(requestIDs, requestID)
 		}
 
-		request := &tikvpb.BatchCommandsRequest{
+		req := &tikvpb.BatchCommandsRequest{
 			Requests:   requests,
 			RequestIds: requestIDs,
 		}
 
-		// Use the lock to protect the stream client won't be replaced by RecvLoop,
-		// and new added request won't be removed by `failPendingRequests`.
-		batchCommandsClient.clientLock.Lock()
-		for i, requestID := range request.RequestIds {
-			batchCommandsClient.batched.Store(requestID, entries[i])
-		}
-		err := batchCommandsClient.client.Send(request)
-		batchCommandsClient.clientLock.Unlock()
-		if err != nil {
-			logutil.BgLogger().Error(
-				"batch commands send error",
-				zap.String("target", a.target),
-				zap.Error(err),
-			)
-			batchCommandsClient.failPendingRequests(err)
-		}
+		batchCommandsClient.send(req, entries)
 	}
 }
 
@@ -547,6 +571,11 @@ func newRPCClient(security config.Security) *rpcClient {
 	}
 }
 
+// NewTestRPCClient is for some external tests.
+func NewTestRPCClient() Client {
+	return newRPCClient(config.Security{})
+}
+
 func (c *rpcClient) getConnArray(addr string) (*connArray, error) {
 	c.RLock()
 	if c.isClosed {

store/tikv/client_fail_test.go

@@ -0,0 +1,56 @@
+// Copyright 2016 PingCAP, Inc.
+//
+// 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,
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package tikv
+
+import (
+	"context"
+	"fmt"
+	"time"
+
+	. "github.com/pingcap/check"
+	"github.com/pingcap/failpoint"
+	"github.com/pingcap/kvproto/pkg/kvrpcpb"
+	"github.com/pingcap/tidb/config"
+	"github.com/pingcap/tidb/store/tikv/tikvrpc"
+)
+
+func (s *testClientSuite) TestPanicInRecvLoop(c *C) {
+	c.Assert(failpoint.Enable("github.com/pingcap/tidb/store/tikv/panicInFailPendingRequests", `panic`), IsNil)
+	c.Assert(failpoint.Enable("github.com/pingcap/tidb/store/tikv/gotErrorInRecvLoop", `return("0")`), IsNil)
+
+	server, port := startMockTikvService()
+	c.Assert(port > 0, IsTrue)
+
+	grpcConnectionCount := config.GetGlobalConfig().TiKVClient.GrpcConnectionCount
+	config.GetGlobalConfig().TiKVClient.GrpcConnectionCount = 1
+	addr := fmt.Sprintf("%s:%d", "127.0.0.1", port)
+	rpcClient := newRPCClient(config.Security{})
+
+	// Start batchRecvLoop, and it should panic in `failPendingRequests`.
+	_, err := rpcClient.getConnArray(addr)
+	c.Assert(err, IsNil)
+
+	time.Sleep(time.Second)
+	c.Assert(failpoint.Disable("github.com/pingcap/tidb/store/tikv/panicInFailPendingRequests"), IsNil)
+	c.Assert(failpoint.Disable("github.com/pingcap/tidb/store/tikv/gotErrorInRecvLoop"), IsNil)
+
+	req := &tikvrpc.Request{
+		Type: tikvrpc.CmdGet,
+		Get:  &kvrpcpb.GetRequest{},
+	}
+	_, err = rpcClient.SendRequest(context.Background(), addr, req, time.Second)
+	c.Assert(err, IsNil)
+	server.Stop()
+	config.GetGlobalConfig().TiKVClient.GrpcConnectionCount = grpcConnectionCount
+}

store/tikv/mock_tikv_service.go

@@ -0,0 +1,71 @@
+package tikv
+
+import (
+	"fmt"
+	"net"
+	"time"
+
+	"github.com/pingcap/kvproto/pkg/kvrpcpb"
+	"github.com/pingcap/kvproto/pkg/tikvpb"
+	"github.com/pingcap/tidb/util/logutil"
+	"go.uber.org/zap"
+	"google.golang.org/grpc"
+)
+
+type server struct {
+	tikvpb.TikvServer
+}
+
+func (s *server) BatchCommands(ss tikvpb.Tikv_BatchCommandsServer) error {
+	for {
+		req, err := ss.Recv()
+		if err != nil {
+			logutil.BgLogger().Error("batch commands receive fail", zap.Error(err))
+			return err
+		}
+
+		responses := make([]*tikvpb.BatchCommandsResponse_Response, 0, len(req.GetRequestIds()))
+		for i := 0; i < len(req.GetRequestIds()); i++ {
+			responses = append(responses, &tikvpb.BatchCommandsResponse_Response{
+				Cmd: &tikvpb.BatchCommandsResponse_Response_Get{
+					Get: &kvrpcpb.GetResponse{
+						Value: []byte{'a', 'b', 'c'},
+					},
+				},
+			})
+		}
+
+		err = ss.Send(&tikvpb.BatchCommandsResponse{
+			Responses:  responses,
+			RequestIds: req.GetRequestIds(),
+		})
+		if err != nil {
+			logutil.BgLogger().Error("batch commands send fail", zap.Error(err))
+			return err
+		}
+	}
+}
+
+// Try to start a gRPC server and retrun the server instance and binded port.
+func startMockTikvService() (*grpc.Server, int) {
+	for port := 40000; port < 50000; port++ {
+		lis, err := net.Listen("tcp", fmt.Sprintf("%s:%d", "127.0.0.1", port))
+		if err != nil {
+			logutil.BgLogger().Error("can't listen", zap.Error(err))
+			continue
+		}
+		s := grpc.NewServer(grpc.ConnectionTimeout(time.Minute))
+		tikvpb.RegisterTikvServer(s, &server{})
+		go func() {
+			if err = s.Serve(lis); err != nil {
+				logutil.BgLogger().Error(
+					"can't serve gRPC requests",
+					zap.Error(err),
+				)
+			}
+		}()
+		return s, port
+	}
+	logutil.BgLogger().Error("can't start mock tikv service because no available ports")
+	return nil, -1
+}