[release-19.0] Dedicated poolDialer logic for VTOrc, throttler (vites…

…sio#15562) (vitessio#15567) Signed-off-by: Shlomi Noach <[email protected]> Co-authored-by: vitess-bot[bot] <108069721+vitess-bot[bot]@users.noreply.github.com> Co-authored-by: Shlomi Noach <[email protected]>
GrahamCampbell · Mar 26, 2024 · 5950765 · 5950765
1 parent 8ee3294
commit 5950765
Show file tree

Hide file tree

Showing 4 changed files with 337 additions and 13 deletions.
diff --git a/go/test/endtoend/vtorc/general/vtorc_test.go b/go/test/endtoend/vtorc/general/vtorc_test.go
@@ -495,3 +495,76 @@ func TestDurabilityPolicySetLater(t *testing.T) {
 	assert.NotNil(t, primary, "should have elected a primary")
 	utils.CheckReplication(t, newCluster, primary, shard0.Vttablets, 10*time.Second)
 }
+
+func TestFullStatusConnectionPooling(t *testing.T) {
+	defer utils.PrintVTOrcLogsOnFailure(t, clusterInfo.ClusterInstance)
+	defer cluster.PanicHandler(t)
+	utils.SetupVttabletsAndVTOrcs(t, clusterInfo, 4, 0, []string{
+		"--tablet_manager_grpc_concurrency=1",
+	}, cluster.VTOrcConfiguration{
+		PreventCrossDataCenterPrimaryFailover: true,
+	}, 1, "")
+	keyspace := &clusterInfo.ClusterInstance.Keyspaces[0]
+	shard0 := &keyspace.Shards[0]
+	vtorc := clusterInfo.ClusterInstance.VTOrcProcesses[0]
+
+	// find primary from topo
+	curPrimary := utils.ShardPrimaryTablet(t, clusterInfo, keyspace, shard0)
+	assert.NotNil(t, curPrimary, "should have elected a primary")
+	vtOrcProcess := clusterInfo.ClusterInstance.VTOrcProcesses[0]
+	utils.WaitForSuccessfulRecoveryCount(t, vtOrcProcess, logic.ElectNewPrimaryRecoveryName, 1)
+	utils.WaitForSuccessfulPRSCount(t, vtOrcProcess, keyspace.Name, shard0.Name, 1)
+
+	// Kill the current primary.
+	_ = curPrimary.VttabletProcess.Kill()
+
+	// Wait until VTOrc notices some problems
+	status, resp := utils.MakeAPICallRetry(t, vtorc, "/api/replication-analysis", func(_ int, response string) bool {
+		return response == "null"
+	})
+	assert.Equal(t, 200, status)
+	assert.Contains(t, resp, "UnreachablePrimary")
+
+	time.Sleep(1 * time.Minute)
+
+	// Change the primaries ports and restart it.
+	curPrimary.VttabletProcess.Port = clusterInfo.ClusterInstance.GetAndReservePort()
+	curPrimary.VttabletProcess.GrpcPort = clusterInfo.ClusterInstance.GetAndReservePort()
+	err := curPrimary.VttabletProcess.Setup()
+	require.NoError(t, err)
+
+	// See that VTOrc eventually reports no errors.
+	// Wait until there are no problems and the api endpoint returns null
+	status, resp = utils.MakeAPICallRetry(t, vtorc, "/api/replication-analysis", func(_ int, response string) bool {
+		return response != "null"
+	})
+	assert.Equal(t, 200, status)
+	assert.Equal(t, "null", resp)
+
+	// REPEATED
+	// Kill the current primary.
+	_ = curPrimary.VttabletProcess.Kill()
+
+	// Wait until VTOrc notices some problems
+	status, resp = utils.MakeAPICallRetry(t, vtorc, "/api/replication-analysis", func(_ int, response string) bool {
+		return response == "null"
+	})
+	assert.Equal(t, 200, status)
+	assert.Contains(t, resp, "UnreachablePrimary")
+
+	time.Sleep(1 * time.Minute)
+
+	// Change the primaries ports back to original and restart it.
+	curPrimary.VttabletProcess.Port = curPrimary.HTTPPort
+	curPrimary.VttabletProcess.GrpcPort = curPrimary.GrpcPort
+	err = curPrimary.VttabletProcess.Setup()
+	require.NoError(t, err)
+
+	// See that VTOrc eventually reports no errors.
+	// Wait until there are no problems and the api endpoint returns null
+	status, resp = utils.MakeAPICallRetry(t, vtorc, "/api/replication-analysis", func(_ int, response string) bool {
+		return response != "null"
+	})
+	assert.Equal(t, 200, status)
+	assert.Equal(t, "null", resp)
+}
diff --git a/go/test/endtoend/vtorc/utils/utils.go b/go/test/endtoend/vtorc/utils/utils.go
@@ -756,7 +756,7 @@ func MakeAPICall(t *testing.T, vtorc *cluster.VTOrcProcess, url string) (status
 // The function provided takes in the status and response and returns if we should continue to retry or not
 func MakeAPICallRetry(t *testing.T, vtorc *cluster.VTOrcProcess, url string, retry func(int, string) bool) (status int, response string) {
 	t.Helper()
-	timeout := time.After(10 * time.Second)
+	timeout := time.After(30 * time.Second)
 	for {
 		select {
 		case <-timeout:

diff --git a/go/vt/vttablet/grpctmclient/client.go b/go/vt/vttablet/grpctmclient/client.go
@@ -45,6 +45,15 @@ import (
 	topodatapb "vitess.io/vitess/go/vt/proto/topodata"
 )
 
+type DialPoolGroup int
+
+const (
+	dialPoolGroupThrottler DialPoolGroup = iota
+	dialPoolGroupVTOrc
+)
+
+type invalidatorFunc func()
+
 var (
 	concurrency = 8
 	cert        string
@@ -92,14 +101,17 @@ type tmc struct {
 	client tabletmanagerservicepb.TabletManagerClient
 }
 
+type addrTmcMap map[string]*tmc
+
 // grpcClient implements both dialer and poolDialer.
 type grpcClient struct {
 	// This cache of connections is to maximize QPS for ExecuteFetchAs{Dba,App} and
 	// CheckThrottler. Note we'll keep the clients open and close them upon Close() only.
 	// But that's OK because usually the tasks that use them are one-purpose only.
 	// The map is protected by the mutex.
-	mu           sync.Mutex
-	rpcClientMap map[string]chan *tmc
+	mu             sync.Mutex
+	rpcClientMap   map[string]chan *tmc
+	rpcDialPoolMap map[DialPoolGroup]addrTmcMap
 }
 
 type dialer interface {
@@ -109,6 +121,7 @@ type dialer interface {
 
 type poolDialer interface {
 	dialPool(ctx context.Context, tablet *topodatapb.Tablet) (tabletmanagerservicepb.TabletManagerClient, error)
+	dialDedicatedPool(ctx context.Context, dialPoolGroup DialPoolGroup, tablet *topodatapb.Tablet) (tabletmanagerservicepb.TabletManagerClient, invalidatorFunc, error)
 }
 
 // Client implements tmclient.TabletManagerClient.
@@ -152,6 +165,17 @@ func (client *grpcClient) dial(ctx context.Context, tablet *topodatapb.Tablet) (
 	return tabletmanagerservicepb.NewTabletManagerClient(cc), cc, nil
 }
 
+func (client *grpcClient) createTmc(addr string, opt grpc.DialOption) (*tmc, error) {
+	cc, err := grpcclient.Dial(addr, grpcclient.FailFast(false), opt)
+	if err != nil {
+		return nil, err
+	}
+	return &tmc{
+		cc:     cc,
+		client: tabletmanagerservicepb.NewTabletManagerClient(cc),
+	}, nil
+}
+
 func (client *grpcClient) dialPool(ctx context.Context, tablet *topodatapb.Tablet) (tabletmanagerservicepb.TabletManagerClient, error) {
 	addr := netutil.JoinHostPort(tablet.Hostname, int32(tablet.PortMap["grpc"]))
 	opt, err := grpcclient.SecureDialOption(cert, key, ca, crl, name)
@@ -170,14 +194,11 @@ func (client *grpcClient) dialPool(ctx context.Context, tablet *topodatapb.Table
 		client.mu.Unlock()
 
 		for i := 0; i < cap(c); i++ {
-			cc, err := grpcclient.Dial(addr, grpcclient.FailFast(false), opt)
+			tm, err := client.createTmc(addr, opt)
 			if err != nil {
 				return nil, err
 			}
-			c <- &tmc{
-				cc:     cc,
-				client: tabletmanagerservicepb.NewTabletManagerClient(cc),
-			}
+			c <- tm
 		}
 	} else {
 		client.mu.Unlock()
@@ -188,6 +209,38 @@ func (client *grpcClient) dialPool(ctx context.Context, tablet *topodatapb.Table
 	return result.client, nil
 }
 
+func (client *grpcClient) dialDedicatedPool(ctx context.Context, dialPoolGroup DialPoolGroup, tablet *topodatapb.Tablet) (tabletmanagerservicepb.TabletManagerClient, invalidatorFunc, error) {
+	addr := netutil.JoinHostPort(tablet.Hostname, int32(tablet.PortMap["grpc"]))
+	opt, err := grpcclient.SecureDialOption(cert, key, ca, crl, name)
+	if err != nil {
+		return nil, nil, err
+	}
+
+	client.mu.Lock()
+	defer client.mu.Unlock()
+	if client.rpcDialPoolMap == nil {
+		client.rpcDialPoolMap = make(map[DialPoolGroup]addrTmcMap)
+	}
+	if _, ok := client.rpcDialPoolMap[dialPoolGroup]; !ok {
+		client.rpcDialPoolMap[dialPoolGroup] = make(addrTmcMap)
+	}
+	m := client.rpcDialPoolMap[dialPoolGroup]
+	if _, ok := m[addr]; !ok {
+		tm, err := client.createTmc(addr, opt)
+		if err != nil {
+			return nil, nil, err
+		}
+		m[addr] = tm
+	}
+	invalidator := func() {
+		client.mu.Lock()
+		defer client.mu.Unlock()
+		m[addr].cc.Close()
+		delete(m, addr)
+	}
+	return m[addr].client, invalidator, nil
+}
+
 // Close is part of the tmclient.TabletManagerClient interface.
 func (client *grpcClient) Close() {
 	client.mu.Lock()
@@ -570,13 +623,30 @@ func (client *Client) ReplicationStatus(ctx context.Context, tablet *topodatapb.
 
 // FullStatus is part of the tmclient.TabletManagerClient interface.
 func (client *Client) FullStatus(ctx context.Context, tablet *topodatapb.Tablet) (*replicationdatapb.FullStatus, error) {
-	c, closer, err := client.dialer.dial(ctx, tablet)
-	if err != nil {
-		return nil, err
+	var c tabletmanagerservicepb.TabletManagerClient
+	var invalidator invalidatorFunc
+	var err error
+	if poolDialer, ok := client.dialer.(poolDialer); ok {
+		c, invalidator, err = poolDialer.dialDedicatedPool(ctx, dialPoolGroupVTOrc, tablet)
+		if err != nil {
+			return nil, err
+		}
 	}
-	defer closer.Close()
+
+	if c == nil {
+		var closer io.Closer
+		c, closer, err = client.dialer.dial(ctx, tablet)
+		if err != nil {
+			return nil, err
+		}
+		defer closer.Close()
+	}
+
 	response, err := c.FullStatus(ctx, &tabletmanagerdatapb.FullStatusRequest{})
 	if err != nil {
+		if invalidator != nil {
+			invalidator()
+		}
 		return nil, err
 	}
 	return response.Status, nil
@@ -1009,9 +1079,10 @@ func (client *Client) Backup(ctx context.Context, tablet *topodatapb.Tablet, req
 // and dialing the other tablet every time is not practical.
 func (client *Client) CheckThrottler(ctx context.Context, tablet *topodatapb.Tablet, req *tabletmanagerdatapb.CheckThrottlerRequest) (*tabletmanagerdatapb.CheckThrottlerResponse, error) {
 	var c tabletmanagerservicepb.TabletManagerClient
+	var invalidator invalidatorFunc
 	var err error
 	if poolDialer, ok := client.dialer.(poolDialer); ok {
-		c, err = poolDialer.dialPool(ctx, tablet)
+		c, invalidator, err = poolDialer.dialDedicatedPool(ctx, dialPoolGroupThrottler, tablet)
 		if err != nil {
 			return nil, err
 		}
@@ -1028,6 +1099,9 @@ func (client *Client) CheckThrottler(ctx context.Context, tablet *topodatapb.Tab
 
 	response, err := c.CheckThrottler(ctx, req)
 	if err != nil {
+		if invalidator != nil {
+			invalidator()
+		}
 		return nil, err
 	}
 	return response, nil