[release-19.0] SHOW VITESS_REPLICATION_STATUS: Only use replication tracker when it's enabled (#15348)

go/test/endtoend/tabletgateway/vtgate_test.go

@@ -28,15 +28,14 @@
 	"testing"
 	"time"
 
-	"vitess.io/vitess/go/test/endtoend/utils"
-	"vitess.io/vitess/go/vt/proto/topodata"
-
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
 	"vitess.io/vitess/go/mysql"
-
 	"vitess.io/vitess/go/test/endtoend/cluster"
+	"vitess.io/vitess/go/test/endtoend/utils"
+	vtorcutils "vitess.io/vitess/go/test/endtoend/vtorc/utils"
+	"vitess.io/vitess/go/vt/proto/topodata"
 )
 
 func TestVtgateHealthCheck(t *testing.T) {
@@ -59,7 +58,7 @@
 	time.Sleep(2 * time.Second)
 	verifyVtgateVariables(t, clusterInstance.VtgateProcess.VerifyURL)
 	ctx := context.Background()
-	conn, err := mysql.Connect(ctx, &vtParams)
+	conn, err := mysql.Connect(ctx, &vtParams) // VTGate
 	require.NoError(t, err)
 	defer conn.Close()
 
@@ -68,6 +67,38 @@
 	expectNumRows := 2
 	numRows := len(qr.Rows)
 	assert.Equal(t, expectNumRows, numRows, fmt.Sprintf("wrong number of results from show vitess_replication_status. Expected %d, got %d", expectNumRows, numRows))
+
+	// Disable VTOrc(s) recoveries so that it doesn't immediately repair/restart replication.
+	for _, vtorcProcess := range clusterInstance.VTOrcProcesses {
+		vtorcutils.DisableGlobalRecoveries(t, vtorcProcess)
+	}
+	// Re-enable recoveries afterward as the cluster is re-used.
+	defer func() {
+		for _, vtorcProcess := range clusterInstance.VTOrcProcesses {
+			vtorcutils.EnableGlobalRecoveries(t, vtorcProcess)
+		}
+	}()
+	// Stop replication on the non-PRIMARY tablets.
+	_, err = clusterInstance.VtctldClientProcess.ExecuteCommandWithOutput("ExecuteFetchAsDBA", clusterInstance.Keyspaces[0].Shards[0].Replica().Alias, "stop slave")
+	require.NoError(t, err)
+	_, err = clusterInstance.VtctldClientProcess.ExecuteCommandWithOutput("ExecuteFetchAsDBA", clusterInstance.Keyspaces[0].Shards[0].Rdonly().Alias, "stop slave")
+	require.NoError(t, err)
+	// Restart replication afterward as the cluster is re-used.
+	defer func() {
+		_, err = clusterInstance.VtctldClientProcess.ExecuteCommandWithOutput("ExecuteFetchAsDBA", clusterInstance.Keyspaces[0].Shards[0].Replica().Alias, "start slave")
+		require.NoError(t, err)
+		_, err = clusterInstance.VtctldClientProcess.ExecuteCommandWithOutput("ExecuteFetchAsDBA", clusterInstance.Keyspaces[0].Shards[0].Rdonly().Alias, "start slave")
+		require.NoError(t, err)
+	}()
+	time.Sleep(2 * time.Second) // Build up some replication lag
+	res, err := conn.ExecuteFetch("show vitess_replication_status", 2, false)
+	require.NoError(t, err)
+	expectNumRows = 2
+	numRows = len(qr.Rows)
+	assert.Equal(t, expectNumRows, numRows, fmt.Sprintf("wrong number of results from show vitess_replication_status, expected %d, got %d", expectNumRows, numRows))
+	rawLag := res.Named().Rows[0]["ReplicationLag"] // Let's just look at the first row
+	lagInt, _ := rawLag.ToInt64()                   // Don't check the error as the value could be "NULL"
+	assert.True(t, rawLag.IsNull() || lagInt > 0, "replication lag should be NULL or greater than 0 but was: %s", rawLag.ToString())
 }
 
 func TestVtgateReplicationStatusCheckWithTabletTypeChange(t *testing.T) {
@@ -90,6 +121,11 @@
 	rdOnlyTablet := clusterInstance.Keyspaces[0].Shards[0].Rdonly()
 	err = clusterInstance.VtctlclientChangeTabletType(rdOnlyTablet, topodata.TabletType_SPARE)
 	require.NoError(t, err)
+	// Change it back to RDONLY afterward as the cluster is re-used.
+	defer func() {
+		err = clusterInstance.VtctlclientChangeTabletType(rdOnlyTablet, topodata.TabletType_RDONLY)
+		require.NoError(t, err)
+	}()
 
 	// Only returns rows for REPLICA and RDONLY tablets -- so should be 1 of them since we updated 1 to spare
 	qr = utils.Exec(t, conn, "show vitess_replication_status like '%'")

go/vt/vtgate/executor.go

@@ -930,15 +930,15 @@
 			tabletHostPort := ts.GetTabletHostPort()
 			throttlerStatus, err := getTabletThrottlerStatus(tabletHostPort)
 			if err != nil {
-				log.Warningf("Could not get throttler status from %s: %v", tabletHostPort, err)
+				log.Warningf("Could not get throttler status from %s: %v", topoproto.TabletAliasString(ts.Tablet.Alias), err)
 			}
 
 			replSourceHost := ""
 			replSourcePort := int64(0)
 			replIOThreadHealth := ""
 			replSQLThreadHealth := ""
 			replLastError := ""
-			replLag := int64(-1)
+			replLag := "-1" // A string to support NULL as a value
 			sql := "show slave status"
 			results, err := e.txConn.tabletGateway.Execute(ctx, ts.Target, sql, nil, 0, 0, nil)
 			if err != nil || results == nil {
@@ -949,8 +949,25 @@
 				replIOThreadHealth = row["Slave_IO_Running"].ToString()
 				replSQLThreadHealth = row["Slave_SQL_Running"].ToString()
 				replLastError = row["Last_Error"].ToString()
-				if ts.Stats != nil {
-					replLag = int64(ts.Stats.ReplicationLagSeconds)
+				// We cannot check the tablet's tabletenv config from here so
+				// we only use the tablet's stat -- which is managed by the
+				// ReplicationTracker -- if we can tell that it's enabled,
+				// meaning that it has a non-zero value. If it's actually
+				// enabled AND zero (rather than the zeroval), then mysqld
+				// should also return 0 so in this case the value is correct
+				// and equivalent either way. The only reason that we would
+				// want to use the ReplicationTracker based value, when we
+				// can, is because the polling method allows us to get the
+				// estimated lag value when replication is not running (based
+				// on how long we've seen that it's not been running).
+				if ts.Stats != nil && ts.Stats.ReplicationLagSeconds > 0 { // Use the value we get from the ReplicationTracker
+					replLag = fmt.Sprintf("%d", ts.Stats.ReplicationLagSeconds)
+				} else { // Use the value from mysqld
+					if row["Seconds_Behind_Master"].IsNull() {
+						replLag = strings.ToUpper(sqltypes.NullStr) // Uppercase to match mysqld's output in SHOW REPLICA STATUS
+					} else {
+						replLag = row["Seconds_Behind_Master"].ToString()
+					}
 				}
 			}
 			replicationHealth := fmt.Sprintf("{\"EventStreamRunning\":\"%s\",\"EventApplierRunning\":\"%s\",\"LastError\":\"%s\"}", replIOThreadHealth, replSQLThreadHealth, replLastError)
@@ -963,7 +980,7 @@
 				ts.Tablet.Hostname,
 				fmt.Sprintf("%s:%d", replSourceHost, replSourcePort),
 				replicationHealth,
-				fmt.Sprintf("%d", replLag),
+				replLag,
 				throttlerStatus,
 			))
 		}
@@ -1478,11 +1495,14 @@
 	return vterrors.Errorf(vtrpcpb.Code_INVALID_ARGUMENT, "plan includes scatter, which is disallowed using the `no_scatter` command line argument")
 }
 
+// getTabletThrottlerStatus uses HTTP to get the throttler status
+// on a tablet. It uses HTTP because the CheckThrottler RPC is a
+// tmclient RPC and you cannot use tmclient outside of a tablet.
 func getTabletThrottlerStatus(tabletHostPort string) (string, error) {
 	client := http.Client{
 		Timeout: 100 * time.Millisecond,
 	}
-	resp, err := client.Get(fmt.Sprintf("http://%s/throttler/check?app=vtgate", tabletHostPort))
+	resp, err := client.Get(fmt.Sprintf("http://%s/throttler/check-self", tabletHostPort))
 	if err != nil {
 		return "", err
 	}

go/vt/vtgate/executor_test.go

@@ -42,11 +42,6 @@ import (
 	"vitess.io/vitess/go/test/utils"
 	"vitess.io/vitess/go/vt/callerid"
 	"vitess.io/vitess/go/vt/discovery"
-	querypb "vitess.io/vitess/go/vt/proto/query"
-	topodatapb "vitess.io/vitess/go/vt/proto/topodata"
-	vschemapb "vitess.io/vitess/go/vt/proto/vschema"
-	vtgatepb "vitess.io/vitess/go/vt/proto/vtgate"
-	vtrpcpb "vitess.io/vitess/go/vt/proto/vtrpc"
 	"vitess.io/vitess/go/vt/sqlparser"
 	"vitess.io/vitess/go/vt/topo"
 	"vitess.io/vitess/go/vt/vtgate/buffer"
@@ -55,6 +50,12 @@ import (
 	"vitess.io/vitess/go/vt/vtgate/vindexes"
 	"vitess.io/vitess/go/vt/vtgate/vschemaacl"
 	"vitess.io/vitess/go/vt/vtgate/vtgateservice"
+
+	querypb "vitess.io/vitess/go/vt/proto/query"
+	topodatapb "vitess.io/vitess/go/vt/proto/topodata"
+	vschemapb "vitess.io/vitess/go/vt/proto/vschema"
+	vtgatepb "vitess.io/vitess/go/vt/proto/vtgate"
+	vtrpcpb "vitess.io/vitess/go/vt/proto/vtrpc"
 )
 
 func TestExecutorResultsExceeded(t *testing.T) {

go/vt/vttablet/tabletserver/repltracker/poller.go

@@ -21,10 +21,10 @@ import (
 	"time"
 
 	"vitess.io/vitess/go/stats"
-
 	"vitess.io/vitess/go/vt/mysqlctl"
-	vtrpcpb "vitess.io/vitess/go/vt/proto/vtrpc"
 	"vitess.io/vitess/go/vt/vterrors"
+
+	vtrpcpb "vitess.io/vitess/go/vt/proto/vtrpc"
 )
 
 var replicationLagSeconds = stats.NewGauge("replicationLagSec", "replication lag in seconds")

go/vt/vttablet/tabletserver/repltracker/repltracker.go

@@ -23,10 +23,11 @@ import (
 	"vitess.io/vitess/go/stats"
 	"vitess.io/vitess/go/vt/log"
 	"vitess.io/vitess/go/vt/mysqlctl"
-	querypb "vitess.io/vitess/go/vt/proto/query"
-	topodatapb "vitess.io/vitess/go/vt/proto/topodata"
 	"vitess.io/vitess/go/vt/vttablet/tabletserver/heartbeat"
 	"vitess.io/vitess/go/vt/vttablet/tabletserver/tabletenv"
+
+	querypb "vitess.io/vitess/go/vt/proto/query"
+	topodatapb "vitess.io/vitess/go/vt/proto/topodata"
 )
 
 var (