// CreateFakeServers returns the servers to use for these tests
func CreateFakeServers(t *testing.T) (*tabletconntest.FakeQueryService, topo.Server, string) {
cell := "local"
// the FakeServer is just slightly modified
f := tabletconntest.CreateFakeServer(t)
f.TestingGateway = true
f.StreamHealthResponse = &querypb.StreamHealthResponse{
Target: tabletconntest.TestTarget,
Serving: true,
TabletExternallyReparentedTimestamp: 1234589,
RealtimeStats: &querypb.RealtimeStats{
SecondsBehindMaster: 1,
},
}
// The topo server has a single SrvKeyspace
ts := zktestserver.New(t, []string{cell})
if err := ts.UpdateSrvKeyspace(context.Background(), cell, tabletconntest.TestTarget.Keyspace, &topodatapb.SrvKeyspace{
Partitions: []*topodatapb.SrvKeyspace_KeyspacePartition{
{
ServedType: topodatapb.TabletType_MASTER,
ShardReferences: []*topodatapb.ShardReference{
{
Name: tabletconntest.TestTarget.Shard,
},
},
},
},
}); err != nil {
t.Fatalf("can't add srvKeyspace: %v", err)
}
return f, ts, cell
}
// TestCreateShard tests a few cases for CreateShard
func TestCreateShard(t *testing.T) {
ctx := context.Background()
cells := []string{"test_cell"}
// Set up topology.
ts := zktestserver.New(t, cells)
keyspace := "test_keyspace"
shard := "0"
// create shard in a non-existing keyspace
if err := CreateShard(ctx, ts, keyspace, shard); err == nil {
t.Fatalf("CreateShard(invalid keyspace) didn't fail")
}
// create keyspace
if err := ts.CreateKeyspace(ctx, keyspace, &topodatapb.Keyspace{}); err != nil {
t.Fatalf("CreateKeyspace failed: %v", err)
}
// create shard should now work
if err := CreateShard(ctx, ts, keyspace, shard); err != nil {
t.Fatalf("CreateShard failed: %v", err)
}
}
// TestGetOrCreateShard will create / get 100 shards in a keyspace
// for a long time in parallel, making sure the locking and everything
// works correctly.
func TestGetOrCreateShard(t *testing.T) {
ctx := context.Background()
cells := []string{"test_cell"}
// Set up topology.
ts := zktestserver.New(t, cells)
// and do massive parallel GetOrCreateShard
keyspace := "test_keyspace"
wg := sync.WaitGroup{}
rand.Seed(time.Now().UnixNano())
for i := 0; i < 100; i++ {
wg.Add(1)
go func(i int) {
defer wg.Done()
for j := 0; j < 100; j++ {
index := rand.Intn(10)
shard := fmt.Sprintf("%v", index)
si, err := GetOrCreateShard(ctx, ts, keyspace, shard)
if err != nil {
t.Errorf("GetOrCreateShard(%v, %v) failed: %v", i, shard, err)
}
if si.ShardName() != shard {
t.Errorf("si.ShardName() is wrong, got %v expected %v", si.ShardName(), shard)
}
}
}(i)
}
wg.Wait()
}
// TestInitMasterShardChecks makes sure the safety checks work
func TestInitMasterShardChecks(t *testing.T) {
ctx := context.Background()
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
master := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db)
// InitShardMaster with an unknown tablet
if err := wr.InitShardMaster(ctx, master.Tablet.Keyspace, master.Tablet.Shard, &topodatapb.TabletAlias{
Cell: master.Tablet.Alias.Cell,
Uid: master.Tablet.Alias.Uid + 1,
}, false /*force*/, 10*time.Second); err == nil || !strings.Contains(err.Error(), "is not in the shard") {
t.Errorf("InitShardMaster with unknown alias returned wrong error: %v", err)
}
// InitShardMaster with two masters in the shard, no force flag
// (master2 needs to run InitTablet with -force, as it is the second
// master in the same shard)
master2 := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_MASTER, db, ForceInitTablet())
if err := wr.InitShardMaster(ctx, master2.Tablet.Keyspace, master2.Tablet.Shard, master2.Tablet.Alias, false /*force*/, 10*time.Second); err == nil || !strings.Contains(err.Error(), "is not the only master in the shard") {
t.Errorf("InitShardMaster with two masters returned wrong error: %v", err)
}
// InitShardMaster where the new master fails (use force flag
// as we have 2 masters). We force the failure by making the
// SQL commands executed on the master unexpected by the test fixture
master.StartActionLoop(t, wr)
defer master.StopActionLoop(t)
master2.StartActionLoop(t, wr)
defer master2.StopActionLoop(t)
if err := wr.InitShardMaster(ctx, master.Tablet.Keyspace, master.Tablet.Shard, master.Tablet.Alias, true /*force*/, 10*time.Second); err == nil || !strings.Contains(err.Error(), "unexpected extra query") {
t.Errorf("InitShardMaster with new master failing in new master InitMaster returned wrong error: %v", err)
}
}
// TestNewRestartableResultReader tests the correct error handling e.g.
// if the connection to a tablet fails due to a canceled context.
func TestNewRestartableResultReader(t *testing.T) {
wantErr := errors.New("restartable_result_reader_test.go: context canceled")
tabletconn.RegisterDialer("fake_dialer", func(tablet *topodatapb.Tablet, timeout time.Duration) (tabletconn.TabletConn, error) {
return nil, wantErr
})
protocol := flag.CommandLine.Lookup("tablet_protocol").Value.String()
flag.Set("tablet_protocol", "fake_dialer")
// Restore the previous flag value after the test.
defer flag.Set("tablet_protocol", protocol)
// Create dependencies e.g. a "singleTabletProvider" instance.
ts := zktestserver.New(t, []string{"cell1"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
alias := &topodatapb.TabletAlias{
Cell: "cell1",
Uid: 1,
}
tablet := &topodatapb.Tablet{
Keyspace: "ks1",
Shard: "-80",
Alias: alias,
}
ctx := context.Background()
if err := ts.CreateTablet(ctx, tablet); err != nil {
t.Fatalf("CreateTablet failed: %v", err)
}
tp := newSingleTabletProvider(ctx, ts, alias)
_, err := NewRestartableResultReader(ctx, wr.Logger(), tp, nil /* td */, chunk{}, false)
if err == nil || !strings.Contains(err.Error(), wantErr.Error()) {
t.Fatalf("NewRestartableResultReader() should have failed because the context is canceled: %v", err)
}
}
func createTestAgent(ctx context.Context, t *testing.T) *ActionAgent {
ts := zktestserver.New(t, []string{"cell1"})
if err := ts.CreateKeyspace(ctx, "test_keyspace", &topodatapb.Keyspace{}); err != nil {
t.Fatalf("CreateKeyspace failed: %v", err)
}
if err := ts.CreateShard(ctx, "test_keyspace", "0"); err != nil {
t.Fatalf("CreateShard failed: %v", err)
}
port := int32(1234)
tablet := &topodatapb.Tablet{
Alias: tabletAlias,
Hostname: "host",
PortMap: map[string]int32{
"vt": port,
},
Ip: "1.0.0.1",
Keyspace: "test_keyspace",
Shard: "0",
Type: topodatapb.TabletType_SPARE,
}
if err := ts.CreateTablet(ctx, tablet); err != nil {
t.Fatalf("CreateTablet failed: %v", err)
}
mysqlDaemon := &mysqlctl.FakeMysqlDaemon{MysqlPort: 3306}
agent := NewTestActionAgent(ctx, ts, tabletAlias, port, 0, mysqlDaemon)
agent.BinlogPlayerMap = NewBinlogPlayerMap(ts, nil, nil)
agent.HealthReporter = &fakeHealthCheck{}
return agent
}
// TestVtctlThrottlerCommands tests all vtctl commands from the
// "Resharding Throttler" group.
func TestVtctlThrottlerCommands(t *testing.T) {
// Run a throttler server using the default process throttle manager.
listener, err := net.Listen("tcp", ":0")
if err != nil {
t.Fatalf("Cannot listen: %v", err)
}
s := grpc.NewServer()
go s.Serve(listener)
grpcthrottlerserver.StartServer(s, throttler.GlobalManager)
addr := fmt.Sprintf("localhost:%v", listener.Addr().(*net.TCPAddr).Port)
ts := zktestserver.New(t, []string{"cell1", "cell2"})
vp := NewVtctlPipe(t, ts)
defer vp.Close()
// Get and set rate commands do not fail when no throttler is registered.
{
got, err := vp.RunAndOutput([]string{"ThrottlerMaxRates", "-server", addr})
if err != nil {
t.Fatalf("VtctlPipe.RunAndStreamOutput() failed: %v", err)
}
want := "no active throttlers"
if !strings.Contains(got, want) {
t.Fatalf("ThrottlerMaxRates() = %v, want substring = %v", got, want)
}
}
{
got, err := vp.RunAndOutput([]string{"ThrottlerSetMaxRate", "-server", addr, "23"})
if err != nil {
t.Fatalf("VtctlPipe.RunAndStreamOutput() failed: %v", err)
}
want := "no active throttlers"
if !strings.Contains(got, want) {
t.Fatalf("ThrottlerSetMaxRate(23) = %v, want substring = %v", got, want)
}
}
// Add a throttler and check the commands again.
t1, err := throttler.NewThrottler("t1", "TPS", 1 /* threadCount */, 2323, throttler.ReplicationLagModuleDisabled)
if err != nil {
t.Fatal(err)
}
defer t1.Close()
// MaxRates() will return the initial rate.
expectRate(t, vp, addr, "2323")
// Disable the module by setting the rate to 'unlimited'.
setRate(t, vp, addr, "unlimited")
expectRate(t, vp, addr, "unlimited")
// Re-enable it by setting a limit.
setRate(t, vp, addr, "9999")
expectRate(t, vp, addr, "9999")
}
// TestCreateTablet tests all the logic in the topo.CreateTablet method.
func TestCreateTablet(t *testing.T) {
cell := "cell1"
keyspace := "ks1"
shard := "shard1"
ctx := context.Background()
ts := zktestserver.New(t, []string{cell})
// Create a tablet.
alias := &topodatapb.TabletAlias{
Cell: cell,
Uid: 1,
}
tablet := &topodatapb.Tablet{
Keyspace: keyspace,
Shard: shard,
Alias: alias,
}
if err := ts.CreateTablet(ctx, tablet); err != nil {
t.Fatalf("CreateTablet failed: %v", err)
}
// Get the tablet, make sure it's good. Also check ShardReplication.
ti, err := ts.GetTablet(ctx, alias)
if err != nil || !proto.Equal(ti.Tablet, tablet) {
t.Fatalf("Created Tablet doesn't match: %v %v", ti, err)
}
sri, err := ts.GetShardReplication(ctx, cell, keyspace, shard)
if err != nil || len(sri.Nodes) != 1 || !proto.Equal(sri.Nodes[0].TabletAlias, alias) {
t.Fatalf("Created ShardReplication doesn't match: %v %v", sri, err)
}
// Create the same tablet again, make sure it fails with ErrNodeExists.
if err := ts.CreateTablet(ctx, tablet); err != topo.ErrNodeExists {
t.Fatalf("CreateTablet(again) returned: %v", err)
}
// Remove the ShardReplication record, try to create the
// tablets again, make sure it's fixed.
if err := topo.RemoveShardReplicationRecord(ctx, ts, cell, keyspace, shard, alias); err != nil {
t.Fatalf("RemoveShardReplicationRecord failed: %v", err)
}
sri, err = ts.GetShardReplication(ctx, cell, keyspace, shard)
if err != nil || len(sri.Nodes) != 0 {
t.Fatalf("Modifed ShardReplication doesn't match: %v %v", sri, err)
}
if err := ts.CreateTablet(ctx, tablet); err != topo.ErrNodeExists {
t.Fatalf("CreateTablet(again and again) returned: %v", err)
}
sri, err = ts.GetShardReplication(ctx, cell, keyspace, shard)
if err != nil || len(sri.Nodes) != 1 || !proto.Equal(sri.Nodes[0].TabletAlias, alias) {
t.Fatalf("Created ShardReplication doesn't match: %v %v", sri, err)
}
}
func TestTabletData(t *testing.T) {
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
if err := ts.CreateKeyspace(context.Background(), "ks", &topodatapb.Keyspace{
ShardingColumnName: "keyspace_id",
ShardingColumnType: topodatapb.KeyspaceIdType_UINT64,
}); err != nil {
t.Fatalf("CreateKeyspace failed: %v", err)
}
tablet1 := testlib.NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(t, "ks", "-80"))
tablet1.StartActionLoop(t, wr)
defer tablet1.StopActionLoop(t)
shsq := newStreamHealthTabletServer(t)
grpcqueryservice.Register(tablet1.RPCServer, shsq)
thc := newTabletHealthCache(ts)
stats := &querypb.RealtimeStats{
HealthError: "testHealthError",
SecondsBehindMaster: 72,
CpuUsage: 1.1,
}
// Keep broadcasting until the first result goes through.
stop := make(chan struct{})
go func() {
for {
select {
case <-stop:
return
default:
shsq.BroadcastHealth(42, stats)
}
}
}()
// Start streaming and wait for the first result.
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
result, err := thc.Get(ctx, tablet1.Tablet.Alias)
cancel()
close(stop)
if err != nil {
t.Fatalf("thc.Get failed: %v", err)
}
if got, want := result.RealtimeStats, stats; !proto.Equal(got, want) {
t.Errorf("RealtimeStats = %#v, want %#v", got, want)
}
}
func TestTabletExternallyReparentedFailedOldMaster(t *testing.T) {
tabletmanager.SetReparentFlags(time.Minute /* finalizeTimeout */)
ctx := context.Background()
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
// Create an old master, a new master, and a good slave.
oldMaster := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db)
newMaster := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_REPLICA, db)
goodSlave := NewFakeTablet(t, wr, "cell1", 2, topodatapb.TabletType_REPLICA, db)
// Reparent to a replica, and pretend the old master is not responding.
// On the elected master, we will respond to
// TabletActionSlaveWasPromoted
newMaster.StartActionLoop(t, wr)
defer newMaster.StopActionLoop(t)
// On the old master, we will only get a
// TabletActionSlaveWasRestarted call, let's just not
// respond to it at all
// On the good slave, we will respond to
// TabletActionSlaveWasRestarted.
goodSlave.StartActionLoop(t, wr)
defer goodSlave.StopActionLoop(t)
// The reparent should work as expected here
t.Logf("TabletExternallyReparented(new master) expecting success")
tmc := tmclient.NewTabletManagerClient()
ti, err := ts.GetTablet(ctx, newMaster.Tablet.Alias)
if err != nil {
t.Fatalf("GetTablet failed: %v", err)
}
waitID := makeWaitID()
if err := tmc.TabletExternallyReparented(context.Background(), ti.Tablet, waitID); err != nil {
t.Fatalf("TabletExternallyReparented(replica) failed: %v", err)
}
waitForExternalReparent(t, waitID)
// check the old master was converted to replica
tablet, err := ts.GetTablet(ctx, oldMaster.Tablet.Alias)
if err != nil {
t.Fatalf("GetTablet(%v) failed: %v", oldMaster.Tablet.Alias, err)
}
if tablet.Type != topodatapb.TabletType_REPLICA {
t.Fatalf("old master should be spare but is: %v", tablet.Type)
}
}
func TestReparentTablet(t *testing.T) {
ctx := context.Background()
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
// create shard and tablets
if err := ts.CreateShard(ctx, "test_keyspace", "0"); err != nil {
t.Fatalf("CreateShard failed: %v", err)
}
master := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_MASTER, db)
slave := NewFakeTablet(t, wr, "cell1", 2, topodatapb.TabletType_REPLICA, db)
// mark the master inside the shard
si, err := ts.GetShard(ctx, "test_keyspace", "0")
if err != nil {
t.Fatalf("GetShard failed: %v", err)
}
si.MasterAlias = master.Tablet.Alias
if err := ts.UpdateShard(ctx, si); err != nil {
t.Fatalf("UpdateShard failed: %v", err)
}
// master action loop (to initialize host and port)
master.StartActionLoop(t, wr)
defer master.StopActionLoop(t)
// slave loop
slave.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", master.Tablet.Hostname, master.Tablet.PortMap["mysql"])
slave.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
slave.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"set master cmd 1",
}
slave.StartActionLoop(t, wr)
defer slave.StopActionLoop(t)
// run ReparentTablet
if err := wr.ReparentTablet(ctx, slave.Tablet.Alias); err != nil {
t.Fatalf("ReparentTablet failed: %v", err)
}
// check what was run
if err := slave.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Fatalf("slave.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
checkSemiSyncEnabled(t, false, true, slave)
}
// TestTabletExternallyReparentedWithDifferentMysqlPort makes sure
// that if mysql is restarted on the master-elect tablet and has a different
// port, we pick it up correctly.
func TestTabletExternallyReparentedWithDifferentMysqlPort(t *testing.T) {
tabletmanager.SetReparentFlags(time.Minute /* finalizeTimeout */)
ctx := context.Background()
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
// Create an old master, a new master, two good slaves, one bad slave
oldMaster := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db)
newMaster := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_REPLICA, db)
goodSlave := NewFakeTablet(t, wr, "cell1", 2, topodatapb.TabletType_REPLICA, db)
// Now we're restarting mysql on a different port, 3301->3303
// but without updating the Tablet record in topology.
// On the elected master, we will respond to
// TabletActionSlaveWasPromoted, so we need a MysqlDaemon
// that returns no master, and the new port (as returned by mysql)
newMaster.FakeMysqlDaemon.MysqlPort = 3303
newMaster.StartActionLoop(t, wr)
defer newMaster.StopActionLoop(t)
// On the old master, we will only respond to
// TabletActionSlaveWasRestarted and point to the new mysql port
oldMaster.StartActionLoop(t, wr)
defer oldMaster.StopActionLoop(t)
// On the good slaves, we will respond to
// TabletActionSlaveWasRestarted and point to the new mysql port
goodSlave.StartActionLoop(t, wr)
defer goodSlave.StopActionLoop(t)
// This tests the good case, where everything works as planned
t.Logf("TabletExternallyReparented(new master) expecting success")
tmc := tmclient.NewTabletManagerClient()
ti, err := ts.GetTablet(ctx, newMaster.Tablet.Alias)
if err != nil {
t.Fatalf("GetTablet failed: %v", err)
}
waitID := makeWaitID()
if err := tmc.TabletExternallyReparented(context.Background(), ti.Tablet, waitID); err != nil {
t.Fatalf("TabletExternallyReparented(replica) failed: %v", err)
}
waitForExternalReparent(t, waitID)
}
func TestShard(t *testing.T) {
cell := "cell1"
keyspace := "ks1"
shard := "sh1"
ctx := context.Background()
ts := zktestserver.New(t, []string{cell})
// Create a Keyspace / Shard
if err := ts.CreateKeyspace(ctx, keyspace, &topodatapb.Keyspace{}); err != nil {
t.Fatalf("CreateKeyspace failed: %v", err)
}
if err := ts.CreateShard(ctx, keyspace, shard); err != nil {
t.Fatalf("CreateShard failed: %v", err)
}
// Hack the zookeeper backend to create an error for GetShard.
zconn := ts.Impl.(*zktestserver.TestServer).Impl.(*zktopo.Server).GetZConn()
if _, err := zconn.Set(path.Join(zktopo.GlobalKeyspacesPath, keyspace, "shards", shard), []byte{}, -1); err != nil {
t.Fatalf("failed to hack the shard: %v", err)
}
// Create the workflow, run the validator.
w := &Workflow{
logger: logutil.NewMemoryLogger(),
}
sv := &ShardValidator{}
if err := sv.Audit(ctx, ts, w); err != nil {
t.Fatalf("Audit failed: %v", err)
}
if len(w.fixers) != 1 {
t.Fatalf("fixer not added: %v", w.fixers)
}
if !strings.Contains(w.fixers[0].message, "bad shard data") {
t.Errorf("bad message: %v ", w.fixers[0].message)
}
// Run Delete, make sure the entry is removed.
if err := w.fixers[0].fixer.Action(ctx, "Delete"); err != nil {
t.Fatalf("Action failed: %v", err)
}
shards, err := ts.GetShardNames(ctx, keyspace)
if err != nil || len(shards) != 0 {
t.Errorf("bad GetShardNames output: %v %v ", shards, err)
}
}
func TestVersion(t *testing.T) {
// We need to run this test with the /debug/vars version of the
// plugin.
wrangler.ResetDebugVarsGetVersion()
// Initialize our environment
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
vp := NewVtctlPipe(t, ts)
defer vp.Close()
// couple tablets is enough
sourceMaster := NewFakeTablet(t, wr, "cell1", 10, topodatapb.TabletType_MASTER, db,
TabletKeyspaceShard(t, "source", "0"),
StartHTTPServer())
sourceReplica := NewFakeTablet(t, wr, "cell1", 11, topodatapb.TabletType_REPLICA, db,
TabletKeyspaceShard(t, "source", "0"),
StartHTTPServer())
// sourceMaster loop
sourceMasterGitRev := "fake git rev"
sourceMaster.StartActionLoop(t, wr)
sourceMaster.HTTPServer.Handler.(*http.ServeMux).HandleFunc("/debug/vars", expvarHandler(&sourceMasterGitRev))
defer sourceMaster.StopActionLoop(t)
// sourceReplica loop
sourceReplicaGitRev := "fake git rev"
sourceReplica.StartActionLoop(t, wr)
sourceReplica.HTTPServer.Handler.(*http.ServeMux).HandleFunc("/debug/vars", expvarHandler(&sourceReplicaGitRev))
defer sourceReplica.StopActionLoop(t)
// test when versions are the same
if err := vp.Run([]string{"ValidateVersionKeyspace", sourceMaster.Tablet.Keyspace}); err != nil {
t.Fatalf("ValidateVersionKeyspace(same) failed: %v", err)
}
// test when versions are different
sourceReplicaGitRev = "different fake git rev"
if err := vp.Run([]string{"ValidateVersionKeyspace", sourceMaster.Tablet.Keyspace}); err == nil || !strings.Contains(err.Error(), "is different than slave") {
t.Fatalf("ValidateVersionKeyspace(different) returned an unexpected error: %v", err)
}
}
func TestHandleExplorerRedirect(t *testing.T) {
ctx := context.Background()
ts := zktestserver.New(t, []string{"cell1"})
if err := ts.CreateTablet(ctx, &topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{
Cell: "cell1",
Uid: 123,
},
Keyspace: "test_keyspace",
Shard: "123-456",
}); err != nil {
t.Fatalf("CreateTablet failed: %v", err)
}
table := map[string]string{
"/explorers/redirect?type=keyspace&keyspace=test_keyspace": "/app/#/keyspaces/",
"/explorers/redirect?type=shard&keyspace=test_keyspace&shard=-80": "/app/#/shard/test_keyspace/-80",
"/explorers/redirect?type=srv_keyspace&keyspace=test_keyspace&cell=cell1": "/app/#/keyspaces/",
"/explorers/redirect?type=srv_shard&keyspace=test_keyspace&shard=-80&cell=cell1": "/app/#/shard/test_keyspace/-80",
"/explorers/redirect?type=srv_type&keyspace=test_keyspace&shard=-80&cell=cell1&tablet_type=replica": "/app/#/shard/test_keyspace/-80",
"/explorers/redirect?type=tablet&alias=cell1-123": "/app/#/shard/test_keyspace/123-456",
"/explorers/redirect?type=replication&keyspace=test_keyspace&shard=-80&cell=cell1": "/app/#/shard/test_keyspace/-80",
}
for input, want := range table {
request, err := http.NewRequest("GET", input, nil)
if err != nil {
t.Fatalf("NewRequest error: %v", err)
}
if err := request.ParseForm(); err != nil {
t.Fatalf("ParseForm error: %v", err)
}
got, err := handleExplorerRedirect(ctx, ts, request)
if err != nil {
t.Fatalf("handleExplorerRedirect error: %v", err)
}
if got != want {
t.Errorf("handlExplorerRedirect(%#v) = %#v, want %#v", input, got, want)
}
}
}
func newReplica(lagUpdateInterval, degrationInterval, degrationDuration time.Duration) *replica {
t := &testing.T{}
ts := zktestserver.New(t, []string{"cell1"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
db := fakesqldb.Register()
fakeTablet := testlib.NewFakeTablet(t, wr, "cell1", 0,
topodatapb.TabletType_REPLICA, db, testlib.TabletKeyspaceShard(t, "ks", "-80"))
fakeTablet.StartActionLoop(t, wr)
target := querypb.Target{
Keyspace: "ks",
Shard: "-80",
TabletType: topodatapb.TabletType_REPLICA,
}
qs := fakes.NewStreamHealthQueryService(target)
grpcqueryservice.Register(fakeTablet.RPCServer, qs)
throttler, err := throttler.NewThrottler("replica", "TPS", 1, *rate, throttler.ReplicationLagModuleDisabled)
if err != nil {
log.Fatal(err)
}
var nextDegration time.Time
if degrationInterval != time.Duration(0) {
nextDegration = time.Now().Add(degrationInterval)
}
r := &replica{
fakeTablet: fakeTablet,
qs: qs,
throttler: throttler,
replicationStream: make(chan time.Time, 1*1024*1024),
lagUpdateInterval: lagUpdateInterval,
degrationInterval: degrationInterval,
degrationDuration: degrationDuration,
nextDegration: nextDegration,
stopChan: make(chan struct{}),
}
r.wg.Add(1)
go r.processReplicationStream()
return r
}
// TestCreateShardCustomSharding checks ServedTypes is set correctly
// when creating multiple custom sharding shards
func TestCreateShardCustomSharding(t *testing.T) {
ctx := context.Background()
cells := []string{"test_cell"}
// Set up topology.
ts := zktestserver.New(t, cells)
// create keyspace
keyspace := "test_keyspace"
if err := ts.CreateKeyspace(ctx, keyspace, &topodatapb.Keyspace{}); err != nil {
t.Fatalf("CreateKeyspace failed: %v", err)
}
// create first shard in keyspace
shard0 := "0"
if err := CreateShard(ctx, ts, keyspace, shard0); err != nil {
t.Fatalf("CreateShard(shard0) failed: %v", err)
}
if si, err := ts.GetShard(ctx, keyspace, shard0); err != nil {
t.Fatalf("GetShard(shard0) failed: %v", err)
} else {
if len(si.ServedTypes) != 3 {
t.Fatalf("shard0 should have all 3 served types")
}
}
// create second shard in keyspace
shard1 := "1"
if err := CreateShard(ctx, ts, keyspace, shard1); err != nil {
t.Fatalf("CreateShard(shard1) failed: %v", err)
}
if si, err := ts.GetShard(ctx, keyspace, shard1); err != nil {
t.Fatalf("GetShard(shard1) failed: %v", err)
} else {
if len(si.ServedTypes) != 3 {
t.Fatalf("shard1 should have all 3 served types")
}
}
}
func createTestAgent(ctx context.Context, t *testing.T, preStart func(*ActionAgent)) (*ActionAgent, chan<- *binlogplayer.VtClientMock) {
ts := zktestserver.New(t, []string{"cell1"})
if err := ts.CreateKeyspace(ctx, "test_keyspace", &topodatapb.Keyspace{}); err != nil {
t.Fatalf("CreateKeyspace failed: %v", err)
}
if err := ts.CreateShard(ctx, "test_keyspace", "0"); err != nil {
t.Fatalf("CreateShard failed: %v", err)
}
port := int32(1234)
tablet := &topodatapb.Tablet{
Alias: tabletAlias,
Hostname: "host",
PortMap: map[string]int32{
"vt": port,
},
Ip: "1.0.0.1",
Keyspace: "test_keyspace",
Shard: "0",
Type: topodatapb.TabletType_REPLICA,
}
if err := ts.CreateTablet(ctx, tablet); err != nil {
t.Fatalf("CreateTablet failed: %v", err)
}
mysqlDaemon := &mysqlctl.FakeMysqlDaemon{MysqlPort: 3306}
agent := NewTestActionAgent(ctx, ts, tabletAlias, port, 0, mysqlDaemon, preStart)
vtClientMocksChannel := make(chan *binlogplayer.VtClientMock, 1)
agent.BinlogPlayerMap = NewBinlogPlayerMap(ts, mysqlDaemon, func() binlogplayer.VtClient {
return <-vtClientMocksChannel
})
agent.HealthReporter = &fakeHealthCheck{}
return agent, vtClientMocksChannel
}
func TestPlannedReparentShard(t *testing.T) {
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
vp := NewVtctlPipe(t, ts)
defer vp.Close()
// Create a master, a couple good slaves
oldMaster := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db)
newMaster := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_REPLICA, db)
goodSlave1 := NewFakeTablet(t, wr, "cell1", 2, topodatapb.TabletType_REPLICA, db)
goodSlave2 := NewFakeTablet(t, wr, "cell2", 3, topodatapb.TabletType_REPLICA, db)
// new master
newMaster.FakeMysqlDaemon.ReadOnly = true
newMaster.FakeMysqlDaemon.Replicating = true
newMaster.FakeMysqlDaemon.WaitMasterPosition = replication.Position{
GTIDSet: replication.MariadbGTID{
Domain: 7,
Server: 123,
Sequence: 990,
},
}
newMaster.FakeMysqlDaemon.PromoteSlaveResult = replication.Position{
GTIDSet: replication.MariadbGTID{
Domain: 7,
Server: 456,
Sequence: 991,
},
}
newMaster.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"CREATE DATABASE IF NOT EXISTS _vt",
"SUBCREATE TABLE IF NOT EXISTS _vt.reparent_journal",
"SUBINSERT INTO _vt.reparent_journal (time_created_ns, action_name, master_alias, replication_position) VALUES",
}
newMaster.StartActionLoop(t, wr)
defer newMaster.StopActionLoop(t)
// old master
oldMaster.FakeMysqlDaemon.ReadOnly = false
oldMaster.FakeMysqlDaemon.Replicating = false
oldMaster.FakeMysqlDaemon.DemoteMasterPosition = newMaster.FakeMysqlDaemon.WaitMasterPosition
oldMaster.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", newMaster.Tablet.Hostname, newMaster.Tablet.PortMap["mysql"])
oldMaster.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
oldMaster.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"set master cmd 1",
"START SLAVE",
}
oldMaster.StartActionLoop(t, wr)
defer oldMaster.StopActionLoop(t)
oldMaster.Agent.QueryServiceControl.(*tabletservermock.Controller).SetQueryServiceEnabledForTests(true)
// good slave 1 is replicating
goodSlave1.FakeMysqlDaemon.ReadOnly = true
goodSlave1.FakeMysqlDaemon.Replicating = true
goodSlave1.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", newMaster.Tablet.Hostname, newMaster.Tablet.PortMap["mysql"])
goodSlave1.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
goodSlave1.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"STOP SLAVE",
"set master cmd 1",
"START SLAVE",
}
goodSlave1.StartActionLoop(t, wr)
defer goodSlave1.StopActionLoop(t)
// good slave 2 is not replicating
goodSlave2.FakeMysqlDaemon.ReadOnly = true
goodSlave2.FakeMysqlDaemon.Replicating = false
goodSlave2.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", newMaster.Tablet.Hostname, newMaster.Tablet.PortMap["mysql"])
goodSlave2.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
goodSlave2.StartActionLoop(t, wr)
goodSlave2.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"set master cmd 1",
}
defer goodSlave2.StopActionLoop(t)
// run PlannedReparentShard
if err := vp.Run([]string{"PlannedReparentShard", "-wait_slave_timeout", "10s", newMaster.Tablet.Keyspace + "/" + newMaster.Tablet.Shard, topoproto.TabletAliasString(newMaster.Tablet.Alias)}); err != nil {
t.Fatalf("PlannedReparentShard failed: %v", err)
}
// check what was run
if err := newMaster.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Errorf("newMaster.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if err := oldMaster.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Errorf("oldMaster.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if err := goodSlave1.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Errorf("goodSlave1.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if err := goodSlave2.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Errorf("goodSlave2.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if newMaster.FakeMysqlDaemon.ReadOnly {
t.Errorf("newMaster.FakeMysqlDaemon.ReadOnly set")
}
if !oldMaster.FakeMysqlDaemon.ReadOnly {
t.Errorf("oldMaster.FakeMysqlDaemon.ReadOnly not set")
}
if !goodSlave1.FakeMysqlDaemon.ReadOnly {
t.Errorf("goodSlave1.FakeMysqlDaemon.ReadOnly not set")
}
if !goodSlave2.FakeMysqlDaemon.ReadOnly {
t.Errorf("goodSlave2.FakeMysqlDaemon.ReadOnly not set")
}
if !oldMaster.Agent.QueryServiceControl.IsServing() {
t.Errorf("oldMaster...QueryServiceControl not serving")
}
// verify the old master was told to start replicating (and not
// the slave that wasn't replicating in the first place)
if !oldMaster.FakeMysqlDaemon.Replicating {
t.Errorf("oldMaster.FakeMysqlDaemon.Replicating not set")
}
if !goodSlave1.FakeMysqlDaemon.Replicating {
t.Errorf("goodSlave1.FakeMysqlDaemon.Replicating not set")
}
if goodSlave2.FakeMysqlDaemon.Replicating {
t.Errorf("goodSlave2.FakeMysqlDaemon.Replicating set")
}
checkSemiSyncEnabled(t, true, true, newMaster)
checkSemiSyncEnabled(t, false, true, goodSlave1, goodSlave2, oldMaster)
}
func TestVerticalSplitDiff(t *testing.T) {
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
ctx := context.Background()
wi := NewInstance(ts, "cell1", time.Second)
sourceMaster := testlib.NewFakeTablet(t, wi.wr, "cell1", 0,
topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(t, "source_ks", "0"))
sourceRdonly1 := testlib.NewFakeTablet(t, wi.wr, "cell1", 1,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(t, "source_ks", "0"))
sourceRdonly2 := testlib.NewFakeTablet(t, wi.wr, "cell1", 2,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(t, "source_ks", "0"))
// Create the destination keyspace with the appropriate ServedFromMap
ki := &topodatapb.Keyspace{
ServedFroms: []*topodatapb.Keyspace_ServedFrom{
{
TabletType: topodatapb.TabletType_MASTER,
Keyspace: "source_ks",
},
{
TabletType: topodatapb.TabletType_REPLICA,
Keyspace: "source_ks",
},
{
TabletType: topodatapb.TabletType_RDONLY,
Keyspace: "source_ks",
},
},
}
wi.wr.TopoServer().CreateKeyspace(ctx, "destination_ks", ki)
destMaster := testlib.NewFakeTablet(t, wi.wr, "cell1", 10,
topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(t, "destination_ks", "0"))
destRdonly1 := testlib.NewFakeTablet(t, wi.wr, "cell1", 11,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(t, "destination_ks", "0"))
destRdonly2 := testlib.NewFakeTablet(t, wi.wr, "cell1", 12,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(t, "destination_ks", "0"))
for _, ft := range []*testlib.FakeTablet{sourceMaster, sourceRdonly1, sourceRdonly2, destMaster, destRdonly1, destRdonly2} {
ft.StartActionLoop(t, wi.wr)
defer ft.StopActionLoop(t)
}
wi.wr.SetSourceShards(ctx, "destination_ks", "0", []*topodatapb.TabletAlias{sourceRdonly1.Tablet.Alias}, []string{"moving.*", "view1"})
// add the topo and schema data we'll need
if err := wi.wr.RebuildKeyspaceGraph(ctx, "source_ks", nil); err != nil {
t.Fatalf("RebuildKeyspaceGraph failed: %v", err)
}
if err := wi.wr.RebuildKeyspaceGraph(ctx, "destination_ks", nil); err != nil {
t.Fatalf("RebuildKeyspaceGraph failed: %v", err)
}
for _, rdonly := range []*testlib.FakeTablet{sourceRdonly1, sourceRdonly2, destRdonly1, destRdonly2} {
// both source and destination have the table definition for 'moving1'.
// source also has "staying1" while destination has "extra1".
// (Both additional tables should be ignored by the diff.)
extraTable := "staying1"
if rdonly == destRdonly1 || rdonly == destRdonly2 {
extraTable = "extra1"
}
rdonly.FakeMysqlDaemon.Schema = &tabletmanagerdatapb.SchemaDefinition{
DatabaseSchema: "",
TableDefinitions: []*tabletmanagerdatapb.TableDefinition{
{
Name: "moving1",
Columns: []string{"id", "msg"},
PrimaryKeyColumns: []string{"id"},
Type: tmutils.TableBaseTable,
},
{
Name: extraTable,
Columns: []string{"id", "msg"},
PrimaryKeyColumns: []string{"id"},
Type: tmutils.TableBaseTable,
},
{
Name: "view1",
Type: tmutils.TableView,
},
},
}
qs := fakes.NewStreamHealthQueryService(rdonly.Target())
qs.AddDefaultHealthResponse()
grpcqueryservice.Register(rdonly.RPCServer, &verticalDiffTabletServer{
t: t,
StreamHealthQueryService: qs,
})
}
// Run the vtworker command.
args := []string{"VerticalSplitDiff", "destination_ks/0"}
// We need to use FakeTabletManagerClient because we don't
// have a good way to fake the binlog player yet, which is
// necessary for synchronizing replication.
wr := wrangler.New(logutil.NewConsoleLogger(), ts, newFakeTMCTopo(ts))
if err := runCommand(t, wi, wr, args); err != nil {
t.Fatal(err)
}
}
// CreateWorkerInstance returns a properly configured vtworker instance.
func CreateWorkerInstance(t *testing.T) *worker.Instance {
ts := zktestserver.New(t, []string{"cell1", "cell2"})
return worker.NewInstance(context.Background(), ts, "cell1", 1*time.Second)
}
func testWaitForDrain(t *testing.T, desc, cells string, drain drainDirective, expectedErrors []string) {
const keyspace = "ks"
const shard = "-80"
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
flag.Set("vtctl_healthcheck_timeout", "0.25s")
vp := NewVtctlPipe(t, ts)
defer vp.Close()
// Create keyspace.
if err := ts.CreateKeyspace(context.Background(), keyspace, &topodatapb.Keyspace{
ShardingColumnName: "keyspace_id",
ShardingColumnType: topodatapb.KeyspaceIdType_UINT64,
}); err != nil {
t.Fatalf("CreateKeyspace failed: %v", err)
}
t1 := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_REPLICA, db,
TabletKeyspaceShard(t, keyspace, shard))
t2 := NewFakeTablet(t, wr, "cell2", 1, topodatapb.TabletType_REPLICA, db,
TabletKeyspaceShard(t, keyspace, shard))
for _, ft := range []*FakeTablet{t1, t2} {
ft.StartActionLoop(t, wr)
defer ft.StopActionLoop(t)
}
target := querypb.Target{
Keyspace: keyspace,
Shard: shard,
TabletType: topodatapb.TabletType_REPLICA,
}
fqs1 := newFakeQueryService(target)
fqs2 := newFakeQueryService(target)
grpcqueryservice.RegisterForTest(t1.RPCServer, fqs1)
grpcqueryservice.RegisterForTest(t2.RPCServer, fqs2)
// Run vtctl WaitForDrain and react depending on its output.
timeout := "0.5s"
if len(expectedErrors) == 0 {
// Tests with a positive outcome should have a more generous timeout to
// avoid flakyness.
timeout = "30s"
}
stream, err := vp.RunAndStreamOutput(
[]string{"WaitForDrain", "-cells", cells, "-retry_delay", "100ms", "-timeout", timeout,
keyspace + "/" + shard, topodatapb.TabletType_REPLICA.String()})
if err != nil {
t.Fatalf("VtctlPipe.RunAndStreamOutput() failed: %v", err)
}
// QPS = 1.0. Tablets are not drained yet.
fqs1.addHealthResponse(1.0)
fqs2.addHealthResponse(1.0)
var le *logutilpb.Event
for {
le, err = stream.Recv()
if err != nil {
break
}
line := logutil.EventString(le)
t.Logf(line)
if strings.Contains(line, "for all healthy tablets to be drained") {
t.Log("Successfully waited for WaitForDrain to be blocked because tablets have a QPS rate > 0.0")
break
} else {
t.Log("waiting for WaitForDrain to see a QPS rate > 0.0")
}
}
if drain&DrainCell1 != 0 {
fqs1.addHealthResponse(0.0)
} else {
fqs1.addHealthResponse(2.0)
}
if drain&DrainCell2 != 0 {
fqs2.addHealthResponse(0.0)
} else {
fqs2.addHealthResponse(2.0)
}
// If a cell was drained, rate should go below <0.0 now.
// If not all selected cells were drained, this will end after "-timeout".
for {
le, err = stream.Recv()
if err == nil {
vp.t.Logf(logutil.EventString(le))
} else {
break
}
}
if len(expectedErrors) == 0 {
if err != io.EOF {
t.Fatalf("TestWaitForDrain: %v: no error expected but got: %v", desc, err)
}
// else: Success.
} else {
if err == nil || err == io.EOF {
t.Fatalf("TestWaitForDrain: %v: error expected but got none", desc)
}
for _, errString := range expectedErrors {
if !strings.Contains(err.Error(), errString) {
t.Fatalf("TestWaitForDrain: %v: error does not include expected string. got: %v want: %v", desc, err, errString)
}
}
// Success.
}
}
func TestFindAllKeyspaceShards(t *testing.T) {
ctx := context.Background()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
// no keyspace / shards
ks, err := findAllKeyspaceShards(ctx, ts, "cell1")
if err != nil {
t.Errorf("unexpected error: %v", err)
}
if len(ks) > 0 {
t.Errorf("why did I get anything? %v", ks)
}
// add one
if err := ts.UpdateSrvKeyspace(ctx, "cell1", "test_keyspace", &topodatapb.SrvKeyspace{
Partitions: []*topodatapb.SrvKeyspace_KeyspacePartition{
{
ServedType: topodatapb.TabletType_MASTER,
ShardReferences: []*topodatapb.ShardReference{
{
Name: "test_shard0",
},
},
},
},
}); err != nil {
t.Fatalf("can't add srvKeyspace: %v", err)
}
// get it
ks, err = findAllKeyspaceShards(ctx, ts, "cell1")
if err != nil {
t.Errorf("unexpected error: %v", err)
}
if !reflect.DeepEqual(ks, map[keyspaceShard]bool{
keyspaceShard{
keyspace: "test_keyspace",
shard: "test_shard0",
}: true,
}) {
t.Errorf("got wrong value: %v", ks)
}
// add another one
if err := ts.UpdateSrvKeyspace(ctx, "cell1", "test_keyspace2", &topodatapb.SrvKeyspace{
Partitions: []*topodatapb.SrvKeyspace_KeyspacePartition{
{
ServedType: topodatapb.TabletType_MASTER,
ShardReferences: []*topodatapb.ShardReference{
{
Name: "test_shard1",
},
},
},
{
ServedType: topodatapb.TabletType_MASTER,
ShardReferences: []*topodatapb.ShardReference{
{
Name: "test_shard2",
},
},
},
},
}); err != nil {
t.Fatalf("can't add srvKeyspace: %v", err)
}
// get it
ks, err = findAllKeyspaceShards(ctx, ts, "cell1")
if err != nil {
t.Errorf("unexpected error: %v", err)
}
if !reflect.DeepEqual(ks, map[keyspaceShard]bool{
keyspaceShard{
keyspace: "test_keyspace",
shard: "test_shard0",
}: true,
keyspaceShard{
keyspace: "test_keyspace2",
shard: "test_shard1",
}: true,
keyspaceShard{
keyspace: "test_keyspace2",
shard: "test_shard2",
}: true,
}) {
t.Errorf("got wrong value: %v", ks)
}
}
// TestInitMasterShard is the good scenario test, where everything
// works as planned
func TestInitMasterShard(t *testing.T) {
ctx := context.Background()
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
vp := NewVtctlPipe(t, ts)
defer vp.Close()
db.AddQuery("CREATE DATABASE IF NOT EXISTS `vt_test_keyspace`", &sqltypes.Result{})
// Create a master, a couple good slaves
master := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db)
goodSlave1 := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_REPLICA, db)
goodSlave2 := NewFakeTablet(t, wr, "cell2", 2, topodatapb.TabletType_REPLICA, db)
// Master: set a plausible ReplicationPosition to return,
// and expect to add entry in _vt.reparent_journal
master.FakeMysqlDaemon.CurrentMasterPosition = replication.Position{
GTIDSet: replication.MariadbGTID{
Domain: 5,
Server: 456,
Sequence: 890,
},
}
master.FakeMysqlDaemon.ReadOnly = true
master.FakeMysqlDaemon.ResetReplicationResult = []string{"reset rep 1"}
master.FakeMysqlDaemon.SetSlavePositionCommandsResult = []string{"new master shouldn't use this"}
master.FakeMysqlDaemon.SetMasterCommandsResult = []string{"new master shouldn't use this"}
master.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"reset rep 1",
"CREATE DATABASE IF NOT EXISTS _vt",
"SUBCREATE TABLE IF NOT EXISTS _vt.reparent_journal",
"CREATE DATABASE IF NOT EXISTS _vt",
"SUBCREATE TABLE IF NOT EXISTS _vt.reparent_journal",
"SUBINSERT INTO _vt.reparent_journal (time_created_ns, action_name, master_alias, replication_position) VALUES",
}
master.StartActionLoop(t, wr)
defer master.StopActionLoop(t)
// Slave1: expect to be reset and re-parented
goodSlave1.FakeMysqlDaemon.ReadOnly = true
goodSlave1.FakeMysqlDaemon.ResetReplicationResult = []string{"reset rep 1"}
goodSlave1.FakeMysqlDaemon.SetSlavePositionCommandsPos = master.FakeMysqlDaemon.CurrentMasterPosition
goodSlave1.FakeMysqlDaemon.SetSlavePositionCommandsResult = []string{"cmd1"}
goodSlave1.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", master.Tablet.Hostname, master.Tablet.PortMap["mysql"])
goodSlave1.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
goodSlave1.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"reset rep 1",
"cmd1",
"set master cmd 1",
"START SLAVE",
}
goodSlave1.StartActionLoop(t, wr)
defer goodSlave1.StopActionLoop(t)
// Slave2: expect to be re-parented
goodSlave2.FakeMysqlDaemon.ReadOnly = true
goodSlave2.FakeMysqlDaemon.ResetReplicationResult = []string{"reset rep 2"}
goodSlave2.FakeMysqlDaemon.SetSlavePositionCommandsPos = master.FakeMysqlDaemon.CurrentMasterPosition
goodSlave2.FakeMysqlDaemon.SetSlavePositionCommandsResult = []string{"cmd1", "cmd2"}
goodSlave2.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", master.Tablet.Hostname, master.Tablet.PortMap["mysql"])
goodSlave2.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
goodSlave2.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"reset rep 2",
"cmd1",
"cmd2",
"set master cmd 1",
"START SLAVE",
}
goodSlave2.StartActionLoop(t, wr)
defer goodSlave2.StopActionLoop(t)
// run InitShardMaster
if err := vp.Run([]string{"InitShardMaster", "-wait_slave_timeout", "10s", master.Tablet.Keyspace + "/" + master.Tablet.Shard, topoproto.TabletAliasString(master.Tablet.Alias)}); err != nil {
t.Fatalf("InitShardMaster failed: %v", err)
}
// check what was run
if master.FakeMysqlDaemon.ReadOnly {
t.Errorf("master was not turned read-write")
}
si, err := ts.GetShard(ctx, master.Tablet.Keyspace, master.Tablet.Shard)
if err != nil {
t.Fatalf("GetShard failed: %v", err)
}
if !topoproto.TabletAliasEqual(si.MasterAlias, master.Tablet.Alias) {
t.Errorf("unexpected shard master alias, got %v expected %v", si.MasterAlias, master.Tablet.Alias)
}
if err := master.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Fatalf("master.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if err := goodSlave1.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Fatalf("goodSlave1.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if err := goodSlave2.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Fatalf("goodSlave2.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
}
// TestInitMasterShardOneSlaveFails makes sure that if one slave fails to
// proceed, the action completes anyway
func TestInitMasterShardOneSlaveFails(t *testing.T) {
ctx := context.Background()
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
// Create a master, a couple slaves
master := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db)
goodSlave := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_REPLICA, db)
badSlave := NewFakeTablet(t, wr, "cell2", 2, topodatapb.TabletType_REPLICA, db)
// Master: set a plausible ReplicationPosition to return,
// and expect to add entry in _vt.reparent_journal
master.FakeMysqlDaemon.CurrentMasterPosition = replication.Position{
GTIDSet: replication.MariadbGTID{
Domain: 5,
Server: 456,
Sequence: 890,
},
}
master.FakeMysqlDaemon.ReadOnly = true
master.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"CREATE DATABASE IF NOT EXISTS _vt",
"SUBCREATE TABLE IF NOT EXISTS _vt.reparent_journal",
"CREATE DATABASE IF NOT EXISTS _vt",
"SUBCREATE TABLE IF NOT EXISTS _vt.reparent_journal",
"SUBINSERT INTO _vt.reparent_journal (time_created_ns, action_name, master_alias, replication_position) VALUES",
}
master.StartActionLoop(t, wr)
defer master.StopActionLoop(t)
// goodSlave: expect to be re-parented
goodSlave.FakeMysqlDaemon.ReadOnly = true
goodSlave.FakeMysqlDaemon.SetSlavePositionCommandsPos = master.FakeMysqlDaemon.CurrentMasterPosition
goodSlave.FakeMysqlDaemon.SetSlavePositionCommandsResult = []string{"cmd1"}
goodSlave.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", master.Tablet.Hostname, master.Tablet.PortMap["mysql"])
goodSlave.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
goodSlave.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"cmd1",
"set master cmd 1",
"START SLAVE",
}
goodSlave.StartActionLoop(t, wr)
defer goodSlave.StopActionLoop(t)
// badSlave: insert an error by failing the master hostname input
// on purpose
badSlave.FakeMysqlDaemon.ReadOnly = true
badSlave.FakeMysqlDaemon.SetSlavePositionCommandsPos = master.FakeMysqlDaemon.CurrentMasterPosition
badSlave.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", "", master.Tablet.PortMap["mysql"])
badSlave.StartActionLoop(t, wr)
defer badSlave.StopActionLoop(t)
// also change the master alias in the Shard object, to make sure it
// is set back.
si, err := ts.GetShard(ctx, master.Tablet.Keyspace, master.Tablet.Shard)
if err != nil {
t.Fatalf("GetShard failed: %v", err)
}
si.MasterAlias.Uid++
if err := ts.UpdateShard(ctx, si); err != nil {
t.Fatalf("UpdateShard failed: %v", err)
}
// run InitShardMaster without force, it fails because master is
// changing.
if err := wr.InitShardMaster(ctx, master.Tablet.Keyspace, master.Tablet.Shard, master.Tablet.Alias, false /*force*/, 10*time.Second); err == nil || !strings.Contains(err.Error(), "is not the shard master") {
t.Errorf("InitShardMaster with mismatched new master returned wrong error: %v", err)
}
// run InitShardMaster
if err := wr.InitShardMaster(ctx, master.Tablet.Keyspace, master.Tablet.Shard, master.Tablet.Alias, true /*force*/, 10*time.Second); err == nil || !strings.Contains(err.Error(), "wrong input for SetMasterCommands") {
t.Errorf("InitShardMaster with one failed slave returned wrong error: %v", err)
}
// check what was run: master should still be good
if master.FakeMysqlDaemon.ReadOnly {
t.Errorf("master was not turned read-write")
}
si, err = ts.GetShard(ctx, master.Tablet.Keyspace, master.Tablet.Shard)
if err != nil {
t.Fatalf("GetShard failed: %v", err)
}
if !topoproto.TabletAliasEqual(si.MasterAlias, master.Tablet.Alias) {
t.Errorf("unexpected shard master alias, got %v expected %v", si.MasterAlias, master.Tablet.Alias)
}
}
func TestAPI(t *testing.T) {
ctx := context.Background()
cells := []string{"cell1", "cell2"}
ts := zktestserver.New(t, cells)
actionRepo := NewActionRepository(ts)
server := httptest.NewServer(nil)
defer server.Close()
// Populate topo.
ts.CreateKeyspace(ctx, "ks1", &topodatapb.Keyspace{ShardingColumnName: "shardcol"})
ts.Impl.CreateShard(ctx, "ks1", "-80", &topodatapb.Shard{
Cells: cells,
KeyRange: &topodatapb.KeyRange{Start: nil, End: []byte{0x80}},
})
ts.Impl.CreateShard(ctx, "ks1", "80-", &topodatapb.Shard{
Cells: cells,
KeyRange: &topodatapb.KeyRange{Start: []byte{0x80}, End: nil},
})
tablet1 := topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{Cell: "cell1", Uid: 100},
Keyspace: "ks1",
Shard: "-80",
Type: topodatapb.TabletType_REPLICA,
KeyRange: &topodatapb.KeyRange{Start: nil, End: []byte{0x80}},
PortMap: map[string]int32{"vt": 100},
}
ts.CreateTablet(ctx, &tablet1)
tablet2 := topodatapb.Tablet{
Alias: &topodatapb.TabletAlias{Cell: "cell2", Uid: 200},
Keyspace: "ks1",
Shard: "-80",
Type: topodatapb.TabletType_REPLICA,
KeyRange: &topodatapb.KeyRange{Start: nil, End: []byte{0x80}},
PortMap: map[string]int32{"vt": 200},
}
ts.CreateTablet(ctx, &tablet2)
// Populate fake actions.
actionRepo.RegisterKeyspaceAction("TestKeyspaceAction",
func(ctx context.Context, wr *wrangler.Wrangler, keyspace string, r *http.Request) (string, error) {
return "TestKeyspaceAction Result", nil
})
actionRepo.RegisterShardAction("TestShardAction",
func(ctx context.Context, wr *wrangler.Wrangler, keyspace, shard string, r *http.Request) (string, error) {
return "TestShardAction Result", nil
})
actionRepo.RegisterTabletAction("TestTabletAction", "",
func(ctx context.Context, wr *wrangler.Wrangler, tabletAlias *topodatapb.TabletAlias, r *http.Request) (string, error) {
return "TestTabletAction Result", nil
})
realtimeStats := newRealtimeStatsForTesting()
initAPI(ctx, ts, actionRepo, realtimeStats)
target := &querypb.Target{
Keyspace: "ks1",
Shard: "-80",
TabletType: topodatapb.TabletType_REPLICA,
}
stats := &querypb.RealtimeStats{
HealthError: "",
SecondsBehindMaster: 2,
BinlogPlayersCount: 0,
CpuUsage: 12.1,
Qps: 5.6,
}
tabletStats := &discovery.TabletStats{
Tablet: &tablet1,
Target: target,
Up: true,
Serving: true,
TabletExternallyReparentedTimestamp: 5,
Stats: stats,
LastError: nil,
}
realtimeStats.tabletStats.StatsUpdate(tabletStats)
// Test cases.
table := []struct {
method, path, want string
}{
// Cells
{"GET", "cells", `["cell1","cell2"]`},
// Keyspaces
{"GET", "keyspaces", `["ks1"]`},
{"GET", "keyspaces/ks1", `{
"sharding_column_name": "shardcol"
}`},
{"POST", "keyspaces/ks1?action=TestKeyspaceAction", `{
"Name": "TestKeyspaceAction",
"Parameters": "ks1",
"Output": "TestKeyspaceAction Result",
"Error": false
}`},
// Shards
{"GET", "shards/ks1/", `["-80","80-"]`},
{"GET", "shards/ks1/-80", `{
"key_range": {"end":"gA=="},
"cells": ["cell1", "cell2"]
}`},
{"POST", "shards/ks1/-80?action=TestShardAction", `{
"Name": "TestShardAction",
"Parameters": "ks1/-80",
"Output": "TestShardAction Result",
"Error": false
}`},
// Tablets
{"GET", "tablets/?shard=ks1%2F-80", `[
{"cell":"cell1","uid":100},
{"cell":"cell2","uid":200}
]`},
{"GET", "tablets/?cell=cell1", `[
{"cell":"cell1","uid":100}
]`},
{"GET", "tablets/?shard=ks1%2F-80&cell=cell2", `[
{"cell":"cell2","uid":200}
]`},
{"GET", "tablets/cell1-100", `{
"alias": {"cell": "cell1", "uid": 100},
"port_map": {"vt": 100},
"keyspace": "ks1",
"shard": "-80",
"key_range": {"end": "gA=="},
"type": 2
}`},
{"POST", "tablets/cell1-100?action=TestTabletAction", `{
"Name": "TestTabletAction",
"Parameters": "cell1-0000000100",
"Output": "TestTabletAction Result",
"Error": false
}`},
//Tablet Updates
{"GET", "tablet_statuses/cell1/ks1/-80/REPLICA", `{"100":{"Tablet":{"alias":{"cell":"cell1","uid":100},"port_map":{"vt":100},"keyspace":"ks1","shard":"-80","key_range":{"end":"gA=="},"type":2},"Name":"","Target":{"keyspace":"ks1","shard":"-80","tablet_type":2},"Up":true,"Serving":true,"TabletExternallyReparentedTimestamp":5,"Stats":{"seconds_behind_master":2,"cpu_usage":12.1,"qps":5.6},"LastError":null}}`},
{"GET", "tablet_statuses/cell1/ks1/replica", "can't get tablet_statuses: invalid target path: \"cell1/ks1/replica\" expected path: <cell>/<keyspace>/<shard>/<type>"},
{"GET", "tablet_statuses/cell1/ks1/-80/hello", "can't get tablet_statuses: invalid tablet type: hello"},
}
for _, in := range table {
var resp *http.Response
var err error
switch in.method {
case "GET":
resp, err = http.Get(server.URL + apiPrefix + in.path)
case "POST":
resp, err = http.Post(server.URL+apiPrefix+in.path, "", nil)
default:
t.Errorf("[%v] unknown method: %v", in.path, in.method)
continue
}
if err != nil {
t.Errorf("[%v] http error: %v", in.path, err)
continue
}
body, err := ioutil.ReadAll(resp.Body)
resp.Body.Close()
if err != nil {
t.Errorf("[%v] ioutil.ReadAll(resp.Body) error: %v", in.path, err)
continue
}
got := compactJSON(body)
want := compactJSON([]byte(in.want))
if want == "" {
// want is no valid JSON. Fallback to a string comparison.
want = in.want
// For unknown reasons errors have a trailing "\n\t\t". Remove it.
got = strings.TrimSpace(string(body))
}
if got != want {
t.Errorf("[%v] got '%v', want '%v'", in.path, got, want)
continue
}
}
}
func TestBackupRestore(t *testing.T) {
// Initialize our environment
ctx := context.Background()
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
wr := wrangler.New(logutil.NewConsoleLogger(), ts, tmclient.NewTabletManagerClient())
vp := NewVtctlPipe(t, ts)
defer vp.Close()
// Set up mock query results.
db.AddQuery("CREATE DATABASE IF NOT EXISTS _vt", &sqltypes.Result{})
db.AddQuery("BEGIN", &sqltypes.Result{})
db.AddQuery("COMMIT", &sqltypes.Result{})
db.AddQueryPattern(`SET @@session\.sql_log_bin = .*`, &sqltypes.Result{})
db.AddQueryPattern(`CREATE TABLE IF NOT EXISTS _vt\.shard_metadata .*`, &sqltypes.Result{})
db.AddQueryPattern(`CREATE TABLE IF NOT EXISTS _vt\.local_metadata .*`, &sqltypes.Result{})
db.AddQueryPattern(`INSERT INTO _vt\.local_metadata .*`, &sqltypes.Result{})
// Initialize our temp dirs
root, err := ioutil.TempDir("", "backuptest")
if err != nil {
t.Fatalf("os.TempDir failed: %v", err)
}
defer os.RemoveAll(root)
// Initialize BackupStorage
fbsRoot := path.Join(root, "fbs")
*filebackupstorage.FileBackupStorageRoot = fbsRoot
*backupstorage.BackupStorageImplementation = "file"
// Initialize the fake mysql root directories
sourceInnodbDataDir := path.Join(root, "source_innodb_data")
sourceInnodbLogDir := path.Join(root, "source_innodb_log")
sourceDataDir := path.Join(root, "source_data")
sourceDataDbDir := path.Join(sourceDataDir, "vt_db")
for _, s := range []string{sourceInnodbDataDir, sourceInnodbLogDir, sourceDataDbDir} {
if err := os.MkdirAll(s, os.ModePerm); err != nil {
t.Fatalf("failed to create directory %v: %v", s, err)
}
}
if err := ioutil.WriteFile(path.Join(sourceInnodbDataDir, "innodb_data_1"), []byte("innodb data 1 contents"), os.ModePerm); err != nil {
t.Fatalf("failed to write file innodb_data_1: %v", err)
}
if err := ioutil.WriteFile(path.Join(sourceInnodbLogDir, "innodb_log_1"), []byte("innodb log 1 contents"), os.ModePerm); err != nil {
t.Fatalf("failed to write file innodb_log_1: %v", err)
}
if err := ioutil.WriteFile(path.Join(sourceDataDbDir, "db.opt"), []byte("db opt file"), os.ModePerm); err != nil {
t.Fatalf("failed to write file db.opt: %v", err)
}
// create a master tablet, not started, just for shard health
master := NewFakeTablet(t, wr, "cell1", 0, topodatapb.TabletType_MASTER, db)
// create a single tablet, set it up so we can do backups
sourceTablet := NewFakeTablet(t, wr, "cell1", 1, topodatapb.TabletType_REPLICA, db)
sourceTablet.FakeMysqlDaemon.ReadOnly = true
sourceTablet.FakeMysqlDaemon.Replicating = true
sourceTablet.FakeMysqlDaemon.CurrentMasterPosition = replication.Position{
GTIDSet: replication.MariadbGTID{
Domain: 2,
Server: 123,
Sequence: 457,
},
}
sourceTablet.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"STOP SLAVE",
"START SLAVE",
}
sourceTablet.FakeMysqlDaemon.Mycnf = &mysqlctl.Mycnf{
DataDir: sourceDataDir,
InnodbDataHomeDir: sourceInnodbDataDir,
InnodbLogGroupHomeDir: sourceInnodbLogDir,
}
sourceTablet.StartActionLoop(t, wr)
defer sourceTablet.StopActionLoop(t)
// run the backup
if err := vp.Run([]string{"Backup", topoproto.TabletAliasString(sourceTablet.Tablet.Alias)}); err != nil {
t.Fatalf("Backup failed: %v", err)
}
// verify the full status
if err := sourceTablet.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Errorf("sourceTablet.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if !sourceTablet.FakeMysqlDaemon.Replicating {
t.Errorf("sourceTablet.FakeMysqlDaemon.Replicating not set")
}
if !sourceTablet.FakeMysqlDaemon.Running {
t.Errorf("sourceTablet.FakeMysqlDaemon.Running not set")
}
// create a destination tablet, set it up so we can do restores
destTablet := NewFakeTablet(t, wr, "cell1", 2, topodatapb.TabletType_REPLICA, db)
destTablet.FakeMysqlDaemon.ReadOnly = true
destTablet.FakeMysqlDaemon.Replicating = true
destTablet.FakeMysqlDaemon.CurrentMasterPosition = replication.Position{
GTIDSet: replication.MariadbGTID{
Domain: 2,
Server: 123,
Sequence: 457,
},
}
destTablet.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"cmd1",
"set master cmd 1",
"START SLAVE",
}
destTablet.FakeMysqlDaemon.Mycnf = &mysqlctl.Mycnf{
DataDir: sourceDataDir,
InnodbDataHomeDir: sourceInnodbDataDir,
InnodbLogGroupHomeDir: sourceInnodbLogDir,
BinLogPath: path.Join(root, "bin-logs/filename_prefix"),
RelayLogPath: path.Join(root, "relay-logs/filename_prefix"),
RelayLogIndexPath: path.Join(root, "relay-log.index"),
RelayLogInfoPath: path.Join(root, "relay-log.info"),
}
destTablet.FakeMysqlDaemon.FetchSuperQueryMap = map[string]*sqltypes.Result{
"SHOW DATABASES": {},
}
destTablet.FakeMysqlDaemon.SetSlavePositionCommandsPos = sourceTablet.FakeMysqlDaemon.CurrentMasterPosition
destTablet.FakeMysqlDaemon.SetSlavePositionCommandsResult = []string{"cmd1"}
destTablet.FakeMysqlDaemon.SetMasterCommandsInput = fmt.Sprintf("%v:%v", master.Tablet.Hostname, master.Tablet.PortMap["mysql"])
destTablet.FakeMysqlDaemon.SetMasterCommandsResult = []string{"set master cmd 1"}
destTablet.StartActionLoop(t, wr)
defer destTablet.StopActionLoop(t)
if err := destTablet.Agent.RestoreData(ctx, logutil.NewConsoleLogger(), false /* deleteBeforeRestore */); err != nil {
t.Fatalf("RestoreData failed: %v", err)
}
// verify the full status
if err := destTablet.FakeMysqlDaemon.CheckSuperQueryList(); err != nil {
t.Errorf("destTablet.FakeMysqlDaemon.CheckSuperQueryList failed: %v", err)
}
if !destTablet.FakeMysqlDaemon.Replicating {
t.Errorf("destTablet.FakeMysqlDaemon.Replicating not set")
}
if !destTablet.FakeMysqlDaemon.Running {
t.Errorf("destTablet.FakeMysqlDaemon.Running not set")
}
}
func (tc *splitCloneTestCase) setUpWithConcurreny(v3 bool, concurrency, writeQueryMaxRows, rowsCount int) {
*useV3ReshardingMode = v3
db := fakesqldb.Register()
tc.ts = zktestserver.New(tc.t, []string{"cell1", "cell2"})
ctx := context.Background()
tc.wi = NewInstance(tc.ts, "cell1", time.Second)
if v3 {
if err := tc.ts.CreateKeyspace(ctx, "ks", &topodatapb.Keyspace{}); err != nil {
tc.t.Fatalf("CreateKeyspace v3 failed: %v", err)
}
vs := &vschemapb.Keyspace{
Sharded: true,
Vindexes: map[string]*vschemapb.Vindex{
"table1_index": {
Type: "numeric",
},
},
Tables: map[string]*vschemapb.Table{
"table1": {
ColumnVindexes: []*vschemapb.ColumnVindex{
{
Column: "keyspace_id",
Name: "table1_index",
},
},
},
},
}
if err := tc.ts.SaveVSchema(ctx, "ks", vs); err != nil {
tc.t.Fatalf("SaveVSchema v3 failed: %v", err)
}
} else {
if err := tc.ts.CreateKeyspace(ctx, "ks", &topodatapb.Keyspace{
ShardingColumnName: "keyspace_id",
ShardingColumnType: topodatapb.KeyspaceIdType_UINT64,
}); err != nil {
tc.t.Fatalf("CreateKeyspace v2 failed: %v", err)
}
}
sourceMaster := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 0,
topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(tc.t, "ks", "-80"))
sourceRdonly1 := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 1,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(tc.t, "ks", "-80"))
sourceRdonly2 := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 2,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(tc.t, "ks", "-80"))
leftMaster := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 10,
topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(tc.t, "ks", "-40"))
// leftReplica is used by the reparent test.
leftReplica := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 11,
topodatapb.TabletType_REPLICA, db, testlib.TabletKeyspaceShard(tc.t, "ks", "-40"))
tc.leftReplica = leftReplica
leftRdonly1 := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 12,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(tc.t, "ks", "-40"))
leftRdonly2 := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 13,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(tc.t, "ks", "-40"))
rightMaster := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 20,
topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(tc.t, "ks", "40-80"))
rightRdonly1 := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 22,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(tc.t, "ks", "40-80"))
rightRdonly2 := testlib.NewFakeTablet(tc.t, tc.wi.wr, "cell1", 23,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(tc.t, "ks", "40-80"))
tc.tablets = []*testlib.FakeTablet{sourceMaster, sourceRdonly1, sourceRdonly2,
leftMaster, tc.leftReplica, leftRdonly1, leftRdonly2, rightMaster, rightRdonly1, rightRdonly2}
for _, ft := range tc.tablets {
ft.StartActionLoop(tc.t, tc.wi.wr)
}
// add the topo and schema data we'll need
if err := tc.ts.CreateShard(ctx, "ks", "80-"); err != nil {
tc.t.Fatalf("CreateShard(\"-80\") failed: %v", err)
}
if err := tc.wi.wr.SetKeyspaceShardingInfo(ctx, "ks", "keyspace_id", topodatapb.KeyspaceIdType_UINT64, false); err != nil {
tc.t.Fatalf("SetKeyspaceShardingInfo failed: %v", err)
}
if err := tc.wi.wr.RebuildKeyspaceGraph(ctx, "ks", nil); err != nil {
tc.t.Fatalf("RebuildKeyspaceGraph failed: %v", err)
}
for _, sourceRdonly := range []*testlib.FakeTablet{sourceRdonly1, sourceRdonly2} {
sourceRdonly.FakeMysqlDaemon.Schema = &tabletmanagerdatapb.SchemaDefinition{
DatabaseSchema: "",
TableDefinitions: []*tabletmanagerdatapb.TableDefinition{
{
Name: "table1",
// "id" is the last column in the list on purpose to test for
// regressions. The reconciliation code will SELECT with the primary
// key columns first. The same ordering must be used throughout the
// process e.g. by RowAggregator or the v2Resolver.
Columns: []string{"msg", "keyspace_id", "id"},
PrimaryKeyColumns: []string{"id"},
Type: tmutils.TableBaseTable,
// Set the row count to avoid that --min_rows_per_chunk reduces the
// number of chunks.
RowCount: uint64(rowsCount),
},
},
}
sourceRdonly.FakeMysqlDaemon.DbAppConnectionFactory = sourceRdonlyFactory(
tc.t, "vt_ks", "table1", splitCloneTestMin, splitCloneTestMax)
sourceRdonly.FakeMysqlDaemon.CurrentMasterPosition = replication.Position{
GTIDSet: replication.MariadbGTID{Domain: 12, Server: 34, Sequence: 5678},
}
sourceRdonly.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"STOP SLAVE",
"START SLAVE",
}
shqs := fakes.NewStreamHealthQueryService(sourceRdonly.Target())
shqs.AddDefaultHealthResponse()
qs := newTestQueryService(tc.t, sourceRdonly.Target(), shqs, 0, 1, topoproto.TabletAliasString(sourceRdonly.Tablet.Alias), false /* omitKeyspaceID */)
qs.addGeneratedRows(100, 100+rowsCount)
grpcqueryservice.Register(sourceRdonly.RPCServer, qs)
tc.sourceRdonlyQs = append(tc.sourceRdonlyQs, qs)
}
// Set up destination rdonlys which will be used as input for the diff during the clone.
for i, destRdonly := range []*testlib.FakeTablet{leftRdonly1, rightRdonly1, leftRdonly2, rightRdonly2} {
shqs := fakes.NewStreamHealthQueryService(destRdonly.Target())
shqs.AddDefaultHealthResponse()
qs := newTestQueryService(tc.t, destRdonly.Target(), shqs, i%2, 2, topoproto.TabletAliasString(destRdonly.Tablet.Alias), false /* omitKeyspaceID */)
grpcqueryservice.Register(destRdonly.RPCServer, qs)
if i%2 == 0 {
tc.leftRdonlyQs = append(tc.leftRdonlyQs, qs)
} else {
tc.rightRdonlyQs = append(tc.rightRdonlyQs, qs)
}
}
tc.leftMasterFakeDb = NewFakePoolConnectionQuery(tc.t, "leftMaster")
tc.leftReplicaFakeDb = NewFakePoolConnectionQuery(tc.t, "leftReplica")
tc.rightMasterFakeDb = NewFakePoolConnectionQuery(tc.t, "rightMaster")
// In the default test case there will be 30 inserts per destination shard
// because 10 writer threads will insert 5 rows on each destination shard.
// (100 rowsCount / 10 writers / 2 shards = 5 rows.)
// Due to --write_query_max_rows=2 there will be 3 inserts for 5 rows.
rowsPerDestinationShard := rowsCount / 2
rowsPerThread := rowsPerDestinationShard / concurrency
insertsPerThread := math.Ceil(float64(rowsPerThread) / float64(writeQueryMaxRows))
insertsTotal := int(insertsPerThread) * concurrency
for i := 1; i <= insertsTotal; i++ {
tc.leftMasterFakeDb.addExpectedQuery("INSERT INTO `vt_ks`.`table1` (`id`, `msg`, `keyspace_id`) VALUES (*", nil)
// leftReplica is unused by default.
tc.rightMasterFakeDb.addExpectedQuery("INSERT INTO `vt_ks`.`table1` (`id`, `msg`, `keyspace_id`) VALUES (*", nil)
}
expectBlpCheckpointCreationQueries(tc.leftMasterFakeDb)
expectBlpCheckpointCreationQueries(tc.rightMasterFakeDb)
leftMaster.FakeMysqlDaemon.DbAppConnectionFactory = tc.leftMasterFakeDb.getFactory()
leftReplica.FakeMysqlDaemon.DbAppConnectionFactory = tc.leftReplicaFakeDb.getFactory()
rightMaster.FakeMysqlDaemon.DbAppConnectionFactory = tc.rightMasterFakeDb.getFactory()
// Fake stream health reponses because vtworker needs them to find the master.
tc.leftMasterQs = fakes.NewStreamHealthQueryService(leftMaster.Target())
tc.leftMasterQs.AddDefaultHealthResponse()
tc.leftReplicaQs = fakes.NewStreamHealthQueryService(leftReplica.Target())
tc.leftReplicaQs.AddDefaultHealthResponse()
tc.rightMasterQs = fakes.NewStreamHealthQueryService(rightMaster.Target())
tc.rightMasterQs.AddDefaultHealthResponse()
grpcqueryservice.Register(leftMaster.RPCServer, tc.leftMasterQs)
grpcqueryservice.Register(leftReplica.RPCServer, tc.leftReplicaQs)
grpcqueryservice.Register(rightMaster.RPCServer, tc.rightMasterQs)
tc.defaultWorkerArgs = []string{
"SplitClone",
"-online=false",
// --max_tps is only specified to enable the throttler and ensure that the
// code is executed. But the intent here is not to throttle the test, hence
// the rate limit is set very high.
"-max_tps", "9999",
"-write_query_max_rows", strconv.Itoa(writeQueryMaxRows),
"-chunk_count", strconv.Itoa(concurrency),
"-min_rows_per_chunk", strconv.Itoa(rowsPerThread),
"-source_reader_count", strconv.Itoa(concurrency),
"-destination_writer_count", strconv.Itoa(concurrency),
"ks/-80"}
}
// CreateTopoServer returns the test topo server properly configured
func CreateTopoServer(t *testing.T) topo.Server {
return zktestserver.New(t, []string{"cell1", "cell2"})
}
// TestVerticalSplitClone will run VerticalSplitClone in the combined
// online and offline mode. The online phase will copy 100 rows from the source
// to the destination and the offline phase won't copy any rows as the source
// has not changed in the meantime.
func TestVerticalSplitClone(t *testing.T) {
db := fakesqldb.Register()
ts := zktestserver.New(t, []string{"cell1", "cell2"})
ctx := context.Background()
wi := NewInstance(ctx, ts, "cell1", time.Second)
sourceMaster := testlib.NewFakeTablet(t, wi.wr, "cell1", 0,
topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(t, "source_ks", "0"))
sourceRdonly := testlib.NewFakeTablet(t, wi.wr, "cell1", 1,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(t, "source_ks", "0"))
// Create the destination keyspace with the appropriate ServedFromMap
ki := &topodatapb.Keyspace{
ServedFroms: []*topodatapb.Keyspace_ServedFrom{
{
TabletType: topodatapb.TabletType_MASTER,
Keyspace: "source_ks",
},
{
TabletType: topodatapb.TabletType_REPLICA,
Keyspace: "source_ks",
},
{
TabletType: topodatapb.TabletType_RDONLY,
Keyspace: "source_ks",
},
},
}
wi.wr.TopoServer().CreateKeyspace(ctx, "destination_ks", ki)
destMaster := testlib.NewFakeTablet(t, wi.wr, "cell1", 10,
topodatapb.TabletType_MASTER, db, testlib.TabletKeyspaceShard(t, "destination_ks", "0"))
destRdonly := testlib.NewFakeTablet(t, wi.wr, "cell1", 11,
topodatapb.TabletType_RDONLY, db, testlib.TabletKeyspaceShard(t, "destination_ks", "0"))
for _, ft := range []*testlib.FakeTablet{sourceMaster, sourceRdonly, destMaster, destRdonly} {
ft.StartActionLoop(t, wi.wr)
defer ft.StopActionLoop(t)
}
// add the topo and schema data we'll need
if err := wi.wr.RebuildKeyspaceGraph(ctx, "source_ks", nil); err != nil {
t.Fatalf("RebuildKeyspaceGraph failed: %v", err)
}
if err := wi.wr.RebuildKeyspaceGraph(ctx, "destination_ks", nil); err != nil {
t.Fatalf("RebuildKeyspaceGraph failed: %v", err)
}
// Set up source rdonly which will be used as input for the diff during the clone.
sourceRdonly.FakeMysqlDaemon.Schema = &tabletmanagerdatapb.SchemaDefinition{
DatabaseSchema: "",
TableDefinitions: []*tabletmanagerdatapb.TableDefinition{
{
Name: "moving1",
Columns: []string{"id", "msg"},
PrimaryKeyColumns: []string{"id"},
Type: tmutils.TableBaseTable,
// Set the row count to avoid that --min_rows_per_chunk reduces the
// number of chunks.
RowCount: 100,
},
{
Name: "view1",
Type: tmutils.TableView,
},
},
}
sourceRdonly.FakeMysqlDaemon.DbAppConnectionFactory = sourceRdonlyFactory(
t, "vt_source_ks", "moving1", verticalSplitCloneTestMin, verticalSplitCloneTestMax)
sourceRdonly.FakeMysqlDaemon.CurrentMasterPosition = replication.Position{
GTIDSet: replication.MariadbGTID{Domain: 12, Server: 34, Sequence: 5678},
}
sourceRdonly.FakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
"STOP SLAVE",
"START SLAVE",
}
sourceRdonlyShqs := fakes.NewStreamHealthQueryService(sourceRdonly.Target())
sourceRdonlyShqs.AddDefaultHealthResponse()
sourceRdonlyQs := newTestQueryService(t, sourceRdonly.Target(), sourceRdonlyShqs, 0, 1, topoproto.TabletAliasString(sourceRdonly.Tablet.Alias), true /* omitKeyspaceID */)
sourceRdonlyQs.addGeneratedRows(verticalSplitCloneTestMin, verticalSplitCloneTestMax)
grpcqueryservice.Register(sourceRdonly.RPCServer, sourceRdonlyQs)
// Set up destination rdonly which will be used as input for the diff during the clone.
destRdonlyShqs := fakes.NewStreamHealthQueryService(destRdonly.Target())
destRdonlyShqs.AddDefaultHealthResponse()
destRdonlyQs := newTestQueryService(t, destRdonly.Target(), destRdonlyShqs, 0, 1, topoproto.TabletAliasString(destRdonly.Tablet.Alias), true /* omitKeyspaceID */)
// This tablet is empty and does not return any rows.
grpcqueryservice.Register(destRdonly.RPCServer, destRdonlyQs)
// We read 100 source rows. sourceReaderCount is set to 10, so
// we'll have 100/10=10 rows per table chunk.
// destinationPackCount is set to 4, so we take 4 source rows
// at once. So we'll process 4 + 4 + 2 rows to get to 10.
// That means 3 insert statements on the target. So 3 * 10
// = 30 insert statements on the destination.
destMasterFakeDb := createVerticalSplitCloneDestinationFakeDb(t, "destMaster", 30)
defer destMasterFakeDb.verifyAllExecutedOrFail()
destMaster.FakeMysqlDaemon.DbAppConnectionFactory = destMasterFakeDb.getFactory()
// Fake stream health reponses because vtworker needs them to find the master.
qs := fakes.NewStreamHealthQueryService(destMaster.Target())
qs.AddDefaultHealthResponse()
grpcqueryservice.Register(destMaster.RPCServer, qs)
// Only wait 1 ms between retries, so that the test passes faster
*executeFetchRetryTime = (1 * time.Millisecond)
// When the online clone inserted the last rows, modify the destination test
// query service such that it will return them as well.
destMasterFakeDb.getEntry(29).AfterFunc = func() {
destRdonlyQs.addGeneratedRows(verticalSplitCloneTestMin, verticalSplitCloneTestMax)
}
// Run the vtworker command.
args := []string{
"VerticalSplitClone",
// --max_tps is only specified to enable the throttler and ensure that the
// code is executed. But the intent here is not to throttle the test, hence
// the rate limit is set very high.
"-max_tps", "9999",
"-tables", "moving.*,view1",
"-source_reader_count", "10",
// Each chunk pipeline will process 10 rows. To spread them out across 3
// write queries, set the max row count per query to 4. (10 = 4+4+2)
"-write_query_max_rows", "4",
"-min_rows_per_chunk", "10",
"-destination_writer_count", "10",
// This test uses only one healthy RDONLY tablet.
"-min_healthy_rdonly_tablets", "1",
"destination_ks/0",
}
if err := runCommand(t, wi, wi.wr, args); err != nil {
t.Fatal(err)
}
if inserts := statsOnlineInsertsCounters.Counts()["moving1"]; inserts != 100 {
t.Errorf("wrong number of rows inserted: got = %v, want = %v", inserts, 100)
}
if updates := statsOnlineUpdatesCounters.Counts()["moving1"]; updates != 0 {
t.Errorf("wrong number of rows updated: got = %v, want = %v", updates, 0)
}
if deletes := statsOnlineDeletesCounters.Counts()["moving1"]; deletes != 0 {
t.Errorf("wrong number of rows deleted: got = %v, want = %v", deletes, 0)
}
if inserts := statsOfflineInsertsCounters.Counts()["moving1"]; inserts != 0 {
t.Errorf("no stats for the offline clone phase should have been modified. got inserts = %v", inserts)
}
if updates := statsOfflineUpdatesCounters.Counts()["moving1"]; updates != 0 {
t.Errorf("no stats for the offline clone phase should have been modified. got updates = %v", updates)
}
if deletes := statsOfflineDeletesCounters.Counts()["moving1"]; deletes != 0 {
t.Errorf("no stats for the offline clone phase should have been modified. got deletes = %v", deletes)
}
wantRetryCount := int64(1)
if got := statsRetryCount.Get(); got != wantRetryCount {
t.Errorf("Wrong statsRetryCounter: got %v, wanted %v", got, wantRetryCount)
}
wantRetryReadOnlyCount := int64(1)
if got := statsRetryCounters.Counts()[retryCategoryReadOnly]; got != wantRetryReadOnlyCount {
t.Errorf("Wrong statsRetryCounters: got %v, wanted %v", got, wantRetryReadOnlyCount)
}
}