// InitTablet creates or updates a tablet. If no parent is specified
// in the tablet, and the tablet has a slave type, we will find the
// appropriate parent. If createShardAndKeyspace is true and the
// parent keyspace or shard don't exist, they will be created. If
// allowUpdate is true, and a tablet with the same ID exists, just update it.
// If a tablet is created as master, and there is already a different
// master in the shard, allowMasterOverride must be set.
func (wr *Wrangler) InitTablet(ctx context.Context, tablet *topodatapb.Tablet, allowMasterOverride, createShardAndKeyspace, allowUpdate bool) error {
if err := topo.TabletComplete(tablet); err != nil {
return err
}
// get the shard, possibly creating it
var err error
var si *topo.ShardInfo
if createShardAndKeyspace {
// create the parent keyspace and shard if needed
si, err = wr.ts.GetOrCreateShard(ctx, tablet.Keyspace, tablet.Shard)
} else {
si, err = wr.ts.GetShard(ctx, tablet.Keyspace, tablet.Shard)
if err == topo.ErrNoNode {
return fmt.Errorf("missing parent shard, use -parent option to create it, or CreateKeyspace / CreateShard")
}
}
// get the shard, checks a couple things
if err != nil {
return fmt.Errorf("cannot get (or create) shard %v/%v: %v", tablet.Keyspace, tablet.Shard, err)
}
if !key.KeyRangeEqual(si.KeyRange, tablet.KeyRange) {
return fmt.Errorf("shard %v/%v has a different KeyRange: %v != %v", tablet.Keyspace, tablet.Shard, si.KeyRange, tablet.KeyRange)
}
if tablet.Type == topodatapb.TabletType_MASTER && si.HasMaster() && !topoproto.TabletAliasEqual(si.MasterAlias, tablet.Alias) && !allowMasterOverride {
return fmt.Errorf("creating this tablet would override old master %v in shard %v/%v, use allow_master_override flag", topoproto.TabletAliasString(si.MasterAlias), tablet.Keyspace, tablet.Shard)
}
// update the shard record if needed
if err := wr.updateShardCellsAndMaster(ctx, si, tablet.Alias, tablet.Type, allowMasterOverride); err != nil {
return err
}
err = wr.ts.CreateTablet(ctx, tablet)
if err == topo.ErrNodeExists && allowUpdate {
// Try to update then
oldTablet, err := wr.ts.GetTablet(ctx, tablet.Alias)
if err != nil {
return fmt.Errorf("failed reading existing tablet %v: %v", topoproto.TabletAliasString(tablet.Alias), err)
}
// Check we have the same keyspace / shard, and if not,
// require the allowDifferentShard flag.
if oldTablet.Keyspace != tablet.Keyspace || oldTablet.Shard != tablet.Shard {
return fmt.Errorf("old tablet has shard %v/%v. Cannot override with shard %v/%v. Delete and re-add tablet if you want to change the tablet's keyspace/shard", oldTablet.Keyspace, oldTablet.Shard, tablet.Keyspace, tablet.Shard)
}
*(oldTablet.Tablet) = *tablet
if err := wr.ts.UpdateTablet(ctx, oldTablet); err != nil {
return fmt.Errorf("failed updating tablet %v: %v", topoproto.TabletAliasString(tablet.Alias), err)
}
}
return nil
}
// InitTablet initializes the tablet record if necessary.
func (agent *ActionAgent) InitTablet(port, gRPCPort int32) error {
// only enabled if one of init_tablet_type (when healthcheck
// is disabled) or init_keyspace (when healthcheck is enabled)
// is passed in, then check other parameters
if *initTabletType == "" && *initKeyspace == "" {
return nil
}
// figure out our default target type
var tabletType topodatapb.TabletType
if *initTabletType != "" {
if *targetTabletType != "" {
log.Fatalf("cannot specify both target_tablet_type and init_tablet_type parameters (as they might conflict)")
}
// use the type specified on the command line
var err error
tabletType, err = topoproto.ParseTabletType(*initTabletType)
if err != nil {
log.Fatalf("Invalid init tablet type %v: %v", *initTabletType, err)
}
if tabletType == topodatapb.TabletType_MASTER {
// We disallow MASTER, so we don't have to change
// shard.MasterAlias, and deal with the corner cases.
log.Fatalf("init_tablet_type cannot be %v", tabletType)
}
} else if *targetTabletType != "" {
if strings.ToUpper(*targetTabletType) == topodatapb.TabletType_name[int32(topodatapb.TabletType_MASTER)] {
log.Fatalf("target_tablet_type cannot be '%v'. Use '%v' instead.", tabletType, topodatapb.TabletType_REPLICA)
}
// use spare, the healthcheck will turn us into what
// we need to be eventually
tabletType = topodatapb.TabletType_SPARE
} else {
log.Fatalf("if init tablet is enabled, one of init_tablet_type or target_tablet_type needs to be specified")
}
// create a context for this whole operation
ctx, cancel := context.WithTimeout(agent.batchCtx, *initTimeout)
defer cancel()
// since we're assigned to a shard, make sure it exists, see if
// we are its master, and update its cells list if necessary
if *initKeyspace == "" || *initShard == "" {
log.Fatalf("if init tablet is enabled and the target type is not idle, init_keyspace and init_shard also need to be specified")
}
shard, _, err := topo.ValidateShardName(*initShard)
if err != nil {
log.Fatalf("cannot validate shard name: %v", err)
}
log.Infof("Reading shard record %v/%v", *initKeyspace, shard)
// read the shard, create it if necessary
si, err := topotools.GetOrCreateShard(ctx, agent.TopoServer, *initKeyspace, shard)
if err != nil {
return fmt.Errorf("InitTablet cannot GetOrCreateShard shard: %v", err)
}
if si.MasterAlias != nil && topoproto.TabletAliasEqual(si.MasterAlias, agent.TabletAlias) {
// we are the current master for this shard (probably
// means the master tablet process was just restarted),
// so InitTablet as master.
tabletType = topodatapb.TabletType_MASTER
}
// See if we need to add the tablet's cell to the shard's cell
// list. If we do, it has to be under the shard lock.
if !si.HasCell(agent.TabletAlias.Cell) {
actionNode := actionnode.UpdateShard()
lockPath, err := actionNode.LockShard(ctx, agent.TopoServer, *initKeyspace, shard)
if err != nil {
return fmt.Errorf("LockShard(%v/%v) failed: %v", *initKeyspace, shard, err)
}
// re-read the shard with the lock
si, err = agent.TopoServer.GetShard(ctx, *initKeyspace, shard)
if err != nil {
return actionNode.UnlockShard(ctx, agent.TopoServer, *initKeyspace, shard, lockPath, err)
}
// see if we really need to update it now
if !si.HasCell(agent.TabletAlias.Cell) {
si.Cells = append(si.Cells, agent.TabletAlias.Cell)
// write it back
if err := agent.TopoServer.UpdateShard(ctx, si); err != nil {
return actionNode.UnlockShard(ctx, agent.TopoServer, *initKeyspace, shard, lockPath, err)
}
}
// and unlock
if err := actionNode.UnlockShard(ctx, agent.TopoServer, *initKeyspace, shard, lockPath, nil); err != nil {
return err
}
}
log.Infof("Initializing the tablet for type %v", tabletType)
// figure out the hostname
hostname := *tabletHostname
if hostname != "" {
log.Infof("Using hostname: %v from -tablet_hostname flag.", hostname)
} else {
hostname, err := netutil.FullyQualifiedHostname()
if err != nil {
return err
}
log.Infof("Using detected machine hostname: %v To change this, fix your machine network configuration or override it with -tablet_hostname.", hostname)
}
// create and populate tablet record
tablet := &topodatapb.Tablet{
Alias: agent.TabletAlias,
Hostname: hostname,
PortMap: make(map[string]int32),
Keyspace: *initKeyspace,
Shard: *initShard,
Type: tabletType,
DbNameOverride: *initDbNameOverride,
Tags: initTags,
}
if port != 0 {
tablet.PortMap["vt"] = port
}
if gRPCPort != 0 {
tablet.PortMap["grpc"] = gRPCPort
}
if err := topo.TabletComplete(tablet); err != nil {
return fmt.Errorf("InitTablet TabletComplete failed: %v", err)
}
// now try to create the record
err = agent.TopoServer.CreateTablet(ctx, tablet)
switch err {
case nil:
// it worked, we're good, can update the replication graph
if err := topo.UpdateTabletReplicationData(ctx, agent.TopoServer, tablet); err != nil {
return fmt.Errorf("UpdateTabletReplicationData failed: %v", err)
}
case topo.ErrNodeExists:
// The node already exists, will just try to update
// it. So we read it first.
oldTablet, err := agent.TopoServer.GetTablet(ctx, tablet.Alias)
if err != nil {
return fmt.Errorf("InitTablet failed to read existing tablet record: %v", err)
}
// Sanity check the keyspace and shard
if oldTablet.Keyspace != tablet.Keyspace || oldTablet.Shard != tablet.Shard {
return fmt.Errorf("InitTablet failed because existing tablet keyspace and shard %v/%v differ from the provided ones %v/%v", oldTablet.Keyspace, oldTablet.Shard, tablet.Keyspace, tablet.Shard)
}
// And overwrite the rest
*(oldTablet.Tablet) = *tablet
if err := agent.TopoServer.UpdateTablet(ctx, oldTablet); err != nil {
return fmt.Errorf("UpdateTablet failed: %v", err)
}
// Note we don't need to UpdateTabletReplicationData
// as the tablet already existed with the right data
// in the replication graph
default:
return fmt.Errorf("CreateTablet failed: %v", err)
}
// and now update the serving graph. Note we do that in any case,
// to clean any inaccurate record from any part of the serving graph.
if err := topotools.UpdateTabletEndpoints(ctx, agent.TopoServer, tablet); err != nil {
return fmt.Errorf("UpdateTabletEndpoints failed: %v", err)
}
return nil
}
// InitTablet creates or updates a tablet. If no parent is specified
// in the tablet, and the tablet has a slave type, we will find the
// appropriate parent. If createShardAndKeyspace is true and the
// parent keyspace or shard don't exist, they will be created. If
// update is true, and a tablet with the same ID exists, update it.
// If Force is true, and a tablet with the same ID already exists, it
// will be scrapped and deleted, and then recreated.
func (wr *Wrangler) InitTablet(ctx context.Context, tablet *topo.Tablet, force, createShardAndKeyspace, update bool) error {
if err := topo.TabletComplete(tablet); err != nil {
return err
}
if topo.IsInReplicationGraph(tablet.Type) {
// get the shard, possibly creating it
var err error
var si *topo.ShardInfo
if createShardAndKeyspace {
// create the parent keyspace and shard if needed
si, err = topotools.GetOrCreateShard(ctx, wr.ts, tablet.Keyspace, tablet.Shard)
} else {
si, err = wr.ts.GetShard(ctx, tablet.Keyspace, tablet.Shard)
if err == topo.ErrNoNode {
return fmt.Errorf("missing parent shard, use -parent option to create it, or CreateKeyspace / CreateShard")
}
}
// get the shard, checks a couple things
if err != nil {
return fmt.Errorf("cannot get (or create) shard %v/%v: %v", tablet.Keyspace, tablet.Shard, err)
}
if key.ProtoToKeyRange(si.KeyRange) != tablet.KeyRange {
return fmt.Errorf("shard %v/%v has a different KeyRange: %v != %v", tablet.Keyspace, tablet.Shard, si.KeyRange, tablet.KeyRange)
}
if tablet.Type == topo.TYPE_MASTER && !topo.TabletAliasIsZero(si.MasterAlias) && topo.ProtoToTabletAlias(si.MasterAlias) != tablet.Alias && !force {
return fmt.Errorf("creating this tablet would override old master %v in shard %v/%v", si.MasterAlias, tablet.Keyspace, tablet.Shard)
}
// update the shard record if needed
if err := wr.updateShardCellsAndMaster(ctx, si, topo.TabletAliasToProto(tablet.Alias), topo.TabletTypeToProto(tablet.Type), force); err != nil {
return err
}
}
err := topo.CreateTablet(ctx, wr.ts, tablet)
if err != nil && err == topo.ErrNodeExists {
// Try to update nicely, but if it fails fall back to force behavior.
if update || force {
oldTablet, err := wr.ts.GetTablet(ctx, tablet.Alias)
if err != nil {
wr.Logger().Warningf("failed reading tablet %v: %v", tablet.Alias, err)
} else {
if oldTablet.Keyspace == tablet.Keyspace && oldTablet.Shard == tablet.Shard {
*(oldTablet.Tablet) = *tablet
if err := topo.UpdateTablet(ctx, wr.ts, oldTablet); err != nil {
wr.Logger().Warningf("failed updating tablet %v: %v", tablet.Alias, err)
// now fall through the Scrap case
} else {
if !topo.IsInReplicationGraph(tablet.Type) {
return nil
}
if err := topo.UpdateTabletReplicationData(ctx, wr.ts, tablet); err != nil {
wr.Logger().Warningf("failed updating tablet replication data for %v: %v", tablet.Alias, err)
// now fall through the Scrap case
} else {
return nil
}
}
}
}
}
if force {
if err = wr.Scrap(ctx, tablet.Alias, force, false); err != nil {
wr.Logger().Errorf("failed scrapping tablet %v: %v", tablet.Alias, err)
return err
}
if err := wr.ts.DeleteTablet(ctx, tablet.Alias); err != nil {
// we ignore this
wr.Logger().Errorf("failed deleting tablet %v: %v", tablet.Alias, err)
}
return topo.CreateTablet(ctx, wr.ts, tablet)
}
}
return err
}
// InitTablet initializes the tablet record if necessary.
func (agent *ActionAgent) InitTablet(port, gRPCPort int32) error {
// it should be either we have all three of init_keyspace,
// init_shard and init_tablet_type, or none.
if *initKeyspace == "" && *initShard == "" && *initTabletType == "" {
// not initializing the record
return nil
}
if *initKeyspace == "" || *initShard == "" || *initTabletType == "" {
return fmt.Errorf("either need all of init_keyspace, init_shard and init_tablet_type, or none")
}
// parse init_tablet_type
tabletType, err := topoproto.ParseTabletType(*initTabletType)
if err != nil {
return fmt.Errorf("invalid init_tablet_type %v: %v", *initTabletType, err)
}
if tabletType == topodatapb.TabletType_MASTER {
// We disallow MASTER, so we don't have to change
// shard.MasterAlias, and deal with the corner cases.
return fmt.Errorf("init_tablet_type cannot be master, use replica instead")
}
// parse and validate shard name
shard, _, err := topo.ValidateShardName(*initShard)
if err != nil {
return fmt.Errorf("cannot validate shard name %v: %v", *initShard, err)
}
// create a context for this whole operation
ctx, cancel := context.WithTimeout(agent.batchCtx, *initTimeout)
defer cancel()
// read the shard, create it if necessary
log.Infof("Reading shard record %v/%v", *initKeyspace, shard)
si, err := agent.TopoServer.GetOrCreateShard(ctx, *initKeyspace, shard)
if err != nil {
return fmt.Errorf("InitTablet cannot GetOrCreateShard shard: %v", err)
}
if si.MasterAlias != nil && topoproto.TabletAliasEqual(si.MasterAlias, agent.TabletAlias) {
// We're marked as master in the shard record, which could mean the master
// tablet process was just restarted. However, we need to check if a new
// master is in the process of taking over. In that case, it will let us
// know by forcibly updating the old master's tablet record.
oldTablet, err := agent.TopoServer.GetTablet(ctx, agent.TabletAlias)
switch err {
case topo.ErrNoNode:
// There's no existing tablet record, so we can assume
// no one has left us a message to step down.
tabletType = topodatapb.TabletType_MASTER
case nil:
if oldTablet.Type == topodatapb.TabletType_MASTER {
// We're marked as master in the shard record,
// and our existing tablet record agrees.
tabletType = topodatapb.TabletType_MASTER
}
default:
return fmt.Errorf("InitTablet failed to read existing tablet record: %v", err)
}
}
// See if we need to add the tablet's cell to the shard's cell list.
if !si.HasCell(agent.TabletAlias.Cell) {
si, err = agent.TopoServer.UpdateShardFields(ctx, *initKeyspace, shard, func(si *topo.ShardInfo) error {
if si.HasCell(agent.TabletAlias.Cell) {
// Someone else already did it.
return topo.ErrNoUpdateNeeded
}
si.Cells = append(si.Cells, agent.TabletAlias.Cell)
return nil
})
if err != nil {
return fmt.Errorf("couldn't add tablet's cell to shard record: %v", err)
}
}
log.Infof("Initializing the tablet for type %v", tabletType)
// figure out the hostname
hostname := *tabletHostname
if hostname != "" {
log.Infof("Using hostname: %v from -tablet_hostname flag.", hostname)
} else {
hostname, err := netutil.FullyQualifiedHostname()
if err != nil {
return err
}
log.Infof("Using detected machine hostname: %v To change this, fix your machine network configuration or override it with -tablet_hostname.", hostname)
}
// create and populate tablet record
tablet := &topodatapb.Tablet{
Alias: agent.TabletAlias,
Hostname: hostname,
PortMap: make(map[string]int32),
Keyspace: *initKeyspace,
Shard: *initShard,
Type: tabletType,
DbNameOverride: *initDbNameOverride,
Tags: initTags,
}
if port != 0 {
tablet.PortMap["vt"] = port
}
if gRPCPort != 0 {
tablet.PortMap["grpc"] = gRPCPort
}
if err := topo.TabletComplete(tablet); err != nil {
return fmt.Errorf("InitTablet TabletComplete failed: %v", err)
}
// Now try to create the record (it will also fix up the
// ShardReplication record if necessary).
err = agent.TopoServer.CreateTablet(ctx, tablet)
switch err {
case nil:
// It worked, we're good.
case topo.ErrNodeExists:
// The node already exists, will just try to update
// it. So we read it first.
oldTablet, err := agent.TopoServer.GetTablet(ctx, tablet.Alias)
if err != nil {
return fmt.Errorf("InitTablet failed to read existing tablet record: %v", err)
}
// Sanity check the keyspace and shard
if oldTablet.Keyspace != tablet.Keyspace || oldTablet.Shard != tablet.Shard {
return fmt.Errorf("InitTablet failed because existing tablet keyspace and shard %v/%v differ from the provided ones %v/%v", oldTablet.Keyspace, oldTablet.Shard, tablet.Keyspace, tablet.Shard)
}
// Then overwrite everything, ignoring version mismatch.
if err := agent.TopoServer.UpdateTablet(ctx, topo.NewTabletInfo(tablet, -1)); err != nil {
return fmt.Errorf("UpdateTablet failed: %v", err)
}
default:
return fmt.Errorf("CreateTablet failed: %v", err)
}
return nil
}