// Ensure the shard writer returns an error when we can't get a connection.
func TestShardWriter_Write_PoolMax(t *testing.T) {
ts := newTestWriteService(writeShardSlow)
s := cluster.NewService(cluster.Config{
ShardWriterTimeout: toml.Duration(100 * time.Millisecond),
})
s.Listener = ts.muxln
s.TSDBStore = &ts.TSDBStore
if err := s.Open(); err != nil {
t.Fatal(err)
}
defer s.Close()
defer ts.Close()
w := cluster.NewShardWriter(100*time.Millisecond, 1)
w.MetaClient = &metaClient{host: ts.ln.Addr().String()}
now := time.Now()
shardID := uint64(1)
ownerID := uint64(2)
var points []models.Point
points = append(points, models.MustNewPoint(
"cpu", models.Tags{"host": "server01"}, map[string]interface{}{"value": int64(100)}, now,
))
go w.WriteShard(shardID, ownerID, points)
time.Sleep(time.Millisecond)
if err := w.WriteShard(shardID, ownerID, points); err == nil || err.Error() != "timed out waiting for free connection" {
t.Fatalf("unexpected error: %v", err)
}
}
// Ensure the shard writer returns an error when dialing times out.
func TestShardWriter_Write_ErrDialTimeout(t *testing.T) {
ts := newTestWriteService(nil)
ts.TSDBStore.WriteToShardFn = ts.writeShardSuccess
s := cluster.NewService(cluster.Config{})
s.Listener = ts.muxln
s.TSDBStore = &ts.TSDBStore
if err := s.Open(); err != nil {
t.Fatal(err)
}
defer s.Close()
defer ts.Close()
// Zero timeout set to support all platforms.
w := cluster.NewShardWriter(0, 1)
w.MetaClient = &metaClient{host: ts.ln.Addr().String()}
now := time.Now()
shardID := uint64(1)
ownerID := uint64(2)
var points []models.Point
points = append(points, models.MustNewPoint(
"cpu", models.Tags{"host": "server01"}, map[string]interface{}{"value": int64(100)}, now,
))
if err, exp := w.WriteShard(shardID, ownerID, points), "i/o timeout"; err == nil || !strings.Contains(err.Error(), exp) {
t.Fatalf("expected error %v, to contain %s", err, exp)
}
}
// Ensure the shard writer returns an error when the server fails to accept the write.
func TestShardWriter_WriteShard_Error(t *testing.T) {
ts := newTestWriteService(writeShardFail)
s := cluster.NewService(cluster.Config{})
s.Listener = ts.muxln
s.TSDBStore = &ts.TSDBStore
if err := s.Open(); err != nil {
t.Fatal(err)
}
defer s.Close()
defer ts.Close()
w := cluster.NewShardWriter(time.Minute, 1)
w.MetaClient = &metaClient{host: ts.ln.Addr().String()}
now := time.Now()
shardID := uint64(1)
ownerID := uint64(2)
var points []models.Point
points = append(points, models.MustNewPoint(
"cpu", models.Tags{"host": "server01"}, map[string]interface{}{"value": int64(100)}, now,
))
if err := w.WriteShard(shardID, ownerID, points); err == nil || err.Error() != "error code 1: write shard 1: failed to write" {
t.Fatalf("unexpected error: %v", err)
}
}
// Ensure the shard writer can successful write a multiple requests.
func TestShardWriter_WriteShard_Multiple(t *testing.T) {
ts := newTestWriteService(nil)
ts.TSDBStore.WriteToShardFn = ts.writeShardSuccess
s := cluster.NewService(cluster.Config{})
s.Listener = ts.muxln
s.TSDBStore = &ts.TSDBStore
if err := s.Open(); err != nil {
t.Fatal(err)
}
defer s.Close()
defer ts.Close()
w := cluster.NewShardWriter(time.Minute, 1)
w.MetaClient = &metaClient{host: ts.ln.Addr().String()}
// Build a single point.
now := time.Now()
var points []models.Point
points = append(points, models.MustNewPoint("cpu", models.Tags{"host": "server01"}, map[string]interface{}{"value": int64(100)}, now))
// Write to shard twice and close.
if err := w.WriteShard(1, 2, points); err != nil {
t.Fatal(err)
} else if err := w.WriteShard(1, 2, points); err != nil {
t.Fatal(err)
} else if err := w.Close(); err != nil {
t.Fatal(err)
}
// Validate response.
responses, err := ts.ResponseN(1)
if err != nil {
t.Fatal(err)
} else if responses[0].shardID != 1 {
t.Fatalf("unexpected shard id: %d", responses[0].shardID)
}
// Validate point.
if p := responses[0].points[0]; p.Name() != "cpu" {
t.Fatalf("unexpected name: %s", p.Name())
} else if p.Fields()["value"] != int64(100) {
t.Fatalf("unexpected 'value' field: %d", p.Fields()["value"])
} else if p.Tags()["host"] != "server01" {
t.Fatalf("unexpected 'host' tag: %s", p.Tags()["host"])
} else if p.Time().UnixNano() != now.UnixNano() {
t.Fatalf("unexpected time: %s", p.Time())
}
}
// Ensure the shard writer returns an error when reading times out.
func TestShardWriter_Write_ErrReadTimeout(t *testing.T) {
ln, err := net.Listen("tcp", "127.0.0.1:0")
if err != nil {
t.Fatal(err)
}
w := cluster.NewShardWriter(time.Millisecond, 1)
w.MetaClient = &metaClient{host: ln.Addr().String()}
now := time.Now()
shardID := uint64(1)
ownerID := uint64(2)
var points []models.Point
points = append(points, models.MustNewPoint(
"cpu", models.Tags{"host": "server01"}, map[string]interface{}{"value": int64(100)}, now,
))
if err := w.WriteShard(shardID, ownerID, points); err == nil || !strings.Contains(err.Error(), "i/o timeout") {
t.Fatalf("unexpected error: %s", err)
}
}
// NewServer returns a new instance of Server built from a config.
func NewServer(c *Config, buildInfo *BuildInfo) (*Server, error) {
// We need to ensure that a meta directory always exists even if
// we don't start the meta store. node.json is always stored under
// the meta directory.
if err := os.MkdirAll(c.Meta.Dir, 0777); err != nil {
return nil, fmt.Errorf("mkdir all: %s", err)
}
// 0.10-rc1 and prior would sometimes put the node.json at the root
// dir which breaks backup/restore and restarting nodes. This moves
// the file from the root so it's always under the meta dir.
oldPath := filepath.Join(filepath.Dir(c.Meta.Dir), "node.json")
newPath := filepath.Join(c.Meta.Dir, "node.json")
if _, err := os.Stat(oldPath); err == nil {
if err := os.Rename(oldPath, newPath); err != nil {
return nil, err
}
}
nodeAddr, err := meta.DefaultHost(DefaultHostname, c.Meta.HTTPBindAddress)
if err != nil {
return nil, err
}
// load the node information
metaAddresses := []string{nodeAddr}
if !c.Meta.Enabled {
metaAddresses = c.Meta.JoinPeers
}
node, err := influxdb.LoadNode(c.Meta.Dir, metaAddresses)
if err != nil {
if !os.IsNotExist(err) {
return nil, err
} else {
node = influxdb.NewNode(c.Meta.Dir, metaAddresses)
}
}
// In 0.10.0 bind-address got moved to the top level. Check
// The old location to keep things backwards compatible
bind := c.BindAddress
if c.Meta.BindAddress != "" {
bind = c.Meta.BindAddress
}
if !c.Data.Enabled && !c.Meta.Enabled {
return nil, fmt.Errorf("must run as either meta node or data node or both")
}
httpBindAddress, err := meta.DefaultHost(DefaultHostname, c.HTTPD.BindAddress)
if err != nil {
return nil, err
}
tcpBindAddress, err := meta.DefaultHost(DefaultHostname, bind)
if err != nil {
return nil, err
}
s := &Server{
buildInfo: *buildInfo,
err: make(chan error),
closing: make(chan struct{}),
BindAddress: bind,
Node: node,
Monitor: monitor.New(c.Monitor),
reportingDisabled: c.ReportingDisabled,
joinPeers: c.Meta.JoinPeers,
metaUseTLS: c.Meta.HTTPSEnabled,
httpAPIAddr: httpBindAddress,
httpUseTLS: c.HTTPD.HTTPSEnabled,
tcpAddr: tcpBindAddress,
config: c,
}
if c.Meta.Enabled {
s.MetaService = meta.NewService(c.Meta)
}
if c.Data.Enabled {
s.TSDBStore = tsdb.NewStore(c.Data.Dir)
s.TSDBStore.EngineOptions.Config = c.Data
// Copy TSDB configuration.
s.TSDBStore.EngineOptions.EngineVersion = c.Data.Engine
s.TSDBStore.EngineOptions.MaxWALSize = c.Data.MaxWALSize
s.TSDBStore.EngineOptions.WALFlushInterval = time.Duration(c.Data.WALFlushInterval)
s.TSDBStore.EngineOptions.WALPartitionFlushDelay = time.Duration(c.Data.WALPartitionFlushDelay)
// Initialize query executor.
s.QueryExecutor = tsdb.NewQueryExecutor()
s.QueryExecutor.Store = s.TSDBStore
s.QueryExecutor.MonitorStatementExecutor = &monitor.StatementExecutor{Monitor: s.Monitor}
s.QueryExecutor.QueryLogEnabled = c.Data.QueryLogEnabled
// Set the shard writer
s.ShardWriter = cluster.NewShardWriter(time.Duration(c.Cluster.ShardWriterTimeout),
c.Cluster.MaxRemoteWriteConnections)
// Create the hinted handoff service
s.HintedHandoff = hh.NewService(c.HintedHandoff, s.ShardWriter, s.MetaClient)
s.HintedHandoff.Monitor = s.Monitor
// Create the Subscriber service
s.Subscriber = subscriber.NewService(c.Subscriber)
// Initialize points writer.
s.PointsWriter = cluster.NewPointsWriter()
s.PointsWriter.WriteTimeout = time.Duration(c.Cluster.WriteTimeout)
s.PointsWriter.TSDBStore = s.TSDBStore
s.PointsWriter.ShardWriter = s.ShardWriter
s.PointsWriter.HintedHandoff = s.HintedHandoff
s.PointsWriter.Subscriber = s.Subscriber
s.PointsWriter.Node = s.Node
// needed for executing INTO queries.
s.QueryExecutor.IntoWriter = s.PointsWriter
// Initialize the monitor
s.Monitor.Version = s.buildInfo.Version
s.Monitor.Commit = s.buildInfo.Commit
s.Monitor.Branch = s.buildInfo.Branch
s.Monitor.BuildTime = s.buildInfo.Time
s.Monitor.PointsWriter = s.PointsWriter
}
return s, nil
}