pd-client: add some retries when initializing cluster id

TiDB can not start some times if it starts at the same time with PD,
because PD may need some time before serving requests.

pd-client/rpc_worker.go

@@ -29,7 +29,10 @@ import (
 	"github.com/twinj/uuid"
 )
 
-const maxPipelineRequest = 10000
+const (
+	maxPipelineRequest    = 10000
+	maxInitClusterRetries = 300
+)
 
 // errInvalidResponse represents response message is invalid.
 var errInvalidResponse = errors.New("invalid response")
@@ -222,13 +225,17 @@ func (w *rpcWorker) initClusterID() error {
 	}
 	defer conn.Close()
 
-	clusterID, err := w.getClusterID(conn.ReadWriter)
-	if err != nil {
-		return errors.Trace(err)
+	for i := 0; i < maxInitClusterRetries; i++ {
+		clusterID, err := w.getClusterID(conn.ReadWriter)
+		if err == nil {
+			w.clusterID = clusterID
+			return nil
+		}
+		log.Errorf("[pd] failed to get cluster id: %v", err)
+		time.Sleep(time.Second)
 	}
 
-	w.clusterID = clusterID
-	return nil
+	return errors.New("failed to get cluster id")
 }
 
 func (w *rpcWorker) getClusterID(conn *bufio.ReadWriter) (uint64, error) {