add test and fix

coordinator/provider/coordinator.go

@@ -3,7 +3,6 @@ package provider
 import (
 	"context"
 	"crypto/tls"
-	"fmt"
 	"net"
 	"time"
 
@@ -291,7 +290,7 @@ func (qc *qdbCoordinator) lockCoordinator(ctx context.Context, initialRouter boo
 		}
 		router := &topology.Router{
 			ID:      uuid.NewString(),
-			Address: fmt.Sprintf("%s:%s", config.RouterConfig().Host, config.RouterConfig().GrpcApiPort),
+			Address: net.JoinHostPort(config.RouterConfig().Host, config.RouterConfig().GrpcApiPort),
 			State:   qdb.OPENED,
 		}
 		if err := qc.RegisterRouter(ctx, router); err != nil {
@@ -300,7 +299,7 @@ func (qc *qdbCoordinator) lockCoordinator(ctx context.Context, initialRouter boo
 		if err := qc.SyncRouterMetadata(ctx, router); err != nil {
 			spqrlog.Zero.Error().Err(err).Msg("sync router metadata when locking coordinator")
 		}
-		if err := qc.UpdateCoordinator(ctx, fmt.Sprintf("%s:%s", config.CoordinatorConfig().Host, config.CoordinatorConfig().GrpcApiPort)); err != nil {
+		if err := qc.UpdateCoordinator(ctx, net.JoinHostPort(config.CoordinatorConfig().Host, config.CoordinatorConfig().GrpcApiPort)); err != nil {
 			return false
 		}
 		return true

qdb/etcdqdb.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"net"
 	"path"
 	"sort"
 	"sync"
@@ -512,7 +513,7 @@ func (q *EtcdQDB) TryCoordinatorLock(ctx context.Context) error {
 		return err
 	}
 
-	op := clientv3.OpPut(coordLockKey, config.CoordinatorConfig().Host, clientv3.WithLease(clientv3.LeaseID(leaseGrantResp.ID)))
+	op := clientv3.OpPut(coordLockKey, net.JoinHostPort(config.CoordinatorConfig().Host, config.CoordinatorConfig().GrpcApiPort), clientv3.WithLease(clientv3.LeaseID(leaseGrantResp.ID)))
 	tx := q.cli.Txn(ctx).If(clientv3util.KeyMissing(coordLockKey)).Then(op)
 	stat, err := tx.Commit()
 	if err != nil {

test/feature/features/init_etcd.feature

@@ -0,0 +1,48 @@
+Feature: Initialize router metadata from Etcd
+    Background:
+        #
+        # Run routers with coordinators
+        # Stop all coordinators
+        #
+        Given cluster environment is
+        """
+        ROUTER_CONFIG=/spqr/test/feature/conf/router_with_coordinator.yaml
+        COORDINATOR_CONFIG=/spqr/test/feature/conf/router_coordinator.yaml
+        COORDINATOR_CONFIG_2=/spqr/test/feature/conf/router_coordinator_2.yaml
+        """
+        Given cluster is up and running
+        And host "coordinator2" is stopped
+        When I run SQL on host "router-admin"
+        """
+        UNREGISTER ROUTER ALL;
+        REGISTER ROUTER r1 ADDRESS regress_router::7000;
+        REGISTER ROUTER r2 ADDRESS regress_router_2::7000;
+        """
+        Then command return code should be "0"
+        And host "router" is stopped
+        And host "router2" is stopped
+
+    Scenario: Router initialize its metadata from Etcd when no coodinator alive
+        When I run SQL on host "router-admin"
+        """
+        CREATE DISTRIBUTION ds1 COLUMN TYPES integer;
+        CREATE KEY RANGE krid1 FROM 19 ROUTE TO sh1 FOR DISTRIBUTION ds1;
+        """
+        Then command return code should be "0"
+
+        When host "coordinator" is stopped
+        And host "router" is started
+
+        When I run SQL on host "router-admin"
+        """
+        SHOW key_ranges
+        """
+        Then SQL result should match json_exactly
+        """
+        [{
+            "Key range ID":"krid1",
+            "Distribution ID":"ds1",
+            "Lower bound":"19",
+            "Shard ID":"sh1"
+        }]
+        """