diff --git a/internal/app/repair.go b/internal/app/repair.go
index dae3751..99ad99f 100644
--- a/internal/app/repair.go
+++ b/internal/app/repair.go
@@ -88,6 +88,23 @@ func (app *App) repairReplica(node *redis.Node, masterState, state *HostState, m
 				app.logger.Error(fmt.Sprintf("Unable to make %s replica of %s", node.FQDN(), master), "error", err)
 			}
 		case modeCluster:
+			alone, err := node.IsClusterNodeAlone(app.ctx)
+			if err != nil {
+				app.logger.Error(fmt.Sprintf("Unable to check if %s is alone", node.FQDN()), "error", err)
+				return
+			}
+			if alone {
+				masterIP, err := masterNode.GetIP()
+				if err != nil {
+					app.logger.Error(fmt.Sprintf("Unable to make %s replica of %s", node.FQDN(), master), "error", err)
+					return
+				}
+				err = node.ClusterMeet(app.ctx, masterIP, app.config.Redis.Port, app.config.Redis.ClusterBusPort)
+				if err != nil {
+					app.logger.Error(fmt.Sprintf("Unable to make %s meet with master %s at %s:%d:%d", node.FQDN(), master, masterIP, app.config.Redis.Port, app.config.Redis.ClusterBusPort), "error", err)
+					return
+				}
+			}
 			masterID, err := masterNode.ClusterGetID(app.ctx)
 			if err != nil {
 				app.logger.Error(fmt.Sprintf("Unable to get cluster id of %s", master), "error", err.Error())
diff --git a/internal/config/config.go b/internal/config/config.go
index 93ccaa2..f88606b 100644
--- a/internal/config/config.go
+++ b/internal/config/config.go
@@ -16,6 +16,7 @@ import (
 // RedisConfig contains redis connection info and params
 type RedisConfig struct {
 	Port                    int           `yaml:"port"`
+	ClusterBusPort          int           `yaml:"cluster_bus_port"`
 	UseTLS                  bool          `yaml:"use_tls"`
 	TLSCAPath               string        `yaml:"tls_ca_path"`
 	AuthUser                string        `yaml:"auth_user"`
@@ -46,6 +47,7 @@ type RedisRenamesConfig struct {
 	ClusterFailover  string `yaml:"cluster_failover"`
 	ClusterMyID      string `yaml:"cluster_myid"`
 	ClusterReplicate string `yaml:"cluster_replicate"`
+	ClusterMeet      string `yaml:"cluster_meet"`
 	Config           string `yaml:"config"`
 	ReplicaOf        string `yaml:"replicaof"`
 }
@@ -90,6 +92,7 @@ type Config struct {
 func DefaultRedisConfig() RedisConfig {
 	return RedisConfig{
 		Port:                    6379,
+		ClusterBusPort:          16379,
 		UseTLS:                  false,
 		TLSCAPath:               "",
 		AuthUser:                "",
@@ -122,6 +125,7 @@ func DefaultRedisRenamesConfig() RedisRenamesConfig {
 		ClusterFailover:  "FAILOVER",
 		ClusterMyID:      "MYID",
 		ClusterReplicate: "REPLICATE",
+		ClusterMeet:      "MEET",
 		Config:           "CONFIG",
 		ReplicaOf:        "REPLICAOF",
 	}
diff --git a/internal/redis/node.go b/internal/redis/node.go
index bd30083..81154ee 100644
--- a/internal/redis/node.go
+++ b/internal/redis/node.go
@@ -519,3 +519,20 @@ func (n *Node) ClusterPromoteTakeover(ctx context.Context) error {
 	cmd := n.conn.Do(ctx, n.config.Renames.Cluster, n.config.Renames.ClusterFailover, "TAKEOVER")
 	return cmd.Err()
 }
+
+// IsClusterNodeAlone checks if node sees only itself
+func (n *Node) IsClusterNodeAlone(ctx context.Context) (bool, error) {
+	cmd := n.conn.ClusterNodes(ctx)
+	err := cmd.Err()
+	if err != nil {
+		return false, err
+	}
+	lines := strings.Split(cmd.Val(), "\n")
+	return len(lines) == 1, nil
+}
+
+// ClusterMeet makes replica join the cluster
+func (n *Node) ClusterMeet(ctx context.Context, addr string, port, clusterBusPort int) error {
+	cmd := n.conn.Do(ctx, n.config.Renames.Cluster, n.config.Renames.ClusterMeet, addr, strconv.Itoa(port), strconv.Itoa(clusterBusPort))
+	return cmd.Err()
+}
diff --git a/tests/features/05_cluster_replication_fix.feature b/tests/features/05_cluster_replication_fix.feature
index 6b17ef4..6c1dadf 100644
--- a/tests/features/05_cluster_replication_fix.feature
+++ b/tests/features/05_cluster_replication_fix.feature
@@ -154,3 +154,33 @@ Feature: Cluster mode broken replication fix
         When I break replication on host "redis3"
         Then redis host "redis3" should become replica of "redis1" within "15" seconds
         And replication on redis host "redis3" should run fine within "60" seconds
+
+    Scenario: Cluster lone node is joined in cluster back
+        Given clustered shard is up and running
+        Then redis host "redis1" should be master
+        And redis host "redis2" should become replica of "redis1" within "15" seconds
+        And replication on redis host "redis2" should run fine within "15" seconds
+        And redis host "redis3" should become replica of "redis1" within "15" seconds
+        And replication on redis host "redis3" should run fine within "15" seconds
+        And zookeeper node "/test/active_nodes" should match json_exactly within "30" seconds
+        """
+            ["redis1","redis2","redis3"]
+        """
+        When I run command on host "redis3"
+        """
+            rm -f /etc/redis/cluster.conf
+        """
+        And I run command on host "redis3"
+        """
+            supervisorctl signal KILL redis
+        """
+        And I run command on host "redis3"
+        """
+            supervisorctl start redis
+        """
+        Then redis host "redis3" should become replica of "redis1" within "15" seconds
+        And replication on redis host "redis3" should run fine within "15" seconds
+        And zookeeper node "/test/active_nodes" should match json_exactly within "30" seconds
+        """
+            ["redis1","redis2","redis3"]
+        """