Log when command-line peer list overridden by persisted list

Reworked interface to router so we split setting up the initial list
from beginning to make connections - we shouldn't do that until
after router.Start().

prog/weaver/main.go

@@ -127,7 +127,6 @@ func main() {
 	}
 
 	Log.Println("Command line options:", options())
-	Log.Println("Command line peers:", peers)
 
 	if prof != "" {
 		p := *profile.CPUProfile
@@ -169,6 +168,10 @@ func main() {
 	router := weave.NewNetworkRouter(config, networkConfig, name, nickName, overlay, db)
 	Log.Println("Our name is", router.Ourself)
 
+	if peers, err = router.InitialPeers(peers); err != nil {
+		Log.Fatal("Unable to get initial peer set: ", err)
+	}
+
 	var dockerCli *docker.Client
 	if dockerAPI != "" {
 		dc, err := docker.NewClient(dockerAPI)
@@ -215,7 +218,7 @@ func main() {
 	}
 
 	router.Start()
-	if errors := router.SetInitialPeers(peers); len(errors) > 0 {
+	if errors := router.InitiateConnections(peers, false); len(errors) > 0 {
 		Log.Fatal(ErrorMessages(errors))
 	}
 

router/http.go

@@ -14,18 +14,16 @@ func (router *NetworkRouter) HandleHTTP(muxRouter *mux.Router) {
 		if err := r.ParseForm(); err != nil {
 			http.Error(w, fmt.Sprint("unable to parse form: ", err), http.StatusBadRequest)
 		}
-		if errors := router.ConnectionMaker.InitiateConnections(r.Form["peer"], r.FormValue("replace") == "true"); len(errors) > 0 {
+		if errors := router.InitiateConnections(r.Form["peer"], r.FormValue("replace") == "true"); len(errors) > 0 {
 			http.Error(w, common.ErrorMessages(errors), http.StatusBadRequest)
 		}
-		router.persistPeers()
 	})
 
 	muxRouter.Methods("POST").Path("/forget").HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if err := r.ParseForm(); err != nil {
 			http.Error(w, fmt.Sprint("unable to parse form: ", err), http.StatusBadRequest)
 		}
-		router.ConnectionMaker.ForgetConnections(r.Form["peer"])
-		router.persistPeers()
+		router.ForgetConnections(r.Form["peer"])
 	})
 
 }

router/network_router.go

@@ -217,11 +217,40 @@ func (router *NetworkRouter) persistPeers() {
 	}
 }
 
-func (router *NetworkRouter) SetInitialPeers(peers []string) []error {
-	// Load is a no-op if persistence entry has never been saved
-	if err := router.db.Load(peersIdent, &peers); err != nil {
-		return []error{err}
+func (router *NetworkRouter) InitiateConnections(peers []string, replace bool) []error {
+	errors := router.ConnectionMaker.InitiateConnections(peers, replace)
+	router.persistPeers()
+	return errors
+}
+
+func (router *NetworkRouter) ForgetConnections(peers []string) {
+	router.ConnectionMaker.ForgetConnections(peers)
+	router.persistPeers()
+}
+
+func (router *NetworkRouter) InitialPeers(peers []string) ([]string, error) {
+	var storedPeers []string
+	if err := router.db.Load(peersIdent, &storedPeers); err != nil {
+		return nil, err
 	}
-	log.Println("Initiating connections to peers:", peers)
-	return router.ConnectionMaker.InitiateConnections(peers, false)
+
+	if storedPeers != nil && !equal(peers, storedPeers) {
+		log.Println("Overriding initial peer list with stored list:", storedPeers)
+		peers = storedPeers
+	} else {
+		log.Println("Initial set of peers:", peers)
+	}
+	return peers, nil
+}
+
+func equal(a, b []string) bool {
+	if len(a) != len(b) {
+		return false
+	}
+	for i, v := range a {
+		if v != b[i] {
+			return false
+		}
+	}
+	return true
 }

test/162_persist_ipam_2_test.sh

@@ -14,7 +14,7 @@ launch_router_with_db() {
 run_on $HOST1 "sudo rm -f /tmp/test162-*"
 run_on $HOST2 "sudo rm -f /tmp/test162-*"
 
-launch_router_with_db $HOST1
+launch_router_with_db $HOST1 $HOST2
 launch_router_with_db $HOST2 $HOST1
 
 start_container $HOST1 --name=c1
@@ -31,13 +31,8 @@ start_container $HOST2 --name=c3
 C3=$(container_ip $HOST2 c3)
 assert_raises "[ $C3 != $C1 ]"
 
-stop_router_on $HOST2
-
-# Now start HOST1 with HOST2 down and see if it hangs when we launch a container
-launch_router_with_db $HOST1 $HOST2
-start_container $HOST1 --name=c4
-C4=$(container_ip $HOST1 c4)
-assert_raises "[ $C4 != $C1 ]"
-assert_raises "[ $C4 != $C3 ]"
+# Restart HOST1 and see if it remembers to connect to HOST2
+launch_router_with_db $HOST1
+assert_raises "exec_on $HOST2 c2 $PING $C1"
 
 end_suite