prevent capacityd and provisiond to start if networkd is not ready

fixes #632

Author: zaibon

cmds/capacityd/main.go

@@ -26,6 +26,7 @@ const module = "monitor"
 func cap(ctx context.Context, client zbus.Client) {
 	storage := stubs.NewStorageModuleStub(client)
 	identity := stubs.NewIdentityManagerStub(client)
+	network := stubs.NewNetworkerStub(client)
 	cl, err := bcdbClient()
 	if err != nil {
 		log.Fatal().Err(err).Msg("failed to connect to bcdb backend")
@@ -34,6 +35,17 @@ func cap(ctx context.Context, client zbus.Client) {
 	// call this now so we block here until identityd is ready to serve us
 	nodeID := identity.NodeID().Identity()
 
+	// block until networkd is ready to serve request from zbus
+	// this is used to prevent uptime and online status to the explorer if the node is not in a fully ready
+	// https://github.com/threefoldtech/zos/issues/632
+	bo := backoff.NewExponentialBackOff()
+	bo.MaxElapsedTime = 0
+	backoff.RetryNotify(func() error {
+		return network.Ready()
+	}, bo, func(err error, d time.Duration) {
+		log.Error().Err(err).Msgf("networkd is not ready yet")
+	})
+
 	r := capacity.NewResourceOracle(storage)
 
 	log.Info().Msg("inspect hardware resources")
@@ -75,7 +87,7 @@ func cap(ctx context.Context, client zbus.Client) {
 		log.Info().Msg("sends capacity detail to BCDB")
 		return cl.NodeSetCapacity(nodeID, ru, *dmi, disks, hypervisor)
 	}
-	bo := backoff.NewExponentialBackOff()
+	bo = backoff.NewExponentialBackOff()
 	bo.MaxElapsedTime = 0 // retry forever
 	backoff.RetryNotify(setCapacity, bo, func(err error, d time.Duration) {
 		log.Error().

cmds/provisiond/main.go

@@ -7,6 +7,7 @@ import (
 	"path/filepath"
 	"time"
 
+	"github.com/cenkalti/backoff/v3"
 	"github.com/threefoldtech/zos/pkg"
 	"github.com/threefoldtech/zos/pkg/app"
 	"github.com/threefoldtech/zos/pkg/environment"
@@ -92,6 +93,18 @@ func main() {
 	identity := stubs.NewIdentityManagerStub(zbusCl)
 	nodeID := identity.NodeID()
 
+	// block until networkd is ready to serve request from zbus
+	// this is used to prevent uptime and online status to the explorer if the node is not in a fully ready
+	// https://github.com/threefoldtech/zos/issues/632
+	network := stubs.NewNetworkerStub(zbusCl)
+	bo := backoff.NewExponentialBackOff()
+	bo.MaxElapsedTime = 0
+	backoff.RetryNotify(func() error {
+		return network.Ready()
+	}, bo, func(err error, d time.Duration) {
+		log.Error().Err(err).Msgf("networkd is not ready yet")
+	})
+
 	// to get reservation from tnodb
 	e, err := app.ExplorerClient()
 	if err != nil {

pkg/network.go

@@ -20,6 +20,10 @@ type Member struct {
 
 //Networker is the interface for the network module
 type Networker interface {
+	// Ready return nil is networkd is ready to operate
+	// This function is used by other deamon to test if networkd is done booting
+	Ready() error
+
 	// Create a new network resource
 	CreateNR(Network) (string, error)
 	// Delete a network resource

pkg/network/networker.go

@@ -210,6 +210,10 @@ func validatePeer(p pkg.Peer) error {
 	return nil
 }
 
+func (n *networker) Ready() error {
+	return nil
+}
+
 func (n *networker) Join(networkdID pkg.NetID, containerID string, addrs []string, publicIP6 bool) (join pkg.Member, err error) {
 	// TODO:
 	// 1- Make sure this network id is actually deployed

pkg/stubs/network_stub.go

@@ -168,6 +168,19 @@ func (s *NetworkerStub) PublicAddresses(ctx context.Context) (<-chan pkg.Netlink
 	return ch, nil
 }
 
+func (s *NetworkerStub) Ready() (ret0 error) {
+	args := []interface{}{}
+	result, err := s.client.Request(s.module, s.object, "Ready", args...)
+	if err != nil {
+		panic(err)
+	}
+	ret0 = new(zbus.RemoteError)
+	if err := result.Unmarshal(0, &ret0); err != nil {
+		panic(err)
+	}
+	return
+}
+
 func (s *NetworkerStub) RemoveTap(arg0 pkg.NetID) (ret0 error) {
 	args := []interface{}{arg0}
 	result, err := s.client.Request(s.module, s.object, "RemoveTap", args...)