Permit agent to run even when VPP stats are unavailable (#1712)

plugins/govppmux/plugin_impl_govppmux.go

@@ -173,9 +173,10 @@ func (p *Plugin) Init() (err error) {
 		}
 		err := p.startProxy(NewVppAdapter(address, useShm), NewStatsAdapter(statsSocket))
 		if err != nil {
-			return err
+			p.Log.Warnf("VPP proxy failed to start: %v", err)
+		} else {
+			p.Log.Infof("VPP proxy ready")
 		}
-		p.Log.Infof("VPP proxy ready")
 	}
 
 	// register REST API handlers

plugins/telemetry/stats_poller.go

@@ -31,6 +31,9 @@ func (s *statsPollerServer) PollStats(req *configurator.PollStatsRequest, svr co
 	if req.GetPeriodSec() == 0 && req.GetNumPolls() > 1 {
 		return status.Error(codes.InvalidArgument, "period must be > 0 if number of polls is > 1")
 	}
+	if s.handler == nil {
+		return status.Errorf(codes.Unavailable, "VPP telemetry handler not available")
+	}
 
 	ctx := svr.Context()
 

plugins/telemetry/telemetry.go

@@ -159,10 +159,11 @@ func (p *Plugin) AfterInit() error {
 func (p *Plugin) setupStatsPoller() error {
 	h := vppcalls.CompatibleTelemetryHandler(p.VPP)
 	if h == nil {
-		return fmt.Errorf("VPP telemetry handler unavailable")
+		p.Log.Warnf("VPP telemetry handler unavailable")
+	} else {
+		p.statsPollerServer.handler = h
 	}
-	p.statsPollerServer.handler = h
-	p.ifIndex = p.IfPlugin.GetInterfaceIndex()
+	p.statsPollerServer.ifIndex = p.IfPlugin.GetInterfaceIndex()
 
 	if p.GRPC != nil && p.GRPC.GetServer() != nil {
 		configurator.RegisterStatsPollerServiceServer(p.GRPC.GetServer(), &p.statsPollerServer)

plugins/vpp/ifplugin/interface_state.go

@@ -210,7 +210,13 @@ func (c *InterfaceStateUpdater) startReadingCounters(ctx context.Context) {
 	for {
 		select {
 		case <-tick.C:
-			c.doInterfaceStatsRead()
+			statsClient := c.vppClient.Stats()
+			if statsClient == nil {
+				c.log.Warnf("VPP stats client not available")
+				// TODO: use retry with backoff instead of returning here
+				return
+			}
+			c.doInterfaceStatsRead(statsClient)
 
 		case <-ctx.Done():
 			c.log.Debug("Interface state VPP periodic polling stopped")
@@ -276,7 +282,7 @@ func (c *InterfaceStateUpdater) doUpdatesIfStateDetails() {
 }
 
 // doInterfaceStatsRead dumps statistics using interface filter and processes them
-func (c *InterfaceStateUpdater) doInterfaceStatsRead() {
+func (c *InterfaceStateUpdater) doInterfaceStatsRead(statsClient govppapi.StatsProvider) {
 	c.access.Lock()
 	defer c.access.Unlock()
 
@@ -286,7 +292,7 @@ func (c *InterfaceStateUpdater) doInterfaceStatsRead() {
 		return
 	}
 
-	err := c.vppClient.Stats().GetInterfaceStats(&c.ifStats)
+	err := statsClient.GetInterfaceStats(&c.ifStats)
 	if err != nil {
 		// TODO add some counter to prevent it log forever
 		c.log.Errorf("failed to read statistics data: %v", err)