EDGECLOUD-5230: Envoy crashing for UDP packet size larger than 1500 (#…

…1443)

* add debug cmd + use envoy version 1.18.x + update envoy yaml to use v3 api

cloud-resource-manager/crmutil/controller-data.go

@@ -112,6 +112,10 @@ func NewControllerData(pf platform.Platform, key *edgeproto.CloudletKey, nodeMgr
 	cd.updateVMWorkers.Init("vmpool-updatevm", cd.UpdateVMPool)
 	cd.updateTrustPolicyKeyworkers.Init("update-TrustPolicy", cd.UpdateTrustPolicy)
 	cd.settings = *edgeproto.GetDefaultSettings()
+
+	// debug functions
+	nodeMgr.Debug.AddDebugFunc("envoyversioncmd", cd.GetClusterEnvoyVersion)
+
 	return cd
 }
 

cloud-resource-manager/crmutil/exec.go

@@ -44,6 +44,78 @@ func (s *ExecReqHandler) RecvExecRequest(ctx context.Context, msg *edgeproto.Exe
 	}()
 }
 
+type EnvoyContainerVersion struct {
+	ContainerName string
+	EnvoyVersion  string
+	Error         string
+}
+
+type RootLBEnvoyVersion struct {
+	NodeType        string
+	NodeName        string
+	EnvoyContainers []EnvoyContainerVersion
+}
+
+func (cd *ControllerData) GetClusterEnvoyVersion(ctx context.Context, req *edgeproto.DebugRequest) string {
+	clusterInsts := []edgeproto.ClusterInst{}
+	cd.ClusterInstCache.Mux.Lock()
+	for _, v := range cd.ClusterInstCache.Objs {
+		clusterInsts = append(clusterInsts, *v.Obj)
+	}
+	cd.ClusterInstCache.Mux.Unlock()
+	nodes, err := cd.platform.ListCloudletMgmtNodes(ctx, clusterInsts, nil)
+	if err != nil {
+		return fmt.Sprintf("unable to get list of cluster nodes, %v", err)
+	}
+	if len(nodes) == 0 {
+		return fmt.Sprintf("no nodes found")
+	}
+	nodeVersions := []RootLBEnvoyVersion{}
+	for _, node := range nodes {
+		if !strings.Contains(node.Type, "rootlb") {
+			continue
+		}
+		client, err := cd.platform.GetNodePlatformClient(ctx, &node)
+		if err != nil {
+			return fmt.Sprintf("failed to get ssh client for node %s, %v", node.Name, err)
+		}
+		out, err := client.Output(`docker ps --format "{{.Names}}" --filter name="^envoy"`)
+		if err != nil {
+			log.SpanLog(ctx, log.DebugLevelInfra, "failed to find envoy containers on rootlb", "rootlb", node, "err", err, "out", out)
+			return fmt.Sprintf("failed to find envoy containers on rootlb %s, %v", node.Name, err)
+		}
+		nodeVersion := RootLBEnvoyVersion{
+			NodeType: node.Type,
+			NodeName: node.Name,
+		}
+		for _, name := range strings.Split(out, "\n") {
+			name = strings.TrimSpace(name)
+			if name == "" {
+				continue
+			}
+			envoyContainerVers := EnvoyContainerVersion{
+				ContainerName: name,
+			}
+			out, err := client.Output(`docker exec -it %s envoy --version`)
+			if err != nil {
+				log.SpanLog(ctx, log.DebugLevelInfra, "failed to find envoy container version on rootlb", "rootlb", node, "container", name, "err", err, "out", out)
+				envoyContainerVers.Error = err.Error()
+				nodeVersion.EnvoyContainers = append(nodeVersion.EnvoyContainers, envoyContainerVers)
+				continue
+			}
+			version := strings.TrimSpace(out)
+			envoyContainerVers.EnvoyVersion = version
+			nodeVersion.EnvoyContainers = append(nodeVersion.EnvoyContainers, envoyContainerVers)
+		}
+		nodeVersions = append(nodeVersions, nodeVersion)
+	}
+	out, err := json.Marshal(nodeVersions)
+	if err != nil {
+		return fmt.Sprintf("Failed to marshal node versions: %s, %v", string(out), err)
+	}
+	return string(out)
+}
+
 func (cd *ControllerData) ProcessExecReq(ctx context.Context, req *edgeproto.ExecRequest) (reterr error) {
 	var err error
 

cloud-resource-manager/proxy/envoy.go

@@ -195,6 +195,7 @@ func createEnvoyYaml(ctx context.Context, client ssh.Client, yamldir, name, list
 					ListenIP:    listenIP,
 					BackendIP:   backendIP,
 					BackendPort: internalPort,
+					MaxPktSize:  p.MaxPktSize,
 				}
 				udpconns, err := getUDPConcurrentConnections()
 				if err != nil {
@@ -250,12 +251,14 @@ static_resources:
     filter_chains:
     - filters:
       - name: envoy.filters.network.tcp_proxy
-        config:
+        typed_config:
+          '@type': type.googleapis.com/envoy.extensions.filters.network.tcp_proxy.v3.TcpProxy
           stat_prefix: ingress_tcp
           cluster: backend{{.BackendPort}}
           access_log:
             - name: envoy.access_loggers.file
-              config:
+              typed_config:
+                '@type': type.googleapis.com/envoy.extensions.access_loggers.file.v3.FileAccessLog
                 path: /tmp/access.log
                 json_format: {
                   "start_time": "%START_TIME%",
@@ -283,12 +286,21 @@ static_resources:
         protocol: UDP
         address: {{.ListenIP}}
         port_value: {{.ListenPort}}
+    {{if ne .MaxPktSize 0 -}}
+    udp_listener_config:
+      downstream_socket_config:
+        max_rx_datagram_size: {{.MaxPktSize}}
+    {{- end}}
     listener_filters:
       name: envoy.filters.udp_listener.udp_proxy
       typed_config:
         '@type': type.googleapis.com/envoy.extensions.filters.udp.udp_proxy.v3.UdpProxyConfig
         stat_prefix: downstream{{.BackendPort}}
         cluster: udp_backend{{.BackendPort}}
+    {{if ne .MaxPktSize 0 -}}
+        upstream_socket_config:
+          max_rx_datagram_size: {{.MaxPktSize}}
+    {{- end}}
     reuse_port: true
   {{- end}}
   clusters:
@@ -300,10 +312,15 @@ static_resources:
         thresholds:
             max_connections: {{.ConcurrentConns}}
     lb_policy: round_robin
-    hosts:
-    - socket_address:
-        address: {{.BackendIP}}
-        port_value: {{.BackendPort}}
+    load_assignment:
+      cluster_name: backend{{.BackendPort}}
+      endpoints:
+        lb_endpoints:
+        - endpoint:
+            address:
+              socket_address:
+                address: {{.BackendIP}}
+                port_value: {{.BackendPort}}
     {{if .HealthCheck -}}
     health_checks:
       - timeout: 1s

cloud-resource-manager/proxy/nginx.go

@@ -292,6 +292,7 @@ type UDPSpecDetail struct {
 	BackendIP        string
 	BackendPort      int32
 	ConcurrentConns  uint64
+	MaxPktSize       int64
 }
 
 var nginxConf = `

cloudcommon/names.go

@@ -135,7 +135,7 @@ const MaxClusterNameLength = 40
 
 // Common cert name. Cannot use common name as filename since envoy doesn't know if the app is dedicated or not
 const CertName = "envoyTlsCerts"
-const EnvoyImageDigest = "sha256:9bc06553ad6add6bfef1d8a1b04f09721415975e2507da0a2d5b914c066474df"
+const EnvoyImageDigest = "sha256:2b07bb8dd35c2a4bb273652b62e85b0bd27d12da94fa11061a9c365d4352e7f9"
 
 // PlatformApps is the set of all special "platform" developers.   Key
 // is DeveloperName:AppName.  Currently only Samsung's Enabling layer is included.

docker_envoy/Dockerfile

@@ -1,3 +1,3 @@
-FROM envoyproxy/envoy:v1.15-latest
+FROM envoyproxy/envoy:v1.18-latest
 RUN apt-get --assume-yes update && apt-get --assume-yes install curl
 ENV ENVOY_UID=0 ENVOY_GID=0