Merge pull request #19 from alauda/feat/pod-performance

feat: separate static and dynamic address pod and add parallelism to …

pkg/controller/config.go

@@ -35,6 +35,8 @@ type Configuration struct {
 
 	PodName      string
 	PodNamespace string
+
+	WorkerNum int
 }
 
 // ParseFlags parses cmd args then init kubeclient and conf
@@ -58,6 +60,8 @@ func ParseFlags() (*Configuration, error) {
 
 		argClusterTcpLoadBalancer = pflag.String("cluster-tcp-loadbalancer", "cluster-tcp-loadbalancer", "The name for cluster tcp loadbalancer")
 		argClusterUdpLoadBalancer = pflag.String("cluster-udp-loadbalancer", "cluster-udp-loadbalancer", "The name for cluster udp loadbalancer")
+
+		argWorkerNum = pflag.Int("worker-num", 3, "The parallelism of each worker. Default: 3")
 	)
 
 	flag.Set("alsologtostderr", "true")
@@ -93,6 +97,7 @@ func ParseFlags() (*Configuration, error) {
 		NodeSwitchGateway:      *argNodeSwitchGateway,
 		ClusterTcpLoadBalancer: *argClusterTcpLoadBalancer,
 		ClusterUdpLoadBalancer: *argClusterUdpLoadBalancer,
+		WorkerNum:              *argWorkerNum,
 		PodName:                os.Getenv("POD_NAME"),
 		PodNamespace:           os.Getenv("KUBE_NAMESPACE"),
 	}

pkg/controller/controller.go

@@ -38,9 +38,10 @@ type Controller struct {
 	// means we can ensure we only process a fixed amount of resources at a
 	// time, and makes it easy to ensure we are never processing the same item
 	// simultaneously in two different workers.
-	addPodQueue    workqueue.RateLimitingInterface
-	deletePodQueue workqueue.RateLimitingInterface
-	updatePodQueue workqueue.RateLimitingInterface
+	addPodQueue       workqueue.RateLimitingInterface
+	addIpPoolPodQueue workqueue.RateLimitingInterface
+	deletePodQueue    workqueue.RateLimitingInterface
+	updatePodQueue    workqueue.RateLimitingInterface
 
 	namespacesLister     v1.NamespaceLister
 	namespacesSynced     cache.InformerSynced
@@ -95,11 +96,12 @@ func NewController(config *Configuration) *Controller {
 		ovnClient:     ovs.NewClient(config.OvnNbHost, config.OvnNbPort, "", 0, config.ClusterRouter, config.ClusterTcpLoadBalancer, config.ClusterUdpLoadBalancer, config.NodeSwitch, config.NodeSwitchCIDR),
 		kubeclientset: config.KubeClient,
 
-		podsLister:     podInformer.Lister(),
-		podsSynced:     podInformer.Informer().HasSynced,
-		addPodQueue:    workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "AddPod"),
-		deletePodQueue: workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "DeletePod"),
-		updatePodQueue: workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "UpdatePod"),
+		podsLister:        podInformer.Lister(),
+		podsSynced:        podInformer.Informer().HasSynced,
+		addPodQueue:       workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "AddPod"),
+		addIpPoolPodQueue: workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "AddIpPoolPod"),
+		deletePodQueue:    workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "DeletePod"),
+		updatePodQueue:    workqueue.NewNamedRateLimitingQueue(workqueue.DefaultControllerRateLimiter(), "UpdatePod"),
 
 		namespacesLister:     namespaceInformer.Lister(),
 		namespacesSynced:     namespaceInformer.Informer().HasSynced,
@@ -162,10 +164,24 @@ func NewController(config *Configuration) *Controller {
 // workers to finish processing their current work items.
 func (c *Controller) Run(stopCh <-chan struct{}) error {
 	defer utilruntime.HandleCrash()
+
 	defer c.addPodQueue.ShutDown()
+	defer c.addIpPoolPodQueue.ShutDown()
 	defer c.deletePodQueue.ShutDown()
 	defer c.updatePodQueue.ShutDown()
 
+	defer c.addNamespaceQueue.ShutDown()
+	defer c.updateNamespaceQueue.ShutDown()
+	defer c.deleteNamespaceQueue.ShutDown()
+
+	defer c.addNodeQueue.ShutDown()
+	defer c.deleteNodeQueue.ShutDown()
+
+	defer c.addServiceQueue.ShutDown()
+	defer c.updateServiceQueue.ShutDown()
+
+	defer c.updateEndpointQueue.ShutDown()
+
 	// Start the informer factories to begin populating the informer caches
 	klog.Info("Starting OVN controller")
 
@@ -195,21 +211,25 @@ func (c *Controller) Run(stopCh <-chan struct{}) error {
 	klog.Info("Starting workers")
 
 	// Launch workers to process resources
-	go wait.Until(c.runAddPodWorker, time.Second, stopCh)
-	go wait.Until(c.runDeletePodWorker, time.Second, stopCh)
-	go wait.Until(c.runUpdatePodWorker, time.Second, stopCh)
+	go wait.Until(c.runAddIpPoolPodWorker, time.Second, stopCh)
+
+	for i := 0; i < c.config.WorkerNum; i++ {
+		go wait.Until(c.runAddPodWorker, time.Second, stopCh)
+		go wait.Until(c.runDeletePodWorker, time.Second, stopCh)
+		go wait.Until(c.runUpdatePodWorker, time.Second, stopCh)
 
-	go wait.Until(c.runAddNamespaceWorker, time.Second, stopCh)
-	go wait.Until(c.runDeleteNamespaceWorker, time.Second, stopCh)
-	go wait.Until(c.runUpdateNamespaceWorker, time.Second, stopCh)
+		go wait.Until(c.runAddNamespaceWorker, time.Second, stopCh)
+		go wait.Until(c.runDeleteNamespaceWorker, time.Second, stopCh)
+		go wait.Until(c.runUpdateNamespaceWorker, time.Second, stopCh)
 
-	go wait.Until(c.runAddNodeWorker, time.Second, stopCh)
-	go wait.Until(c.runDeleteNodeWorker, time.Second, stopCh)
+		go wait.Until(c.runAddNodeWorker, time.Second, stopCh)
+		go wait.Until(c.runDeleteNodeWorker, time.Second, stopCh)
 
-	go wait.Until(c.runUpdateServiceWorker, time.Second, stopCh)
-	go wait.Until(c.runAddServiceWorker, time.Second, stopCh)
+		go wait.Until(c.runUpdateServiceWorker, time.Second, stopCh)
+		go wait.Until(c.runAddServiceWorker, time.Second, stopCh)
 
-	go wait.Until(c.runUpdateEndpointWorker, time.Second, stopCh)
+		go wait.Until(c.runUpdateEndpointWorker, time.Second, stopCh)
+	}
 
 	klog.Info("Started workers")
 	<-stopCh

pkg/controller/pod.go

@@ -30,6 +30,14 @@ func (c *Controller) enqueueAddPod(obj interface{}) {
 		utilruntime.HandleError(err)
 		return
 	}
+
+	p := obj.(*v1.Pod)
+	if p.Annotations[util.IpPoolAnnotation] != "" && p.Annotations[util.IpAddressAnnotation] == "" {
+		klog.V(3).Infof("enqueue add ip pool address pod %s", key)
+		c.addIpPoolPodQueue.AddRateLimited(key)
+		return
+	}
+
 	klog.V(3).Infof("enqueue add pod %s", key)
 	c.addPodQueue.AddRateLimited(key)
 }
@@ -78,6 +86,11 @@ func (c *Controller) runAddPodWorker() {
 	}
 }
 
+func (c *Controller) runAddIpPoolPodWorker() {
+	for c.processNextAddIpPoolPodWorkItem() {
+	}
+}
+
 // runWorker is a long-running function that will continually call the
 // processNextWorkItem function in order to read and process a message on the
 // workqueue.
@@ -145,6 +158,58 @@ func (c *Controller) processNextAddPodWorkItem() bool {
 	return true
 }
 
+func (c *Controller) processNextAddIpPoolPodWorkItem() bool {
+	obj, shutdown := c.addIpPoolPodQueue.Get()
+
+	if shutdown {
+		return false
+	}
+
+	// We wrap this block in a func so we can defer c.workqueue.Done.
+	err := func(obj interface{}) error {
+		// We call Done here so the workqueue knows we have finished
+		// processing this item. We also must remember to call Forget if we
+		// do not want this work item being re-queued. For example, we do
+		// not call Forget if a transient error occurs, instead the item is
+		// put back on the workqueue and attempted again after a back-off
+		// period.
+		defer c.addIpPoolPodQueue.Done(obj)
+		var key string
+		var ok bool
+		// We expect strings to come off the workqueue. These are of the
+		// form namespace/name. We do this as the delayed nature of the
+		// workqueue means the items in the informer cache may actually be
+		// more up to date that when the item was initially put onto the
+		// workqueue.
+		if key, ok = obj.(string); !ok {
+			// As the item in the workqueue is actually invalid, we call
+			// Forget here else we'd go into a loop of attempting to
+			// process a work item that is invalid.
+			c.addIpPoolPodQueue.Forget(obj)
+			utilruntime.HandleError(fmt.Errorf("expected string in workqueue but got %#v", obj))
+			return nil
+		}
+		// Run the syncHandler, passing it the namespace/name string of the
+		// Foo resource to be synced.
+		if err := c.handleAddIpPoolPod(key); err != nil {
+			// Put the item back on the workqueue to handle any transient errors.
+			c.addIpPoolPodQueue.AddRateLimited(key)
+			return fmt.Errorf("error syncing '%s': %s, requeuing", key, err.Error())
+		}
+		// Finally, if no error occurs we Forget this item so it does not
+		// get queued again until another change happens.
+		c.addIpPoolPodQueue.Forget(obj)
+		return nil
+	}(obj)
+
+	if err != nil {
+		utilruntime.HandleError(err)
+		return true
+	}
+
+	return true
+}
+
 // processNextWorkItem will read a single work item off the workqueue and
 // attempt to process it, by calling the syncHandler.
 func (c *Controller) processNextDeletePodWorkItem() bool {
@@ -283,6 +348,76 @@ func (c *Controller) handleAddPod(key string) error {
 		ls = c.config.DefaultLogicalSwitch
 	}
 
+	// pod address info may already exist in ovn
+	ip := pod.Annotations[util.IpAddressAnnotation]
+	mac := pod.Annotations[util.MacAddressAnnotation]
+
+	nic, err := c.ovnClient.CreatePort(ls, ovs.PodNameToPortName(name, namespace), ip, mac)
+	if err != nil {
+		return err
+	}
+
+	op := "replace"
+	if len(pod.Annotations) == 0 {
+		op = "add"
+	}
+	if pod.Annotations == nil {
+		pod.Annotations = map[string]string{}
+	}
+	pod.Annotations[util.IpAddressAnnotation] = nic.IpAddress
+	pod.Annotations[util.MacAddressAnnotation] = nic.MacAddress
+	pod.Annotations[util.CidrAnnotation] = nic.CIDR
+	pod.Annotations[util.GatewayAnnotation] = nic.Gateway
+	pod.Annotations[util.LogicalSwitchAnnotation] = ls
+
+	patchPayloadTemplate :=
+		`[{
+        "op": "%s",
+        "path": "/metadata/annotations",
+        "value": %s
+    }]`
+
+	raw, _ := json.Marshal(pod.Annotations)
+	patchPayload := fmt.Sprintf(patchPayloadTemplate, op, raw)
+	_, err = c.kubeclientset.CoreV1().Pods(namespace).Patch(name, types.JSONPatchType, []byte(patchPayload))
+	if err != nil {
+		klog.Errorf("patch pod %s/%s failed %v", name, namespace, err)
+	}
+	return err
+}
+
+func (c *Controller) handleAddIpPoolPod(key string) error {
+	// Convert the namespace/name string into a distinct namespace and name
+	namespace, name, err := cache.SplitMetaNamespaceKey(key)
+	if err != nil {
+		utilruntime.HandleError(fmt.Errorf("invalid resource key: %s", key))
+		return nil
+	}
+	pod, err := c.podsLister.Pods(namespace).Get(name)
+	if err != nil {
+		// The Pod resource may no longer exist, in which case we stop
+		// processing.
+		if k8serrors.IsNotFound(err) {
+			return nil
+		}
+		return err
+	}
+	klog.Infof("add ip pool address pod %s/%s", namespace, name)
+	if pod.Spec.HostNetwork {
+		klog.Infof("pod %s/%s in host network mode no need for ovn process", namespace, name)
+		return nil
+	}
+
+	ns, err := c.namespacesLister.Get(namespace)
+	if err != nil {
+		klog.Errorf("get namespace %s failed %v", namespace, err)
+		return err
+	}
+	ls := ns.Annotations[util.LogicalSwitchAnnotation]
+	if ls == "" {
+		ls = c.config.DefaultLogicalSwitch
+	}
+
 	ipPoolAnnotation := pod.Annotations[util.IpPoolAnnotation]
 
 	if ipPoolAnnotation != "" && pod.Annotations[util.IpAddressAnnotation] == "" {

pkg/daemon/controller.go

@@ -307,7 +307,9 @@ func (c *Controller) handlePod(key string) error {
 // Run starts controller
 func (c *Controller) Run(stopCh <-chan struct{}) error {
 	defer utilruntime.HandleCrash()
+
 	defer c.namespaceQueue.ShutDown()
+	defer c.podQueue.ShutDown()
 
 	klog.Info("start watching namespace changes")
 	if ok := cache.WaitForCacheSync(stopCh, c.namespacesSynced, c.podsSynced); !ok {