reload cni network config if has fs change events

With go RWMutex design, no goroutine should expect to be able to
acquire a read lock until the read lock has been released, if one
goroutine call lock.

The original design is to reload cni network config on every single
Status CRI gRPC call. If one RunPodSandbox request holds read lock
to allocate IP for too long, all other RunPodSandbox/StopPodSandbox
requests will wait for the RunPodSandbox request to release read lock.
And the Status CRI call will fail and kubelet becomes NOTReady.

Reload cni network config at every single Status CRI call is not
necessary and also brings NOTReady situation. To lower the possibility
of NOTReady, CRI will reload cni network config if there is any valid fs
change events from the cni network config dir.

Signed-off-by: Wei Fu <fuweid89@gmail.com>

pkg/server/cni_conf_syncer.go

@@ -0,0 +1,121 @@
+/*
+Copyright The Containerd Authors.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+*/
+
+package server
+
+import (
+	"os"
+	"sync"
+
+	cni "github.com/containerd/go-cni"
+	"github.com/fsnotify/fsnotify"
+	"github.com/pkg/errors"
+	"github.com/sirupsen/logrus"
+)
+
+// cniNetConfSyncer is used to reload cni network conf triggered by fs change
+// events.
+type cniNetConfSyncer struct {
+	// only used for lastSyncStatus
+	sync.RWMutex
+	lastSyncStatus error
+
+	watcher   *fsnotify.Watcher
+	confDir   string
+	netPlugin cni.CNI
+	loadOpts  []cni.CNIOpt
+}
+
+// newCNINetConfSyncer creates cni network conf syncer.
+func newCNINetConfSyncer(confDir string, netPlugin cni.CNI, loadOpts []cni.CNIOpt) (*cniNetConfSyncer, error) {
+	watcher, err := fsnotify.NewWatcher()
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to create fsnotify watcher")
+	}
+
+	if err := os.MkdirAll(confDir, 0700); err != nil {
+		return nil, errors.Wrapf(err, "failed to create cni conf dir=%v for watch", confDir)
+	}
+
+	if err := watcher.Add(confDir); err != nil {
+		return nil, errors.Wrapf(err, "failed to watch cni conf dir %s", confDir)
+	}
+
+	syncer := &cniNetConfSyncer{
+		watcher:   watcher,
+		confDir:   confDir,
+		netPlugin: netPlugin,
+		loadOpts:  loadOpts,
+	}
+
+	if err := syncer.netPlugin.Load(syncer.loadOpts...); err != nil {
+		logrus.WithError(err).Error("failed to load cni during init, please check CRI plugin status before setting up network for pods")
+		syncer.updateLastStatus(err)
+	}
+	return syncer, nil
+}
+
+// syncLoop monitors any fs change events from cni conf dir and tries to reload
+// cni configuration.
+func (syncer *cniNetConfSyncer) syncLoop() error {
+	for {
+		select {
+		case event := <-syncer.watcher.Events:
+			// Only reload config when receiving write/rename/remove
+			// events
+			//
+			// TODO(fuweid): Might only reload target cni config
+			// files to prevent no-ops.
+			if event.Op&(fsnotify.Chmod|fsnotify.Create) > 0 {
+				logrus.Debugf("ignore event from cni conf dir: %s", event)
+				continue
+			}
+			logrus.Debugf("receiving change event from cni conf dir: %s", event)
+
+			lerr := syncer.netPlugin.Load(syncer.loadOpts...)
+			if lerr != nil {
+				logrus.WithError(lerr).
+					Errorf("failed to reload cni configuration after receiving fs change event(%s)", event)
+			}
+			syncer.updateLastStatus(lerr)
+
+		case err := <-syncer.watcher.Errors:
+			if err != nil {
+				logrus.WithError(err).Error("failed to continue sync cni conf change")
+				return err
+			}
+		}
+	}
+}
+
+// lastStatus retrieves last sync status.
+func (syncer *cniNetConfSyncer) lastStatus() error {
+	syncer.RLock()
+	defer syncer.RUnlock()
+	return syncer.lastSyncStatus
+}
+
+// updateLastStatus will be called after every single cni load.
+func (syncer *cniNetConfSyncer) updateLastStatus(err error) {
+	syncer.Lock()
+	defer syncer.Unlock()
+	syncer.lastSyncStatus = err
+}
+
+// stop stops watcher in the syncLoop.
+func (syncer *cniNetConfSyncer) stop() error {
+	return syncer.watcher.Close()
+}

pkg/server/service.go

@@ -91,6 +91,9 @@ type criService struct {
 	// initialized indicates whether the server is initialized. All GRPC services
 	// should return error before the server is initialized.
 	initialized atomic.Bool
+	// cniNetConfMonitor is used to reload cni network conf if there is
+	// any valid fs change events from cni network conf dir.
+	cniNetConfMonitor *cniNetConfSyncer
 }
 
 // NewCRIService returns a new instance of CRIService
@@ -128,6 +131,11 @@ func NewCRIService(config criconfig.Config, client *containerd.Client) (CRIServi
 
 	c.eventMonitor = newEventMonitor(c)
 
+	c.cniNetConfMonitor, err = newCNINetConfSyncer(c.config.NetworkPluginConfDir, c.netPlugin, c.cniLoadOptions())
+	if err != nil {
+		return nil, errors.Wrap(err, "failed to create cni conf monitor")
+	}
+
 	return c, nil
 }
 
@@ -169,6 +177,14 @@ func (c *criService) Run() error {
 	)
 	snapshotsSyncer.start()
 
+	// Start CNI network conf syncer
+	logrus.Info("Start cni network conf syncer")
+	cniNetConfMonitorErrCh := make(chan error, 1)
+	go func() {
+		defer close(cniNetConfMonitorErrCh)
+		cniNetConfMonitorErrCh <- c.cniNetConfMonitor.syncLoop()
+	}()
+
 	// Start streaming server.
 	logrus.Info("Start streaming server")
 	streamServerErrCh := make(chan error)
@@ -183,11 +199,12 @@ func (c *criService) Run() error {
 	// Set the server as initialized. GRPC services could start serving traffic.
 	c.initialized.Set()
 
-	var eventMonitorErr, streamServerErr error
+	var eventMonitorErr, streamServerErr, cniNetConfMonitorErr error
 	// Stop the whole CRI service if any of the critical service exits.
 	select {
 	case eventMonitorErr = <-eventMonitorErrCh:
 	case streamServerErr = <-streamServerErrCh:
+	case cniNetConfMonitorErr = <-cniNetConfMonitorErrCh:
 	}
 	if err := c.Close(); err != nil {
 		return errors.Wrap(err, "failed to stop cri service")
@@ -222,13 +239,19 @@ func (c *criService) Run() error {
 	if streamServerErr != nil {
 		return errors.Wrap(streamServerErr, "stream server error")
 	}
+	if cniNetConfMonitorErr != nil {
+		return errors.Wrap(cniNetConfMonitorErr, "cni network conf monitor error")
+	}
 	return nil
 }
 
 // Close stops the CRI service.
 // TODO(random-liu): Make close synchronous.
 func (c *criService) Close() error {
 	logrus.Info("Stop CRI service")
+	if err := c.cniNetConfMonitor.stop(); err != nil {
+		logrus.Errorf("failed to stop cni network conf monitor: %v", err)
+	}
 	c.eventMonitor.stop()
 	if err := c.streamServer.Stop(); err != nil {
 		return errors.Wrap(err, "failed to stop stream server")

pkg/server/service_unix.go

@@ -60,11 +60,6 @@ func (c *criService) initPlatform() error {
 		return errors.Wrap(err, "failed to initialize cni")
 	}
 
-	// Try to load the config if it exists. Just log the error if load fails
-	// This is not disruptive for containerd to panic
-	if err := c.netPlugin.Load(c.cniLoadOptions()...); err != nil {
-		logrus.WithError(err).Error("Failed to load cni during init, please check CRI plugin status before setting up network for pods")
-	}
 	return nil
 }
 

pkg/server/service_windows.go

@@ -21,7 +21,6 @@ package server
 import (
 	cni "github.com/containerd/go-cni"
 	"github.com/pkg/errors"
-	"github.com/sirupsen/logrus"
 )
 
 // windowsNetworkAttachCount is the minimum number of networks the PodSandbox
@@ -44,11 +43,6 @@ func (c *criService) initPlatform() error {
 		return errors.Wrap(err, "failed to initialize cni")
 	}
 
-	// Try to load the config if it exists. Just log the error if load fails
-	// This is not disruptive for containerd to panic
-	if err := c.netPlugin.Load(c.cniLoadOptions()...); err != nil {
-		logrus.WithError(err).Error("Failed to load cni during init, please check CRI plugin status before setting up network for pods")
-	}
 	return nil
 }
 

pkg/server/status.go

@@ -41,10 +41,6 @@ func (c *criService) Status(ctx context.Context, r *runtime.StatusRequest) (*run
 		Type:   runtime.NetworkReady,
 		Status: true,
 	}
-	// Load the latest cni configuration to be in sync with the latest network configuration
-	if err := c.netPlugin.Load(c.cniLoadOptions()...); err != nil {
-		log.G(ctx).WithError(err).Errorf("Failed to load cni configuration")
-	}
 	// Check the status of the cni initialization
 	if err := c.netPlugin.Status(); err != nil {
 		networkCondition.Status = false
@@ -76,6 +72,12 @@ func (c *criService) Status(ctx context.Context, r *runtime.StatusRequest) (*run
 			log.G(ctx).WithError(err).Errorf("Failed to marshal CNI config %v", err)
 		}
 		resp.Info["cniconfig"] = string(cniConfig)
+
+		lastCNILoadStatus := "OK"
+		if lerr := c.cniNetConfMonitor.lastStatus(); lerr != nil {
+			lastCNILoadStatus = lerr.Error()
+		}
+		resp.Info["lastCNILoadStatus"] = lastCNILoadStatus
 	}
 	return resp, nil
 }

vendor.conf

@@ -89,6 +89,7 @@ sigs.k8s.io/yaml                                    9fc95527decd95bb9d28cc2eab08
 github.com/containerd/go-cni                        0d360c50b10b350b6bb23863fd4dfb1c232b01c9
 github.com/containernetworking/cni                  4cfb7b568922a3c79a23e438dc52fe537fc9687e # v0.7.1
 github.com/containernetworking/plugins              9f96827c7cabb03f21d86326000c00f61e181f6a # v0.7.6
+github.com/fsnotify/fsnotify                        c2828203cd70a50dcccfb2761f8b1f8ceef9a8e9 # v1.4.7
 
 # image decrypt depedencies
 github.com/containerd/imgcrypt                      9e761ccd6069fb707ec9493435f31475b5524b38 # v1.0.1