github/kubernetes
4
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Update to latest cadvisor - cleanup mesos/rkt

Browse Source 
Change-Id: Ib5ae0cb13b93f8c87bb74e3ba33040df5f3d6a6f
This commit is contained in:
Davanum Srinivas
2019-04-09 13:11:33 -04:00
parent ca55432599
commit 70d562a6ac
143 changed files with 554 additions and 190406 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
vendor/github.com/google/cadvisor/manager/BUILD generated vendored
View File

@@ -15,28 +15,20 @@ go_library(
"//vendor/github.com/google/cadvisor/cache/memory:go_default_library",
"//vendor/github.com/google/cadvisor/collector:go_default_library",
"//vendor/github.com/google/cadvisor/container:go_default_library",
"//vendor/github.com/google/cadvisor/container/containerd:go_default_library",
"//vendor/github.com/google/cadvisor/container/crio:go_default_library",
"//vendor/github.com/google/cadvisor/container/docker:go_default_library",
"//vendor/github.com/google/cadvisor/container/mesos:go_default_library",
"//vendor/github.com/google/cadvisor/container/raw:go_default_library",
"//vendor/github.com/google/cadvisor/container/rkt:go_default_library",
"//vendor/github.com/google/cadvisor/container/systemd:go_default_library",
"//vendor/github.com/google/cadvisor/events:go_default_library",
"//vendor/github.com/google/cadvisor/fs:go_default_library",
"//vendor/github.com/google/cadvisor/info/v1:go_default_library",
"//vendor/github.com/google/cadvisor/info/v2:go_default_library",
"//vendor/github.com/google/cadvisor/machine:go_default_library",
"//vendor/github.com/google/cadvisor/manager/watcher:go_default_library",
"//vendor/github.com/google/cadvisor/manager/watcher/raw:go_default_library",
"//vendor/github.com/google/cadvisor/manager/watcher/rkt:go_default_library",
"//vendor/github.com/google/cadvisor/summary:go_default_library",
"//vendor/github.com/google/cadvisor/utils/cpuload:go_default_library",
"//vendor/github.com/google/cadvisor/utils/oomparser:go_default_library",
"//vendor/github.com/google/cadvisor/utils/sysfs:go_default_library",
"//vendor/github.com/google/cadvisor/version:go_default_library",
"//vendor/github.com/google/cadvisor/watcher:go_default_library",
"//vendor/github.com/opencontainers/runc/libcontainer/cgroups:go_default_library",
"//vendor/golang.org/x/net/context:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
"//vendor/k8s.io/utils/clock:go_default_library",
],
@@ -51,10 +43,7 @@ filegroup(
filegroup(
name = "all-srcs",
srcs = [
":package-srcs",
"//vendor/github.com/google/cadvisor/manager/watcher:all-srcs",
],
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

109
vendor/github.com/google/cadvisor/manager/manager.go generated vendored
View File

@@ -30,27 +30,19 @@ import (
"github.com/google/cadvisor/cache/memory"
"github.com/google/cadvisor/collector"
"github.com/google/cadvisor/container"
"github.com/google/cadvisor/container/containerd"
"github.com/google/cadvisor/container/crio"
"github.com/google/cadvisor/container/docker"
"github.com/google/cadvisor/container/mesos"
"github.com/google/cadvisor/container/raw"
"github.com/google/cadvisor/container/rkt"
"github.com/google/cadvisor/container/systemd"
"github.com/google/cadvisor/events"
"github.com/google/cadvisor/fs"
info "github.com/google/cadvisor/info/v1"
"github.com/google/cadvisor/info/v2"
"github.com/google/cadvisor/machine"
"github.com/google/cadvisor/manager/watcher"
rawwatcher "github.com/google/cadvisor/manager/watcher/raw"
rktwatcher "github.com/google/cadvisor/manager/watcher/rkt"
"github.com/google/cadvisor/utils/oomparser"
"github.com/google/cadvisor/utils/sysfs"
"github.com/google/cadvisor/version"
"github.com/google/cadvisor/watcher"
"github.com/opencontainers/runc/libcontainer/cgroups"
"golang.org/x/net/context"
"k8s.io/klog"
"k8s.io/utils/clock"
)
@@ -62,8 +54,6 @@ var eventStorageAgeLimit = flag.String("event_storage_age_limit", "default=24h",
var eventStorageEventLimit = flag.String("event_storage_event_limit", "default=100000", "Max number of events to store (per type). Value is a comma separated list of key values, where the keys are event types (e.g.: creation, oom) or \"default\" and the value is an integer. Default is applied to all non-specified event types")
var applicationMetricsCountLimit = flag.Int("application_metrics_count_limit", 100, "Max number of application metrics to store (per container)")
const dockerClientTimeout = 10 * time.Second
// The Manager interface defines operations for starting a manager and getting
// container and machine information.
type Manager interface {
@@ -155,40 +145,12 @@ func New(memoryCache *memory.InMemoryCache, sysfs sysfs.SysFs, maxHousekeepingIn
}
klog.V(2).Infof("cAdvisor running in container: %q", selfContainer)
var (
dockerStatus info.DockerStatus
rktPath string
)
docker.SetTimeout(dockerClientTimeout)
// Try to connect to docker indefinitely on startup.
dockerStatus = retryDockerStatus()
context := fs.Context{}
if tmpRktPath, err := rkt.RktPath(); err != nil {
klog.V(5).Infof("Rkt not connected: %v", err)
} else {
rktPath = tmpRktPath
}
crioClient, err := crio.Client()
if err != nil {
if err := container.InitializeFSContext(&context); err != nil {
return nil, err
}
crioInfo, err := crioClient.Info()
if err != nil {
klog.V(5).Infof("CRI-O not connected: %v", err)
}
context := fs.Context{
Docker: fs.DockerContext{
Root: docker.RootDir(),
Driver: dockerStatus.Driver,
DriverStatus: dockerStatus.DriverStatus,
},
RktPath: rktPath,
Crio: fs.CrioContext{
Root: crioInfo.StorageRoot,
},
}
fsInfo, err := fs.NewFsInfo(context)
if err != nil {
return nil, err
@@ -240,31 +202,6 @@ func New(memoryCache *memory.InMemoryCache, sysfs sysfs.SysFs, maxHousekeepingIn
return newManager, nil
}
func retryDockerStatus() info.DockerStatus {
startupTimeout := dockerClientTimeout
maxTimeout := 4 * startupTimeout
for {
ctx, _ := context.WithTimeout(context.Background(), startupTimeout)
dockerStatus, err := docker.StatusWithContext(ctx)
if err == nil {
return dockerStatus
}
switch err {
case context.DeadlineExceeded:
klog.Warningf("Timeout trying to communicate with docker during initialization, will retry")
default:
klog.V(5).Infof("Docker not connected: %v", err)
return info.DockerStatus{}
}
startupTimeout = 2 * startupTimeout
if startupTimeout > maxTimeout {
startupTimeout = maxTimeout
}
}
}
// A namespaced container name.
type namespacedContainerName struct {
// The namespace of the container. Can be empty for the root namespace.
@@ -300,48 +237,14 @@ type manager struct {
// Start the container manager.
func (self *manager) Start() error {
err := docker.Register(self, self.fsInfo, self.includedMetrics)
if err != nil {
klog.V(5).Infof("Registration of the Docker container factory failed: %v.", err)
}
self.containerWatchers = container.InitializePlugins(self, self.fsInfo, self.includedMetrics)
err = rkt.Register(self, self.fsInfo, self.includedMetrics)
if err != nil {
klog.V(5).Infof("Registration of the rkt container factory failed: %v", err)
} else {
watcher, err := rktwatcher.NewRktContainerWatcher()
if err != nil {
return err
}
self.containerWatchers = append(self.containerWatchers, watcher)
}
err = containerd.Register(self, self.fsInfo, self.includedMetrics)
if err != nil {
klog.V(5).Infof("Registration of the containerd container factory failed: %v", err)
}
err = crio.Register(self, self.fsInfo, self.includedMetrics)
if err != nil {
klog.V(5).Infof("Registration of the crio container factory failed: %v", err)
}
err = mesos.Register(self, self.fsInfo, self.includedMetrics)
if err != nil {
klog.V(5).Infof("Registration of the mesos container factory failed: %v", err)
}
err = systemd.Register(self, self.fsInfo, self.includedMetrics)
if err != nil {
klog.V(5).Infof("Registration of the systemd container factory failed: %v", err)
}
err = raw.Register(self, self.fsInfo, self.includedMetrics, self.rawContainerCgroupPathPrefixWhiteList)
err := raw.Register(self, self.fsInfo, self.includedMetrics, self.rawContainerCgroupPathPrefixWhiteList)
if err != nil {
klog.Errorf("Registration of the raw container factory failed: %v", err)
}
rawWatcher, err := rawwatcher.NewRawContainerWatcher()
rawWatcher, err := raw.NewRawContainerWatcher()
if err != nil {
return err
}

27
vendor/github.com/google/cadvisor/manager/watcher/BUILD generated vendored
View File

@@ -1,27 +0,0 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library")
go_library(
name = "go_default_library",
srcs = ["watcher.go"],
importmap = "k8s.io/kubernetes/vendor/github.com/google/cadvisor/manager/watcher",
importpath = "github.com/google/cadvisor/manager/watcher",
visibility = ["//visibility:public"],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [
":package-srcs",
"//vendor/github.com/google/cadvisor/manager/watcher/raw:all-srcs",
"//vendor/github.com/google/cadvisor/manager/watcher/rkt:all-srcs",
],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

30
vendor/github.com/google/cadvisor/manager/watcher/raw/BUILD generated vendored
View File

@@ -1,30 +0,0 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library")
go_library(
name = "go_default_library",
srcs = ["raw.go"],
importmap = "k8s.io/kubernetes/vendor/github.com/google/cadvisor/manager/watcher/raw",
importpath = "github.com/google/cadvisor/manager/watcher/raw",
visibility = ["//visibility:public"],
deps = [
"//vendor/github.com/google/cadvisor/container/common:go_default_library",
"//vendor/github.com/google/cadvisor/container/libcontainer:go_default_library",
"//vendor/github.com/google/cadvisor/manager/watcher:go_default_library",
"//vendor/github.com/sigma/go-inotify:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

237
vendor/github.com/google/cadvisor/manager/watcher/raw/raw.go generated vendored
View File

@@ -1,237 +0,0 @@
// Copyright 2014 Google Inc. All Rights Reserved.
//
// 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 container defines types for sub-container events and also
// defines an interface for container operation handlers.
package raw
import (
"fmt"
"io/ioutil"
"os"
"path"
"strings"
"github.com/google/cadvisor/container/common"
"github.com/google/cadvisor/container/libcontainer"
"github.com/google/cadvisor/manager/watcher"
inotify "github.com/sigma/go-inotify"
"k8s.io/klog"
)
type rawContainerWatcher struct {
// Absolute path to the root of the cgroup hierarchies
cgroupPaths map[string]string
cgroupSubsystems *libcontainer.CgroupSubsystems
// Inotify event watcher.
watcher *common.InotifyWatcher
// Signal for watcher thread to stop.
stopWatcher chan error
}
func NewRawContainerWatcher() (watcher.ContainerWatcher, error) {
cgroupSubsystems, err := libcontainer.GetAllCgroupSubsystems()
if err != nil {
return nil, fmt.Errorf("failed to get cgroup subsystems: %v", err)
}
if len(cgroupSubsystems.Mounts) == 0 {
return nil, fmt.Errorf("failed to find supported cgroup mounts for the raw factory")
}
watcher, err := common.NewInotifyWatcher()
if err != nil {
return nil, err
}
rawWatcher := &rawContainerWatcher{
cgroupPaths: common.MakeCgroupPaths(cgroupSubsystems.MountPoints, "/"),
cgroupSubsystems: &cgroupSubsystems,
watcher: watcher,
stopWatcher: make(chan error),
}
return rawWatcher, nil
}
func (self *rawContainerWatcher) Start(events chan watcher.ContainerEvent) error {
// Watch this container (all its cgroups) and all subdirectories.
for _, cgroupPath := range self.cgroupPaths {
_, err := self.watchDirectory(events, cgroupPath, "/")
if err != nil {
return err
}
}
// Process the events received from the kernel.
go func() {
for {
select {
case event := <-self.watcher.Event():
err := self.processEvent(event, events)
if err != nil {
klog.Warningf("Error while processing event (%+v): %v", event, err)
}
case err := <-self.watcher.Error():
klog.Warningf("Error while watching %q: %v", "/", err)
case <-self.stopWatcher:
err := self.watcher.Close()
if err == nil {
self.stopWatcher <- err
return
}
}
}
}()
return nil
}
func (self *rawContainerWatcher) Stop() error {
// Rendezvous with the watcher thread.
self.stopWatcher <- nil
return <-self.stopWatcher
}
// Watches the specified directory and all subdirectories. Returns whether the path was
// already being watched and an error (if any).
func (self *rawContainerWatcher) watchDirectory(events chan watcher.ContainerEvent, dir string, containerName string) (bool, error) {
// Don't watch .mount cgroups because they never have containers as sub-cgroups. A single container
// can have many .mount cgroups associated with it which can quickly exhaust the inotify watches on a node.
if strings.HasSuffix(containerName, ".mount") {
return false, nil
}
alreadyWatching, err := self.watcher.AddWatch(containerName, dir)
if err != nil {
return alreadyWatching, err
}
// Remove the watch if further operations failed.
cleanup := true
defer func() {
if cleanup {
_, err := self.watcher.RemoveWatch(containerName, dir)
if err != nil {
klog.Warningf("Failed to remove inotify watch for %q: %v", dir, err)
}
}
}()
// TODO(vmarmol): We should re-do this once we're done to ensure directories were not added in the meantime.
// Watch subdirectories as well.
entries, err := ioutil.ReadDir(dir)
if err != nil {
return alreadyWatching, err
}
for _, entry := range entries {
if entry.IsDir() {
entryPath := path.Join(dir, entry.Name())
subcontainerName := path.Join(containerName, entry.Name())
alreadyWatchingSubDir, err := self.watchDirectory(events, entryPath, subcontainerName)
if err != nil {
klog.Errorf("Failed to watch directory %q: %v", entryPath, err)
if os.IsNotExist(err) {
// The directory may have been removed before watching. Try to watch the other
// subdirectories. (https://github.com/kubernetes/kubernetes/issues/28997)
continue
}
return alreadyWatching, err
}
// since we already missed the creation event for this directory, publish an event here.
if !alreadyWatchingSubDir {
go func() {
events <- watcher.ContainerEvent{
EventType: watcher.ContainerAdd,
Name: subcontainerName,
WatchSource: watcher.Raw,
}
}()
}
}
}
cleanup = false
return alreadyWatching, nil
}
func (self *rawContainerWatcher) processEvent(event *inotify.Event, events chan watcher.ContainerEvent) error {
// Convert the inotify event type to a container create or delete.
var eventType watcher.ContainerEventType
switch {
case (event.Mask & inotify.IN_CREATE) > 0:
eventType = watcher.ContainerAdd
case (event.Mask & inotify.IN_DELETE) > 0:
eventType = watcher.ContainerDelete
case (event.Mask & inotify.IN_MOVED_FROM) > 0:
eventType = watcher.ContainerDelete
case (event.Mask & inotify.IN_MOVED_TO) > 0:
eventType = watcher.ContainerAdd
default:
// Ignore other events.
return nil
}
// Derive the container name from the path name.
var containerName string
for _, mount := range self.cgroupSubsystems.Mounts {
mountLocation := path.Clean(mount.Mountpoint) + "/"
if strings.HasPrefix(event.Name, mountLocation) {
containerName = event.Name[len(mountLocation)-1:]
break
}
}
if containerName == "" {
return fmt.Errorf("unable to detect container from watch event on directory %q", event.Name)
}
// Maintain the watch for the new or deleted container.
switch eventType {
case watcher.ContainerAdd:
// New container was created, watch it.
alreadyWatched, err := self.watchDirectory(events, event.Name, containerName)
if err != nil {
return err
}
// Only report container creation once.
if alreadyWatched {
return nil
}
case watcher.ContainerDelete:
// Container was deleted, stop watching for it.
lastWatched, err := self.watcher.RemoveWatch(containerName, event.Name)
if err != nil {
return err
}
// Only report container deletion once.
if !lastWatched {
return nil
}
default:
return fmt.Errorf("unknown event type %v", eventType)
}
// Deliver the event.
events <- watcher.ContainerEvent{
EventType: eventType,
Name: containerName,
WatchSource: watcher.Raw,
}
return nil
}

30
vendor/github.com/google/cadvisor/manager/watcher/rkt/BUILD generated vendored
View File

@@ -1,30 +0,0 @@
load("@io_bazel_rules_go//go:def.bzl", "go_library")
go_library(
name = "go_default_library",
srcs = ["rkt.go"],
importmap = "k8s.io/kubernetes/vendor/github.com/google/cadvisor/manager/watcher/rkt",
importpath = "github.com/google/cadvisor/manager/watcher/rkt",
visibility = ["//visibility:public"],
deps = [
"//vendor/github.com/coreos/rkt/api/v1alpha:go_default_library",
"//vendor/github.com/google/cadvisor/container/rkt:go_default_library",
"//vendor/github.com/google/cadvisor/manager/watcher:go_default_library",
"//vendor/golang.org/x/net/context:go_default_library",
"//vendor/k8s.io/klog:go_default_library",
],
)
filegroup(
name = "package-srcs",
srcs = glob(["**"]),
tags = ["automanaged"],
visibility = ["//visibility:private"],
)
filegroup(
name = "all-srcs",
srcs = [":package-srcs"],
tags = ["automanaged"],
visibility = ["//visibility:public"],
)

154
vendor/github.com/google/cadvisor/manager/watcher/rkt/rkt.go generated vendored
View File

@@ -1,154 +0,0 @@
// Copyright 2016 Google Inc. All Rights Reserved.
//
// 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 rkt implements the watcher interface for rkt
package rkt
import (
"path/filepath"
"time"
"github.com/google/cadvisor/container/rkt"
"github.com/google/cadvisor/manager/watcher"
rktapi "github.com/coreos/rkt/api/v1alpha"
"golang.org/x/net/context"
"k8s.io/klog"
)
type rktContainerWatcher struct {
// Signal for watcher thread to stop.
stopWatcher chan error
}
func NewRktContainerWatcher() (watcher.ContainerWatcher, error) {
watcher := &rktContainerWatcher{
stopWatcher: make(chan error),
}
return watcher, nil
}
func (self *rktContainerWatcher) Start(events chan watcher.ContainerEvent) error {
go self.detectRktContainers(events)
return nil
}
func (self *rktContainerWatcher) Stop() error {
// Rendezvous with the watcher thread.
self.stopWatcher <- nil
return nil
}
func (self *rktContainerWatcher) detectRktContainers(events chan watcher.ContainerEvent) {
klog.V(1).Infof("Starting detectRktContainers thread")
ticker := time.Tick(10 * time.Second)
curpods := make(map[string]*rktapi.Pod)
for {
select {
case <-ticker:
pods, err := listRunningPods()
if err != nil {
klog.Errorf("detectRktContainers: listRunningPods failed: %v", err)
continue
}
curpods = self.syncRunningPods(pods, events, curpods)
case <-self.stopWatcher:
klog.Infof("Exiting rktContainer Thread")
return
}
}
}
func (self *rktContainerWatcher) syncRunningPods(pods []*rktapi.Pod, events chan watcher.ContainerEvent, curpods map[string]*rktapi.Pod) map[string]*rktapi.Pod {
newpods := make(map[string]*rktapi.Pod)
for _, pod := range pods {
newpods[pod.Id] = pod
// if pods become mutable, have to handle this better
if _, ok := curpods[pod.Id]; !ok {
// should create all cgroups not including system.slice
// i.e. /system.slice/rkt-test.service and /system.slice/rkt-test.service/system.slice/pause.service
for _, cgroup := range podToCgroup(pod) {
self.sendUpdateEvent(cgroup, events)
}
}
}
for id, pod := range curpods {
if _, ok := newpods[id]; !ok {
for _, cgroup := range podToCgroup(pod) {
klog.V(2).Infof("cgroup to delete = %v", cgroup)
self.sendDestroyEvent(cgroup, events)
}
}
}
return newpods
}
func (self *rktContainerWatcher) sendUpdateEvent(cgroup string, events chan watcher.ContainerEvent) {
events <- watcher.ContainerEvent{
EventType: watcher.ContainerAdd,
Name: cgroup,
WatchSource: watcher.Rkt,
}
}
func (self *rktContainerWatcher) sendDestroyEvent(cgroup string, events chan watcher.ContainerEvent) {
events <- watcher.ContainerEvent{
EventType: watcher.ContainerDelete,
Name: cgroup,
WatchSource: watcher.Rkt,
}
}
func listRunningPods() ([]*rktapi.Pod, error) {
client, err := rkt.Client()
if err != nil {
return nil, err
}
resp, err := client.ListPods(context.Background(), &rktapi.ListPodsRequest{
// Specify the request: Fetch and print only running pods and their details.
Detail: true,
Filters: []*rktapi.PodFilter{
{
States: []rktapi.PodState{rktapi.PodState_POD_STATE_RUNNING},
},
},
})
if err != nil {
return nil, err
}
return resp.Pods, nil
}
func podToCgroup(pod *rktapi.Pod) []string {
cgroups := make([]string, 1+len(pod.Apps), 1+len(pod.Apps))
baseCgroup := pod.Cgroup
cgroups[0] = baseCgroup
for i, app := range pod.Apps {
cgroups[i+1] = filepath.Join(baseCgroup, "system.slice", app.Name+".service")
}
return cgroups
}

52
vendor/github.com/google/cadvisor/manager/watcher/watcher.go generated vendored
View File

@@ -1,52 +0,0 @@
// Copyright 2016 Google Inc. All Rights Reserved.
//
// 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 container defines types for sub-container events and also
// defines an interface for container operation handlers.
package watcher
// SubcontainerEventType indicates an addition or deletion event.
type ContainerEventType int
const (
ContainerAdd ContainerEventType = iota
ContainerDelete
)
type ContainerWatchSource int
const (
Raw ContainerWatchSource = iota
Rkt
)
// ContainerEvent represents a
type ContainerEvent struct {
// The type of event that occurred.
EventType ContainerEventType
// The full container name of the container where the event occurred.
Name string
// The watcher that detected this change event
WatchSource ContainerWatchSource
}
type ContainerWatcher interface {
// Registers a channel to listen for events affecting subcontainers (recursively).
Start(events chan ContainerEvent) error
// Stops watching for subcontainer changes.
Stop() error
}