moby/libcontainerd/remote_unix.go

566 lines
14 KiB
Go
Raw Normal View History

// +build linux solaris
package libcontainerd
import (
"fmt"
"io"
"io/ioutil"
"log"
"net"
"os"
"os/exec"
"path/filepath"
goruntime "runtime"
"strconv"
"strings"
"sync"
"time"
"github.com/Sirupsen/logrus"
containerd "github.com/containerd/containerd/api/grpc/types"
"github.com/docker/docker/pkg/locker"
"github.com/docker/docker/pkg/system"
"github.com/golang/protobuf/ptypes"
"github.com/golang/protobuf/ptypes/timestamp"
"golang.org/x/net/context"
"golang.org/x/sys/unix"
"google.golang.org/grpc"
"google.golang.org/grpc/grpclog"
"google.golang.org/grpc/health/grpc_health_v1"
"google.golang.org/grpc/transport"
)
const (
maxConnectionRetryCount = 3
containerdHealthCheckTimeout = 3 * time.Second
containerdShutdownTimeout = 15 * time.Second
containerdBinary = "docker-containerd"
containerdPidFilename = "docker-containerd.pid"
containerdSockFilename = "docker-containerd.sock"
containerdStateDir = "containerd"
eventTimestampFilename = "event.ts"
)
type remote struct {
sync.RWMutex
apiClient containerd.APIClient
daemonPid int
stateDir string
rpcAddr string
startDaemon bool
closedManually bool
debugLog bool
rpcConn *grpc.ClientConn
clients []*client
eventTsPath string
runtime string
runtimeArgs []string
daemonWaitCh chan struct{}
liveRestore bool
oomScore int
restoreFromTimestamp *timestamp.Timestamp
}
// New creates a fresh instance of libcontainerd remote.
func New(stateDir string, options ...RemoteOption) (_ Remote, err error) {
defer func() {
if err != nil {
err = fmt.Errorf("Failed to connect to containerd. Please make sure containerd is installed in your PATH or you have specified the correct address. Got error: %v", err)
}
}()
r := &remote{
stateDir: stateDir,
daemonPid: -1,
eventTsPath: filepath.Join(stateDir, eventTimestampFilename),
}
for _, option := range options {
if err := option.Apply(r); err != nil {
return nil, err
}
}
if err := system.MkdirAll(stateDir, 0700, ""); err != nil {
return nil, err
}
if r.rpcAddr == "" {
r.rpcAddr = filepath.Join(stateDir, containerdSockFilename)
}
if r.startDaemon {
if err := r.runContainerdDaemon(); err != nil {
return nil, err
}
}
// don't output the grpc reconnect logging
grpclog.SetLogger(log.New(ioutil.Discard, "", log.LstdFlags))
dialOpts := []grpc.DialOption{
grpc.WithInsecure(),
grpc.WithBackoffMaxDelay(2 * time.Second),
grpc.WithDialer(func(addr string, timeout time.Duration) (net.Conn, error) {
return net.DialTimeout("unix", addr, timeout)
}),
}
conn, err := grpc.Dial(r.rpcAddr, dialOpts...)
if err != nil {
return nil, fmt.Errorf("error connecting to containerd: %v", err)
}
r.rpcConn = conn
r.apiClient = containerd.NewAPIClient(conn)
// Get the timestamp to restore from
t := r.getLastEventTimestamp()
tsp, err := ptypes.TimestampProto(t)
if err != nil {
logrus.Errorf("libcontainerd: failed to convert timestamp: %q", err)
}
r.restoreFromTimestamp = tsp
go r.handleConnectionChange()
if err := r.startEventsMonitor(); err != nil {
return nil, err
}
return r, nil
}
func (r *remote) UpdateOptions(options ...RemoteOption) error {
for _, option := range options {
if err := option.Apply(r); err != nil {
return err
}
}
return nil
}
func (r *remote) handleConnectionChange() {
var transientFailureCount = 0
ticker := time.NewTicker(500 * time.Millisecond)
defer ticker.Stop()
healthClient := grpc_health_v1.NewHealthClient(r.rpcConn)
for {
<-ticker.C
ctx, cancel := context.WithTimeout(context.Background(), containerdHealthCheckTimeout)
_, err := healthClient.Check(ctx, &grpc_health_v1.HealthCheckRequest{})
cancel()
if err == nil {
continue
}
logrus.Debugf("libcontainerd: containerd health check returned error: %v", err)
if r.daemonPid != -1 {
if r.closedManually {
// Well, we asked for it to stop, just return
return
}
// all other errors are transient
// Reset state to be notified of next failure
transientFailureCount++
if transientFailureCount >= maxConnectionRetryCount {
transientFailureCount = 0
if system.IsProcessAlive(r.daemonPid) {
system.KillProcess(r.daemonPid)
}
<-r.daemonWaitCh
if err := r.runContainerdDaemon(); err != nil { //FIXME: Handle error
logrus.Errorf("libcontainerd: error restarting containerd: %v", err)
}
continue
}
}
}
}
func (r *remote) Cleanup() {
if r.daemonPid == -1 {
return
}
r.closedManually = true
r.rpcConn.Close()
// Ask the daemon to quit
unix.Kill(r.daemonPid, unix.SIGTERM)
// Wait up to 15secs for it to stop
for i := time.Duration(0); i < containerdShutdownTimeout; i += time.Second {
if !system.IsProcessAlive(r.daemonPid) {
break
}
time.Sleep(time.Second)
}
if system.IsProcessAlive(r.daemonPid) {
logrus.Warnf("libcontainerd: containerd (%d) didn't stop within 15 secs, killing it\n", r.daemonPid)
unix.Kill(r.daemonPid, unix.SIGKILL)
}
// cleanup some files
os.Remove(filepath.Join(r.stateDir, containerdPidFilename))
os.Remove(filepath.Join(r.stateDir, containerdSockFilename))
}
func (r *remote) Client(b Backend) (Client, error) {
c := &client{
clientCommon: clientCommon{
backend: b,
containers: make(map[string]*container),
locker: locker.New(),
},
remote: r,
exitNotifiers: make(map[string]*exitNotifier),
liveRestore: r.liveRestore,
}
r.Lock()
r.clients = append(r.clients, c)
r.Unlock()
return c, nil
}
func (r *remote) updateEventTimestamp(t time.Time) {
f, err := os.OpenFile(r.eventTsPath, unix.O_CREAT|unix.O_WRONLY|unix.O_TRUNC, 0600)
if err != nil {
logrus.Warnf("libcontainerd: failed to open event timestamp file: %v", err)
return
}
defer f.Close()
b, err := t.MarshalText()
if err != nil {
logrus.Warnf("libcontainerd: failed to encode timestamp: %v", err)
return
}
n, err := f.Write(b)
if err != nil || n != len(b) {
logrus.Warnf("libcontainerd: failed to update event timestamp file: %v", err)
f.Truncate(0)
return
}
}
func (r *remote) getLastEventTimestamp() time.Time {
t := time.Now()
fi, err := os.Stat(r.eventTsPath)
if os.IsNotExist(err) || fi.Size() == 0 {
return t
}
f, err := os.Open(r.eventTsPath)
if err != nil {
logrus.Warnf("libcontainerd: Unable to access last event ts: %v", err)
return t
}
defer f.Close()
b := make([]byte, fi.Size())
n, err := f.Read(b)
if err != nil || n != len(b) {
logrus.Warnf("libcontainerd: Unable to read last event ts: %v", err)
return t
}
t.UnmarshalText(b)
return t
}
func (r *remote) startEventsMonitor() error {
// First, get past events
t := r.getLastEventTimestamp()
tsp, err := ptypes.TimestampProto(t)
if err != nil {
logrus.Errorf("libcontainerd: failed to convert timestamp: %q", err)
}
er := &containerd.EventsRequest{
Timestamp: tsp,
}
var events containerd.API_EventsClient
for {
events, err = r.apiClient.Events(context.Background(), er, grpc.FailFast(false))
if err == nil {
break
}
logrus.Warnf("libcontainerd: failed to get events from containerd: %q", err)
if r.closedManually {
// ignore error if grpc remote connection is closed manually
return nil
}
<-time.After(100 * time.Millisecond)
}
go r.handleEventStream(events)
return nil
}
func (r *remote) handleEventStream(events containerd.API_EventsClient) {
for {
e, err := events.Recv()
if err != nil {
if grpc.ErrorDesc(err) == transport.ErrConnClosing.Desc &&
r.closedManually {
// ignore error if grpc remote connection is closed manually
return
}
logrus.Errorf("libcontainerd: failed to receive event from containerd: %v", err)
go r.startEventsMonitor()
return
}
logrus.Debugf("libcontainerd: received containerd event: %#v", e)
var container *container
var c *client
r.RLock()
for _, c = range r.clients {
container, err = c.getContainer(e.Id)
if err == nil {
break
}
}
r.RUnlock()
if container == nil {
logrus.Warnf("libcontainerd: unknown container %s", e.Id)
continue
}
if err := container.handleEvent(e); err != nil {
logrus.Errorf("libcontainerd: error processing state change for %s: %v", e.Id, err)
}
tsp, err := ptypes.Timestamp(e.Timestamp)
if err != nil {
logrus.Errorf("libcontainerd: failed to convert event timestamp: %q", err)
continue
}
r.updateEventTimestamp(tsp)
}
}
func (r *remote) runContainerdDaemon() error {
pidFilename := filepath.Join(r.stateDir, containerdPidFilename)
f, err := os.OpenFile(pidFilename, os.O_RDWR|os.O_CREATE, 0600)
if err != nil {
return err
}
defer f.Close()
// File exist, check if the daemon is alive
b := make([]byte, 8)
n, err := f.Read(b)
if err != nil && err != io.EOF {
return err
}
if n > 0 {
pid, err := strconv.ParseUint(string(b[:n]), 10, 64)
if err != nil {
return err
}
if system.IsProcessAlive(int(pid)) {
logrus.Infof("libcontainerd: previous instance of containerd still alive (%d)", pid)
r.daemonPid = int(pid)
return nil
}
}
// rewind the file
_, err = f.Seek(0, os.SEEK_SET)
if err != nil {
return err
}
// Truncate it
err = f.Truncate(0)
if err != nil {
return err
}
// Start a new instance
args := []string{
"-l", fmt.Sprintf("unix://%s", r.rpcAddr),
"--metrics-interval=0",
"--start-timeout", "2m",
"--state-dir", filepath.Join(r.stateDir, containerdStateDir),
}
if goruntime.GOOS == "solaris" {
args = append(args, "--shim", "containerd-shim", "--runtime", "runc")
} else {
args = append(args, "--shim", "docker-containerd-shim")
if r.runtime != "" {
args = append(args, "--runtime")
args = append(args, r.runtime)
}
}
if r.debugLog {
args = append(args, "--debug")
}
if len(r.runtimeArgs) > 0 {
for _, v := range r.runtimeArgs {
args = append(args, "--runtime-args")
args = append(args, v)
}
logrus.Debugf("libcontainerd: runContainerdDaemon: runtimeArgs: %s", args)
}
cmd := exec.Command(containerdBinary, args...)
// redirect containerd logs to docker logs
cmd.Stdout = os.Stdout
cmd.Stderr = os.Stderr
cmd.SysProcAttr = setSysProcAttr(true)
cmd.Env = nil
// clear the NOTIFY_SOCKET from the env when starting containerd
for _, e := range os.Environ() {
if !strings.HasPrefix(e, "NOTIFY_SOCKET") {
cmd.Env = append(cmd.Env, e)
}
}
if err := cmd.Start(); err != nil {
return err
}
libcontainerd: fix reaper goroutine position It has observed defunct containerd processes accumulating over time while dockerd was permanently failing to restart containerd. Due to a bug in the runContainerdDaemon() function, dockerd does not clean up its child process if containerd already exits very soon after the (re)start. The reproducer and analysis below comes from docker 1.12.x but bug still applies on latest master. - from libcontainerd/remote_linux.go: 329 func (r *remote) runContainerdDaemon() error { : : // start the containerd child process : 403 if err := cmd.Start(); err != nil { 404 return err 405 } : : // If containerd exits very soon after (re)start, it is possible : // that containerd is already in defunct state at the time when : // dockerd gets here. The setOOMScore() function tries to write : // to /proc/PID_OF_CONTAINERD/oom_score_adj. However, this fails : // with errno EINVAL because containerd is defunct. Please see : // snippets of kernel source code and further explanation below. : 407 if err := setOOMScore(cmd.Process.Pid, r.oomScore); err != nil { 408 utils.KillProcess(cmd.Process.Pid) : : // Due to the error from write() we return here. As the : // goroutine that would clean up the child has not been : // started yet, containerd remains in the defunct state : // and never gets reaped. : 409 return err 410 } : 417 go func() { 418 cmd.Wait() 419 close(r.daemonWaitCh) 420 }() // Reap our child when needed : 423 } This is the kernel function that gets invoked when dockerd tries to write to /proc/PID_OF_CONTAINERD/oom_score_adj. - from fs/proc/base.c: 1197 static ssize_t oom_score_adj_write(struct file *file, ... 1198 size_t count, loff_t *ppos) 1199 { : 1223 task = get_proc_task(file_inode(file)); : : // The defunct containerd process does not have a virtual : // address space anymore, i.e. task->mm is NULL. Thus the : // following code returns errno EINVAL to dockerd. : 1230 if (!task->mm) { 1231 err = -EINVAL; 1232 goto err_task_lock; 1233 } : 1253 err_task_lock: : 1257 return err < 0 ? err : count; 1258 } The purpose of the following program is to demonstrate the behavior of the oom_score_adj_write() function in connection with a defunct process. $ cat defunct_test.c \#include <unistd.h> main() { pid_t pid = fork(); if (pid == 0) // child _exit(0); // parent pause(); } $ make defunct_test cc defunct_test.c -o defunct_test $ ./defunct_test & [1] 3142 $ ps -f | grep defunct_test | grep -v grep root 3142 2956 0 13:04 pts/0 00:00:00 ./defunct_test root 3143 3142 0 13:04 pts/0 00:00:00 [defunct_test] <defunct> $ echo "ps 3143" | crash -s PID PPID CPU TASK ST %MEM VSZ RSS COMM 3143 3142 2 ffff880035def300 ZO 0.0 0 0 defunct_test $ echo "px ((struct task_struct *)0xffff880035def300)->mm" | crash -s $1 = (struct mm_struct *) 0x0 ^^^ task->mm is NULL $ cat /proc/3143/oom_score_adj 0 $ echo 0 > /proc/3143/oom_score_adj -bash: echo: write error: Invalid argument" --- This patch fixes the above issue by making sure we start the reaper goroutine as soon as possible. Signed-off-by: Antonio Murdaca <runcom@redhat.com>
2017-05-27 13:09:12 +00:00
// unless strictly necessary, do not add anything in between here
// as the reaper goroutine below needs to kick in as soon as possible
// and any "return" from code paths added here will defeat the reaper
// process.
r.daemonWaitCh = make(chan struct{})
go func() {
cmd.Wait()
close(r.daemonWaitCh)
}() // Reap our child when needed
logrus.Infof("libcontainerd: new containerd process, pid: %d", cmd.Process.Pid)
if err := setOOMScore(cmd.Process.Pid, r.oomScore); err != nil {
system.KillProcess(cmd.Process.Pid)
return err
}
if _, err := f.WriteString(fmt.Sprintf("%d", cmd.Process.Pid)); err != nil {
system.KillProcess(cmd.Process.Pid)
return err
}
r.daemonPid = cmd.Process.Pid
return nil
}
// WithRemoteAddr sets the external containerd socket to connect to.
func WithRemoteAddr(addr string) RemoteOption {
return rpcAddr(addr)
}
type rpcAddr string
func (a rpcAddr) Apply(r Remote) error {
if remote, ok := r.(*remote); ok {
remote.rpcAddr = string(a)
return nil
}
return fmt.Errorf("WithRemoteAddr option not supported for this remote")
}
// WithRuntimePath sets the path of the runtime to be used as the
// default by containerd
func WithRuntimePath(rt string) RemoteOption {
return runtimePath(rt)
}
type runtimePath string
func (rt runtimePath) Apply(r Remote) error {
if remote, ok := r.(*remote); ok {
remote.runtime = string(rt)
return nil
}
return fmt.Errorf("WithRuntime option not supported for this remote")
}
// WithRuntimeArgs sets the list of runtime args passed to containerd
func WithRuntimeArgs(args []string) RemoteOption {
return runtimeArgs(args)
}
type runtimeArgs []string
func (rt runtimeArgs) Apply(r Remote) error {
if remote, ok := r.(*remote); ok {
remote.runtimeArgs = rt
return nil
}
return fmt.Errorf("WithRuntimeArgs option not supported for this remote")
}
// WithStartDaemon defines if libcontainerd should also run containerd daemon.
func WithStartDaemon(start bool) RemoteOption {
return startDaemon(start)
}
type startDaemon bool
func (s startDaemon) Apply(r Remote) error {
if remote, ok := r.(*remote); ok {
remote.startDaemon = bool(s)
return nil
}
return fmt.Errorf("WithStartDaemon option not supported for this remote")
}
// WithDebugLog defines if containerd debug logs will be enabled for daemon.
func WithDebugLog(debug bool) RemoteOption {
return debugLog(debug)
}
type debugLog bool
func (d debugLog) Apply(r Remote) error {
if remote, ok := r.(*remote); ok {
remote.debugLog = bool(d)
return nil
}
return fmt.Errorf("WithDebugLog option not supported for this remote")
}
// WithLiveRestore defines if containers are stopped on shutdown or restored.
func WithLiveRestore(v bool) RemoteOption {
return liveRestore(v)
}
type liveRestore bool
func (l liveRestore) Apply(r Remote) error {
if remote, ok := r.(*remote); ok {
remote.liveRestore = bool(l)
for _, c := range remote.clients {
c.liveRestore = bool(l)
}
return nil
}
return fmt.Errorf("WithLiveRestore option not supported for this remote")
}
// WithOOMScore defines the oom_score_adj to set for the containerd process.
func WithOOMScore(score int) RemoteOption {
return oomScore(score)
}
type oomScore int
func (o oomScore) Apply(r Remote) error {
if remote, ok := r.(*remote); ok {
remote.oomScore = int(o)
return nil
}
return fmt.Errorf("WithOOMScore option not supported for this remote")
}