1
0
Fork 0
mirror of https://github.com/moby/moby.git synced 2022-11-09 12:21:53 -05:00
moby--moby/daemon/health.go
Cory Snider a09f8dbe6e daemon: Maintain container exec-inspect invariant
We have integration tests which assert the invariant that a
GET /containers/{id}/json response lists only IDs of execs which are in
the Running state, according to GET /exec/{id}/json. The invariant could
be violated if those requests were to race the handling of the exec's
task-exit event. The coarse-grained locking of the container ExecStore
when starting an exec task was accidentally synchronizing
(*Daemon).ProcessEvent and (*Daemon).ContainerExecInspect to it just
enough to make it improbable for the integration tests to catch the
invariant violation on execs which exit immediately. Removing the
unnecessary locking made the underlying race condition more likely for
the tests to hit.

Maintain the invariant by deleting the exec from its container's
ExecCommands before clearing its Running flag. Additionally, fix other
potential data races with execs by ensuring that the ExecConfig lock is
held whenever a mutable field is read from or written to.

Signed-off-by: Cory Snider <csnider@mirantis.com>
2022-08-24 19:35:07 -04:00

432 lines
12 KiB
Go

package daemon // import "github.com/docker/docker/daemon"
import (
"bytes"
"context"
"fmt"
"runtime"
"strings"
"sync"
"time"
"github.com/docker/docker/api/types"
containertypes "github.com/docker/docker/api/types/container"
"github.com/docker/docker/api/types/strslice"
"github.com/docker/docker/container"
"github.com/sirupsen/logrus"
)
const (
// Longest healthcheck probe output message to store. Longer messages will be truncated.
maxOutputLen = 4096
// Default interval between probe runs (from the end of the first to the start of the second).
// Also the time before the first probe.
defaultProbeInterval = 30 * time.Second
// The maximum length of time a single probe run should take. If the probe takes longer
// than this, the check is considered to have failed.
defaultProbeTimeout = 30 * time.Second
// The time given for the container to start before the health check starts considering
// the container unstable. Defaults to none.
defaultStartPeriod = 0 * time.Second
// Default number of consecutive failures of the health check
// for the container to be considered unhealthy.
defaultProbeRetries = 3
// Maximum number of entries to record
maxLogEntries = 5
)
const (
// Exit status codes that can be returned by the probe command.
exitStatusHealthy = 0 // Container is healthy
)
// probe implementations know how to run a particular type of probe.
type probe interface {
// Perform one run of the check. Returns the exit code and an optional
// short diagnostic string.
run(context.Context, *Daemon, *container.Container) (*types.HealthcheckResult, error)
}
// cmdProbe implements the "CMD" probe type.
type cmdProbe struct {
// Run the command with the system's default shell instead of execing it directly.
shell bool
}
// exec the healthcheck command in the container.
// Returns the exit code and probe output (if any)
func (p *cmdProbe) run(ctx context.Context, d *Daemon, cntr *container.Container) (*types.HealthcheckResult, error) {
startTime := time.Now()
cmdSlice := strslice.StrSlice(cntr.Config.Healthcheck.Test)[1:]
if p.shell {
cmdSlice = append(getShell(cntr), cmdSlice...)
}
entrypoint, args := d.getEntrypointAndArgs(strslice.StrSlice{}, cmdSlice)
execConfig := container.NewExecConfig(cntr)
execConfig.OpenStdin = false
execConfig.OpenStdout = true
execConfig.OpenStderr = true
execConfig.DetachKeys = []byte{}
execConfig.Entrypoint = entrypoint
execConfig.Args = args
execConfig.Tty = false
execConfig.Privileged = false
execConfig.User = cntr.Config.User
execConfig.WorkingDir = cntr.Config.WorkingDir
linkedEnv, err := d.setupLinkedContainers(cntr)
if err != nil {
return nil, err
}
execConfig.Env = container.ReplaceOrAppendEnvValues(cntr.CreateDaemonEnvironment(execConfig.Tty, linkedEnv), execConfig.Env)
d.registerExecCommand(cntr, execConfig)
attributes := map[string]string{
"execID": execConfig.ID,
}
d.LogContainerEventWithAttributes(cntr, "exec_create: "+execConfig.Entrypoint+" "+strings.Join(execConfig.Args, " "), attributes)
output := &limitedBuffer{}
probeCtx, cancelProbe := context.WithCancel(ctx)
defer cancelProbe()
execErr := make(chan error, 1)
options := containertypes.ExecStartOptions{
Stdout: output,
Stderr: output,
}
go func() { execErr <- d.ContainerExecStart(probeCtx, execConfig.ID, options) }()
// Starting an exec can take a significant amount of time: on the order
// of 1s in extreme cases. The time it takes dockerd and containerd to
// start the exec is time that the probe process is not running, and so
// should not count towards the health check's timeout. Apply a separate
// timeout to abort if the exec request is wedged.
tm := time.NewTimer(30 * time.Second)
defer tm.Stop()
select {
case <-tm.C:
return nil, fmt.Errorf("timed out starting health check for container %s", cntr.ID)
case err := <-execErr:
if err != nil {
return nil, err
}
case <-execConfig.Started:
healthCheckStartDuration.UpdateSince(startTime)
}
if !tm.Stop() {
<-tm.C
}
probeTimeout := timeoutWithDefault(cntr.Config.Healthcheck.Timeout, defaultProbeTimeout)
tm.Reset(probeTimeout)
select {
case <-tm.C:
cancelProbe()
logrus.WithContext(ctx).Debugf("Health check for container %s taking too long", cntr.ID)
// Wait for probe to exit (it might take some time to call containerd to kill
// the process and we don't want dying probes to pile up).
<-execErr
return &types.HealthcheckResult{
ExitCode: -1,
Output: fmt.Sprintf("Health check exceeded timeout (%v)", probeTimeout),
End: time.Now(),
}, nil
case err := <-execErr:
if err != nil {
return nil, err
}
}
info, err := d.getExecConfig(execConfig.ID)
if err != nil {
return nil, err
}
exitCode, err := func() (int, error) {
info.Lock()
defer info.Unlock()
if info.ExitCode == nil {
info.Unlock()
return 0, fmt.Errorf("healthcheck for container %s has no exit code", cntr.ID)
}
return *info.ExitCode, nil
}()
if err != nil {
return nil, err
}
// Note: Go's json package will handle invalid UTF-8 for us
out := output.String()
return &types.HealthcheckResult{
End: time.Now(),
ExitCode: exitCode,
Output: out,
}, nil
}
// Update the container's Status.Health struct based on the latest probe's result.
func handleProbeResult(d *Daemon, c *container.Container, result *types.HealthcheckResult, done chan struct{}) {
c.Lock()
defer c.Unlock()
// probe may have been cancelled while waiting on lock. Ignore result then
select {
case <-done:
return
default:
}
retries := c.Config.Healthcheck.Retries
if retries <= 0 {
retries = defaultProbeRetries
}
h := c.State.Health
oldStatus := h.Status()
if len(h.Log) >= maxLogEntries {
h.Log = append(h.Log[len(h.Log)+1-maxLogEntries:], result)
} else {
h.Log = append(h.Log, result)
}
if result.ExitCode == exitStatusHealthy {
h.FailingStreak = 0
h.SetStatus(types.Healthy)
} else { // Failure (including invalid exit code)
shouldIncrementStreak := true
// If the container is starting (i.e. we never had a successful health check)
// then we check if we are within the start period of the container in which
// case we do not increment the failure streak.
if h.Status() == types.Starting {
startPeriod := timeoutWithDefault(c.Config.Healthcheck.StartPeriod, defaultStartPeriod)
timeSinceStart := result.Start.Sub(c.State.StartedAt)
// If still within the start period, then don't increment failing streak.
if timeSinceStart < startPeriod {
shouldIncrementStreak = false
}
}
if shouldIncrementStreak {
h.FailingStreak++
if h.FailingStreak >= retries {
h.SetStatus(types.Unhealthy)
}
}
// Else we're starting or healthy. Stay in that state.
}
// replicate Health status changes
if err := c.CheckpointTo(d.containersReplica); err != nil {
// queries will be inconsistent until the next probe runs or other state mutations
// checkpoint the container
logrus.Errorf("Error replicating health state for container %s: %v", c.ID, err)
}
current := h.Status()
if oldStatus != current {
d.LogContainerEvent(c, "health_status: "+current)
}
}
// Run the container's monitoring thread until notified via "stop".
// There is never more than one monitor thread running per container at a time.
func monitor(d *Daemon, c *container.Container, stop chan struct{}, probe probe) {
probeInterval := timeoutWithDefault(c.Config.Healthcheck.Interval, defaultProbeInterval)
intervalTimer := time.NewTimer(probeInterval)
defer intervalTimer.Stop()
for {
intervalTimer.Reset(probeInterval)
select {
case <-stop:
logrus.Debugf("Stop healthcheck monitoring for container %s (received while idle)", c.ID)
return
case <-intervalTimer.C:
logrus.Debugf("Running health check for container %s ...", c.ID)
startTime := time.Now()
ctx, cancelProbe := context.WithCancel(context.Background())
results := make(chan *types.HealthcheckResult, 1)
go func() {
healthChecksCounter.Inc()
result, err := probe.run(ctx, d, c)
if err != nil {
healthChecksFailedCounter.Inc()
logrus.Warnf("Health check for container %s error: %v", c.ID, err)
results <- &types.HealthcheckResult{
ExitCode: -1,
Output: err.Error(),
Start: startTime,
End: time.Now(),
}
} else {
result.Start = startTime
logrus.Debugf("Health check for container %s done (exitCode=%d)", c.ID, result.ExitCode)
results <- result
}
close(results)
}()
select {
case <-stop:
logrus.Debugf("Stop healthcheck monitoring for container %s (received while probing)", c.ID)
cancelProbe()
// Wait for probe to exit (it might take a while to respond to the TERM
// signal and we don't want dying probes to pile up).
<-results
return
case result := <-results:
handleProbeResult(d, c, result, stop)
cancelProbe()
}
}
}
}
// Get a suitable probe implementation for the container's healthcheck configuration.
// Nil will be returned if no healthcheck was configured or NONE was set.
func getProbe(c *container.Container) probe {
config := c.Config.Healthcheck
if config == nil || len(config.Test) == 0 {
return nil
}
switch config.Test[0] {
case "CMD":
return &cmdProbe{shell: false}
case "CMD-SHELL":
return &cmdProbe{shell: true}
case "NONE":
return nil
default:
logrus.Warnf("Unknown healthcheck type '%s' (expected 'CMD') in container %s", config.Test[0], c.ID)
return nil
}
}
// Ensure the health-check monitor is running or not, depending on the current
// state of the container.
// Called from monitor.go, with c locked.
func (daemon *Daemon) updateHealthMonitor(c *container.Container) {
h := c.State.Health
if h == nil {
return // No healthcheck configured
}
probe := getProbe(c)
wantRunning := c.Running && !c.Paused && probe != nil
if wantRunning {
if stop := h.OpenMonitorChannel(); stop != nil {
go monitor(daemon, c, stop, probe)
}
} else {
h.CloseMonitorChannel()
}
}
// Reset the health state for a newly-started, restarted or restored container.
// initHealthMonitor is called from monitor.go and we should never be running
// two instances at once.
// Called with c locked.
func (daemon *Daemon) initHealthMonitor(c *container.Container) {
// If no healthcheck is setup then don't init the monitor
if getProbe(c) == nil {
return
}
// This is needed in case we're auto-restarting
daemon.stopHealthchecks(c)
if h := c.State.Health; h != nil {
h.SetStatus(types.Starting)
h.FailingStreak = 0
} else {
h := &container.Health{}
h.SetStatus(types.Starting)
c.State.Health = h
}
daemon.updateHealthMonitor(c)
}
// Called when the container is being stopped (whether because the health check is
// failing or for any other reason).
func (daemon *Daemon) stopHealthchecks(c *container.Container) {
h := c.State.Health
if h != nil {
h.CloseMonitorChannel()
}
}
// Buffer up to maxOutputLen bytes. Further data is discarded.
type limitedBuffer struct {
buf bytes.Buffer
mu sync.Mutex
truncated bool // indicates that data has been lost
}
// Append to limitedBuffer while there is room.
func (b *limitedBuffer) Write(data []byte) (int, error) {
b.mu.Lock()
defer b.mu.Unlock()
bufLen := b.buf.Len()
dataLen := len(data)
keep := min(maxOutputLen-bufLen, dataLen)
if keep > 0 {
b.buf.Write(data[:keep])
}
if keep < dataLen {
b.truncated = true
}
return dataLen, nil
}
// The contents of the buffer, with "..." appended if it overflowed.
func (b *limitedBuffer) String() string {
b.mu.Lock()
defer b.mu.Unlock()
out := b.buf.String()
if b.truncated {
out = out + "..."
}
return out
}
// If configuredValue is zero, use defaultValue instead.
func timeoutWithDefault(configuredValue time.Duration, defaultValue time.Duration) time.Duration {
if configuredValue == 0 {
return defaultValue
}
return configuredValue
}
func min(x, y int) int {
if x < y {
return x
}
return y
}
func getShell(cntr *container.Container) []string {
if len(cntr.Config.Shell) != 0 {
return cntr.Config.Shell
}
if runtime.GOOS != "windows" {
return []string{"/bin/sh", "-c"}
}
if cntr.OS != runtime.GOOS {
return []string{"/bin/sh", "-c"}
}
return []string{"cmd", "/S", "/C"}
}