2016-11-26 18:48:49 +00:00
|
|
|
package healthcheck
|
|
|
|
|
|
|
|
import (
|
2016-11-30 21:49:57 +00:00
|
|
|
"context"
|
2021-06-25 19:08:11 +00:00
|
|
|
"errors"
|
2017-03-24 08:36:33 +00:00
|
|
|
"fmt"
|
2017-05-10 18:28:57 +00:00
|
|
|
"net"
|
2016-11-26 18:48:49 +00:00
|
|
|
"net/http"
|
|
|
|
"net/url"
|
2017-05-10 18:28:57 +00:00
|
|
|
"strconv"
|
2016-11-26 18:48:49 +00:00
|
|
|
"time"
|
2017-01-31 21:55:02 +00:00
|
|
|
|
2020-09-26 11:30:03 +00:00
|
|
|
gokitmetrics "github.com/go-kit/kit/metrics"
|
2022-11-21 17:36:05 +00:00
|
|
|
"github.com/rs/zerolog/log"
|
2021-06-25 19:08:11 +00:00
|
|
|
"github.com/traefik/traefik/v2/pkg/config/dynamic"
|
2020-09-16 13:46:04 +00:00
|
|
|
"github.com/traefik/traefik/v2/pkg/config/runtime"
|
2022-09-20 14:54:08 +00:00
|
|
|
"google.golang.org/grpc"
|
|
|
|
"google.golang.org/grpc/codes"
|
2022-10-14 15:16:08 +00:00
|
|
|
"google.golang.org/grpc/credentials/insecure"
|
2022-09-20 14:54:08 +00:00
|
|
|
healthpb "google.golang.org/grpc/health/grpc_health_v1"
|
|
|
|
"google.golang.org/grpc/status"
|
2016-11-26 18:48:49 +00:00
|
|
|
)
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
const modeGRPC = "grpc"
|
2016-11-26 18:48:49 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
// StatusSetter should be implemented by a service that, when the status of a
|
|
|
|
// registered target change, needs to be notified of that change.
|
|
|
|
type StatusSetter interface {
|
|
|
|
SetStatus(ctx context.Context, childName string, up bool)
|
2018-06-11 09:36:03 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
// StatusUpdater should be implemented by a service that, when its status
|
|
|
|
// changes (e.g. all if its children are down), needs to propagate upwards (to
|
|
|
|
// their parent(s)) that change.
|
|
|
|
type StatusUpdater interface {
|
|
|
|
RegisterStatusUpdater(fn func(up bool)) error
|
2021-06-25 19:08:11 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
type metricsHealthCheck interface {
|
|
|
|
ServiceServerUpGauge() gokitmetrics.Gauge
|
2016-11-26 18:48:49 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
type ServiceHealthChecker struct {
|
|
|
|
balancer StatusSetter
|
|
|
|
info *runtime.ServiceInfo
|
2017-03-15 18:16:06 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
config *dynamic.ServerHealthCheck
|
|
|
|
interval time.Duration
|
|
|
|
timeout time.Duration
|
2017-03-24 08:36:33 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
metrics metricsHealthCheck
|
2019-08-07 15:14:04 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
client *http.Client
|
|
|
|
targets map[string]*url.URL
|
2016-11-26 18:48:49 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
func NewServiceHealthChecker(ctx context.Context, metrics metricsHealthCheck, config *dynamic.ServerHealthCheck, service StatusSetter, info *runtime.ServiceInfo, transport http.RoundTripper, targets map[string]*url.URL) *ServiceHealthChecker {
|
2022-11-21 17:36:05 +00:00
|
|
|
logger := log.Ctx(ctx)
|
2016-11-26 18:48:49 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
interval := time.Duration(config.Interval)
|
|
|
|
if interval <= 0 {
|
2022-11-21 17:36:05 +00:00
|
|
|
logger.Error().Msg("Health check interval smaller than zero")
|
2022-11-16 10:38:07 +00:00
|
|
|
interval = time.Duration(dynamic.DefaultHealthCheckInterval)
|
2018-06-11 09:36:03 +00:00
|
|
|
}
|
2016-11-26 18:48:49 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
timeout := time.Duration(config.Timeout)
|
|
|
|
if timeout <= 0 {
|
2022-11-21 17:36:05 +00:00
|
|
|
logger.Error().Msg("Health check timeout smaller than zero")
|
2022-11-16 10:38:07 +00:00
|
|
|
timeout = time.Duration(dynamic.DefaultHealthCheckTimeout)
|
2017-03-09 15:27:31 +00:00
|
|
|
}
|
2016-11-26 18:48:49 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
if timeout >= interval {
|
2022-11-21 17:36:05 +00:00
|
|
|
logger.Warn().Msgf("Health check timeout should be lower than the health check interval. Interval set to timeout + 1 second (%s).", interval)
|
2022-11-16 10:38:07 +00:00
|
|
|
interval = timeout + time.Second
|
2018-06-11 09:36:03 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
client := &http.Client{
|
|
|
|
Transport: transport,
|
2017-03-09 15:27:31 +00:00
|
|
|
}
|
2022-08-08 13:22:07 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
if config.FollowRedirects != nil && !*config.FollowRedirects {
|
|
|
|
client.CheckRedirect = func(req *http.Request, via []*http.Request) error {
|
|
|
|
return http.ErrUseLastResponse
|
|
|
|
}
|
2022-08-08 13:22:07 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
return &ServiceHealthChecker{
|
|
|
|
balancer: service,
|
|
|
|
info: info,
|
|
|
|
config: config,
|
|
|
|
interval: interval,
|
|
|
|
timeout: timeout,
|
|
|
|
targets: targets,
|
|
|
|
client: client,
|
|
|
|
metrics: metrics,
|
2016-11-30 21:49:57 +00:00
|
|
|
}
|
2016-11-26 18:48:49 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
func (shc *ServiceHealthChecker) Launch(ctx context.Context) {
|
|
|
|
ticker := time.NewTicker(shc.interval)
|
2017-03-09 15:27:31 +00:00
|
|
|
defer ticker.Stop()
|
2022-11-16 10:38:07 +00:00
|
|
|
|
2017-04-11 15:10:46 +00:00
|
|
|
for {
|
|
|
|
select {
|
|
|
|
case <-ctx.Done():
|
|
|
|
return
|
2019-09-13 17:28:04 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
case <-ticker.C:
|
|
|
|
for proxyName, target := range shc.targets {
|
|
|
|
select {
|
|
|
|
case <-ctx.Done():
|
|
|
|
return
|
|
|
|
default:
|
|
|
|
}
|
2020-09-26 11:30:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
up := true
|
|
|
|
serverUpMetricValue := float64(1)
|
2020-09-26 11:30:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
if err := shc.executeHealthCheck(ctx, shc.config, target); err != nil {
|
|
|
|
// The context is canceled when the dynamic configuration is refreshed.
|
|
|
|
if errors.Is(err, context.Canceled) {
|
|
|
|
return
|
|
|
|
}
|
2020-09-26 11:30:03 +00:00
|
|
|
|
2022-11-21 17:36:05 +00:00
|
|
|
log.Ctx(ctx).Warn().
|
|
|
|
Str("targetURL", target.String()).
|
|
|
|
Err(err).
|
|
|
|
Msg("Health check failed.")
|
2020-09-26 11:30:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
up = false
|
|
|
|
serverUpMetricValue = float64(0)
|
|
|
|
}
|
2017-03-09 15:27:31 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
shc.balancer.SetStatus(ctx, proxyName, up)
|
2020-09-26 11:30:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
statusStr := runtime.StatusDown
|
|
|
|
if up {
|
|
|
|
statusStr = runtime.StatusUp
|
2019-08-07 15:14:04 +00:00
|
|
|
}
|
2020-09-26 11:30:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
shc.info.UpdateServerStatus(target.String(), statusStr)
|
2020-09-26 11:30:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
shc.metrics.ServiceServerUpGauge().
|
|
|
|
With("service", proxyName).
|
|
|
|
With("url", target.String()).
|
|
|
|
Set(serverUpMetricValue)
|
|
|
|
}
|
2017-04-11 15:10:46 +00:00
|
|
|
}
|
|
|
|
}
|
2017-03-09 15:27:31 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
func (shc *ServiceHealthChecker) executeHealthCheck(ctx context.Context, config *dynamic.ServerHealthCheck, target *url.URL) error {
|
|
|
|
ctx, cancel := context.WithDeadline(ctx, time.Now().Add(shc.timeout))
|
|
|
|
defer cancel()
|
2017-05-10 18:28:57 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
if config.Mode == modeGRPC {
|
|
|
|
return shc.checkHealthGRPC(ctx, target)
|
2018-04-16 09:40:03 +00:00
|
|
|
}
|
2022-11-16 10:38:07 +00:00
|
|
|
return shc.checkHealthHTTP(ctx, target)
|
2018-06-11 09:36:03 +00:00
|
|
|
}
|
2018-05-22 07:22:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
// checkHealthHTTP returns an error with a meaningful description if the health check failed.
|
|
|
|
// Dedicated to HTTP servers.
|
|
|
|
func (shc *ServiceHealthChecker) checkHealthHTTP(ctx context.Context, target *url.URL) error {
|
|
|
|
req, err := shc.newRequest(ctx, target)
|
|
|
|
if err != nil {
|
|
|
|
return fmt.Errorf("create HTTP request: %w", err)
|
2018-04-16 09:40:03 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
resp, err := shc.client.Do(req)
|
|
|
|
if err != nil {
|
|
|
|
return fmt.Errorf("HTTP request failed: %w", err)
|
2022-09-20 14:54:08 +00:00
|
|
|
}
|
2022-11-16 10:38:07 +00:00
|
|
|
|
|
|
|
defer resp.Body.Close()
|
|
|
|
|
|
|
|
if resp.StatusCode < http.StatusOK || resp.StatusCode >= http.StatusBadRequest {
|
|
|
|
return fmt.Errorf("received error status code: %v", resp.StatusCode)
|
|
|
|
}
|
|
|
|
|
|
|
|
return nil
|
2022-09-20 14:54:08 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
func (shc *ServiceHealthChecker) newRequest(ctx context.Context, target *url.URL) (*http.Request, error) {
|
|
|
|
u, err := target.Parse(shc.config.Path)
|
2017-05-10 18:28:57 +00:00
|
|
|
if err != nil {
|
2022-11-16 10:38:07 +00:00
|
|
|
return nil, err
|
2017-05-10 18:28:57 +00:00
|
|
|
}
|
2018-05-22 07:22:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
if len(shc.config.Scheme) > 0 {
|
|
|
|
u.Scheme = shc.config.Scheme
|
2016-11-26 18:48:49 +00:00
|
|
|
}
|
2018-01-03 11:32:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
if shc.config.Port != 0 {
|
|
|
|
u.Host = net.JoinHostPort(u.Hostname(), strconv.Itoa(shc.config.Port))
|
2020-02-26 16:28:04 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
req, err := http.NewRequestWithContext(ctx, shc.config.Method, u.String(), http.NoBody)
|
2018-05-22 07:22:03 +00:00
|
|
|
if err != nil {
|
2022-11-16 10:38:07 +00:00
|
|
|
return nil, fmt.Errorf("failed to create HTTP request: %w", err)
|
2018-01-03 11:32:03 +00:00
|
|
|
}
|
2018-05-22 07:22:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
if shc.config.Hostname != "" {
|
|
|
|
req.Host = shc.config.Hostname
|
|
|
|
}
|
2018-05-22 07:22:03 +00:00
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
for k, v := range shc.config.Headers {
|
|
|
|
req.Header.Set(k, v)
|
2018-05-22 07:22:03 +00:00
|
|
|
}
|
|
|
|
|
2022-11-16 10:38:07 +00:00
|
|
|
return req, nil
|
2016-11-26 18:48:49 +00:00
|
|
|
}
|
2019-05-16 08:58:06 +00:00
|
|
|
|
2022-09-20 14:54:08 +00:00
|
|
|
// checkHealthGRPC returns an error with a meaningful description if the health check failed.
|
|
|
|
// Dedicated to gRPC servers implementing gRPC Health Checking Protocol v1.
|
2022-11-16 10:38:07 +00:00
|
|
|
func (shc *ServiceHealthChecker) checkHealthGRPC(ctx context.Context, serverURL *url.URL) error {
|
|
|
|
u, err := serverURL.Parse(shc.config.Path)
|
2022-09-20 14:54:08 +00:00
|
|
|
if err != nil {
|
|
|
|
return fmt.Errorf("failed to parse server URL: %w", err)
|
|
|
|
}
|
|
|
|
|
|
|
|
port := u.Port()
|
2022-11-16 10:38:07 +00:00
|
|
|
if shc.config.Port != 0 {
|
|
|
|
port = strconv.Itoa(shc.config.Port)
|
2022-09-20 14:54:08 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
serverAddr := net.JoinHostPort(u.Hostname(), port)
|
|
|
|
|
|
|
|
var opts []grpc.DialOption
|
2022-11-16 10:38:07 +00:00
|
|
|
switch shc.config.Scheme {
|
2022-09-20 14:54:08 +00:00
|
|
|
case "http", "h2c", "":
|
2022-10-14 15:16:08 +00:00
|
|
|
opts = append(opts, grpc.WithTransportCredentials(insecure.NewCredentials()))
|
2022-09-20 14:54:08 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
conn, err := grpc.DialContext(ctx, serverAddr, opts...)
|
|
|
|
if err != nil {
|
|
|
|
if errors.Is(err, context.DeadlineExceeded) {
|
2022-11-16 10:38:07 +00:00
|
|
|
return fmt.Errorf("fail to connect to %s within %s: %w", serverAddr, shc.config.Timeout, err)
|
2022-09-20 14:54:08 +00:00
|
|
|
}
|
|
|
|
return fmt.Errorf("fail to connect to %s: %w", serverAddr, err)
|
|
|
|
}
|
|
|
|
defer func() { _ = conn.Close() }()
|
|
|
|
|
|
|
|
resp, err := healthpb.NewHealthClient(conn).Check(ctx, &healthpb.HealthCheckRequest{})
|
|
|
|
if err != nil {
|
|
|
|
if stat, ok := status.FromError(err); ok {
|
|
|
|
switch stat.Code() {
|
|
|
|
case codes.Unimplemented:
|
|
|
|
return fmt.Errorf("gRPC server does not implement the health protocol: %w", err)
|
|
|
|
case codes.DeadlineExceeded:
|
|
|
|
return fmt.Errorf("gRPC health check timeout: %w", err)
|
2022-11-16 10:38:07 +00:00
|
|
|
case codes.Canceled:
|
|
|
|
return context.Canceled
|
2022-09-20 14:54:08 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return fmt.Errorf("gRPC health check failed: %w", err)
|
|
|
|
}
|
|
|
|
|
|
|
|
if resp.Status != healthpb.HealthCheckResponse_SERVING {
|
|
|
|
return fmt.Errorf("received gRPC status code: %v", resp.Status)
|
|
|
|
}
|
|
|
|
|
|
|
|
return nil
|
|
|
|
}
|