[new feature] promtail: Add config reload endoint / signal to promtail

clients/cmd/promtail/main.go

@@ -5,6 +5,7 @@ import (
 	"fmt"
 	"os"
 	"reflect"
+	"sync"
 
 	// embed time zone data
 	_ "time/tzdata"
@@ -13,34 +14,38 @@ import (
 
 	"github.com/go-kit/log/level"
 	"github.com/grafana/dskit/flagext"
+	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/common/version"
 	"github.com/weaveworks/common/logging"
 
 	"github.com/grafana/loki/clients/pkg/logentry/stages"
 	"github.com/grafana/loki/clients/pkg/promtail"
 	"github.com/grafana/loki/clients/pkg/promtail/client"
-	"github.com/grafana/loki/clients/pkg/promtail/config"
+	promtail_config "github.com/grafana/loki/clients/pkg/promtail/config"
 
 	"github.com/grafana/loki/pkg/util"
-	_ "github.com/grafana/loki/pkg/util/build"
 	"github.com/grafana/loki/pkg/util/cfg"
+
+	_ "github.com/grafana/loki/pkg/util/build"
 	util_log "github.com/grafana/loki/pkg/util/log"
 )
 
 func init() {
 	prometheus.MustRegister(version.NewCollector("promtail"))
 }
 
+var mtx sync.Mutex
+
 type Config struct {
-	config.Config   `yaml:",inline"`
-	printVersion    bool
-	printConfig     bool
-	logConfig       bool
-	dryRun          bool
-	configFile      string
-	configExpandEnv bool
-	inspect         bool
+	promtail_config.Config `yaml:",inline"`
+	printVersion           bool
+	printConfig            bool
+	logConfig              bool
+	dryRun                 bool
+	configFile             string
+	configExpandEnv        bool
+	inspect                bool
 }
 
 func (c *Config) RegisterFlags(f *flag.FlagSet) {
@@ -66,11 +71,11 @@ func (c *Config) Clone() flagext.Registerer {
 func main() {
 	// Load config, merging config file and CLI flags
 	var config Config
-	if err := cfg.DefaultUnmarshal(&config, os.Args[1:], flag.CommandLine); err != nil {
+	args := os.Args[1:]
+	if err := cfg.DefaultUnmarshal(&config, args, flag.CommandLine); err != nil {
 		fmt.Println("Unable to parse config:", err)
 		os.Exit(1)
 	}
-
 	// Handle -version CLI flag
 	if config.printVersion {
 		fmt.Println(version.Print("promtail"))
@@ -112,7 +117,17 @@ func main() {
 	}
 
 	clientMetrics := client.NewMetrics(prometheus.DefaultRegisterer, config.Config.Options.StreamLagLabels)
-	p, err := promtail.New(config.Config, clientMetrics, config.dryRun)
+	newConfigFunc := func() (*promtail_config.Config, error) {
+		mtx.Lock()
+		defer mtx.Unlock()
+		var config Config
+		if err := cfg.DefaultUnmarshal(&config, args, flag.NewFlagSet(os.Args[0], flag.ExitOnError)); err != nil {
+			fmt.Println("Unable to parse config:", err)
+			return nil, errors.Wrap(err, "Unable to parse config")
+		}
+		return &config.Config, nil
+	}
+	p, err := promtail.New(config.Config, newConfigFunc, clientMetrics, config.dryRun)
 	if err != nil {
 		level.Error(util_log.Logger).Log("msg", "error creating promtail", "error", err)
 		os.Exit(1)

clients/pkg/promtail/promtail.go

@@ -1,9 +1,14 @@
 package promtail
 
 import (
+	"os"
+	"os/signal"
 	"sync"
+	"syscall"
 
 	"github.com/go-kit/log"
+	"github.com/go-kit/log/level"
+	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"
 
 	"github.com/grafana/loki/clients/pkg/logentry/stages"
@@ -16,6 +21,14 @@ import (
 	util_log "github.com/grafana/loki/pkg/util/log"
 )
 
+var reloadTotal = prometheus.NewCounterVec(prometheus.CounterOpts{
+	Namespace: "promtail",
+	Name:      "config_reload_total",
+	Help:      "Number of reload times.",
+}, []string{"code"})
+
+var errConfigNotChange = errors.New("config has not changed")
+
 // Option is a function that can be passed to the New method of Promtail and
 // customize the Promtail that is created.
 type Option func(p *Promtail)
@@ -42,17 +55,28 @@ type Promtail struct {
 	logger         log.Logger
 	reg            prometheus.Registerer
 
-	stopped bool
-	mtx     sync.Mutex
+	stopped      bool
+	mtx          sync.Mutex
+	configLoaded string
+	newConfig    func() (*config.Config, error)
+	metrics      *client.Metrics
+	dryRun       bool
 }
 
 // New makes a new Promtail.
-func New(cfg config.Config, metrics *client.Metrics, dryRun bool, opts ...Option) (*Promtail, error) {
+func New(cfg config.Config, newConfig func() (*config.Config, error), metrics *client.Metrics, dryRun bool, opts ...Option) (*Promtail, error) {
 	// Initialize promtail with some defaults and allow the options to override
 	// them.
+
 	promtail := &Promtail{
-		logger: util_log.Logger,
-		reg:    prometheus.DefaultRegisterer,
+		logger:  util_log.Logger,
+		reg:     prometheus.DefaultRegisterer,
+		metrics: metrics,
+		dryRun:  dryRun,
+	}
+	err := promtail.reg.Register(reloadTotal)
+	if err != nil {
+		return nil, err
 	}
 	for _, o := range opts {
 		// todo (callum) I don't understand why I needed to add this check
@@ -61,37 +85,70 @@ func New(cfg config.Config, metrics *client.Metrics, dryRun bool, opts ...Option
 		}
 		o(promtail)
 	}
+	err = promtail.reloadConfig(&cfg)
+	if err != nil {
+		return nil, err
+	}
+	server, err := server.New(cfg.ServerConfig, promtail.logger, promtail.targetManagers, cfg.String())
+	if err != nil {
+		return nil, errors.Wrap(err, "error creating loki server")
+	}
+	promtail.server = server
+	promtail.newConfig = newConfig
+
+	return promtail, nil
+}
 
-	cfg.Setup(promtail.logger)
+func (p *Promtail) reloadConfig(cfg *config.Config) error {
+	level.Info(p.logger).Log("msg", "Reloading configuration file")
+	p.mtx.Lock()
+	defer p.mtx.Unlock()
+	newConfigFile := cfg.String()
+	if newConfigFile == p.configLoaded {
+		return errConfigNotChange
+	}
+	newConf := cfg.String()
+	if p.targetManagers != nil {
+		p.targetManagers.Stop()
+	}
+	if p.client != nil {
+		p.client.Stop()
+	}
 
+	cfg.Setup(p.logger)
 	if cfg.LimitsConfig.ReadlineRateEnabled {
 		stages.SetReadLineRateLimiter(cfg.LimitsConfig.ReadlineRate, cfg.LimitsConfig.ReadlineBurst, cfg.LimitsConfig.ReadlineRateDrop)
 	}
 	var err error
-	if dryRun {
-		promtail.client, err = client.NewLogger(metrics, cfg.Options.StreamLagLabels, promtail.logger, cfg.ClientConfigs...)
+	if p.dryRun {
+		p.client, err = client.NewLogger(p.metrics, cfg.Options.StreamLagLabels, p.logger, cfg.ClientConfigs...)
 		if err != nil {
-			return nil, err
+			return err
 		}
 		cfg.PositionsConfig.ReadOnly = true
 	} else {
-		promtail.client, err = client.NewMulti(metrics, cfg.Options.StreamLagLabels, promtail.logger, cfg.LimitsConfig.MaxStreams, cfg.ClientConfigs...)
+		p.client, err = client.NewMulti(p.metrics, cfg.Options.StreamLagLabels, p.logger, cfg.LimitsConfig.MaxStreams, cfg.ClientConfigs...)
 		if err != nil {
-			return nil, err
+			return err
 		}
 	}
 
-	tms, err := targets.NewTargetManagers(promtail, promtail.reg, promtail.logger, cfg.PositionsConfig, promtail.client, cfg.ScrapeConfig, &cfg.TargetConfig)
+	tms, err := targets.NewTargetManagers(p, p.reg, p.logger, cfg.PositionsConfig, p.client, cfg.ScrapeConfig, &cfg.TargetConfig)
 	if err != nil {
-		return nil, err
+		return err
 	}
-	promtail.targetManagers = tms
-	server, err := server.New(cfg.ServerConfig, promtail.logger, tms, cfg.String())
-	if err != nil {
-		return nil, err
+	p.targetManagers = tms
+
+	promServer := p.server
+	if promServer != nil {
+		promtailServer, ok := promServer.(*server.PromtailServer)
+		if !ok {
+			return errors.New("promtailServer cast fail")
+		}
+		promtailServer.ReloadServer(p.targetManagers, cfg.String())
 	}
-	promtail.server = server
-	return promtail, nil
+	p.configLoaded = newConf
+	return nil
 }
 
 // Run the promtail; will block until a signal is received.
@@ -103,6 +160,7 @@ func (p *Promtail) Run() error {
 		return nil
 	}
 	p.mtx.Unlock() // unlock before blocking
+	go p.watchConfig()
 	return p.server.Run()
 }
 
@@ -133,3 +191,48 @@ func (p *Promtail) Shutdown() {
 func (p *Promtail) ActiveTargets() map[string][]target.Target {
 	return p.targetManagers.ActiveTargets()
 }
+
+func (p *Promtail) watchConfig() {
+	// Reload handler.
+	// Make sure that sighup handler is registered with a redirect to the channel before the potentially
+	if p.newConfig == nil {
+		level.Warn(p.logger).Log("msg", "disable watchConfig", "reason", "Promtail newConfig func is Empty")
+		return
+	}
+	promtailServer, ok := p.server.(*server.PromtailServer)
+	if !ok {
+		level.Warn(p.logger).Log("msg", "disable watchConfig", "reason", "promtailServer cast fail")
+		return
+	}
+	level.Warn(p.logger).Log("msg", "enable watchConfig")
+	hup := make(chan os.Signal, 1)
+	signal.Notify(hup, syscall.SIGHUP)
+	for {
+		select {
+		case <-hup:
+			_ = p.reload()
+		case rc := <-promtailServer.Reload():
+			if err := p.reload(); err != nil {
+				rc <- err
+			} else {
+				rc <- nil
+			}
+		}
+	}
+}
+
+func (p *Promtail) reload() error {
+	cfg, err := p.newConfig()
+	if err != nil {
+		reloadTotal.With(prometheus.Labels{"code": "500"}).Inc()
+		return errors.Wrap(err, "Error new Config")
+	}
+	err = p.reloadConfig(cfg)
+	if err != nil {
+		reloadTotal.With(prometheus.Labels{"code": "500"}).Inc()
+		level.Error(p.logger).Log("msg", "Error reloading config", "err", err)
+		return err
+	}
+	reloadTotal.With(prometheus.Labels{"code": "200"}).Inc()
+	return nil
+}