Some more code cleanup and timeouts

README.md

@@ -112,14 +112,10 @@ labelprefix: lazyloader
 
 ### Dependencies
 
-* `lazyloader.needs=a,b,c` -- List of dependencies a container needs (will be started before starting the container)
+* `lazyloader.needs=a,b,c` -- List of dependencies a container needs (will be started before starting the container). Can only be specified on a `lazyloader=true` container
 * `lazyloader.provides=a` -- What dependency name a container provides (Not necessarily a `lazyloader` container)
 * `lazyloader.provides.delay=5s` -- Delay starting other containers for this duration
 
-# Features
-
-- [ ] Dependencies & groups (eg. shut down DB if all dependent apps are down)
-
 # License
 
 GPLv3

config.yaml

@@ -17,6 +17,9 @@ splash: splash.html
 stopdelay: 5m # How long to wait before stopping container
 pollfreq: 10s # How often to check
 
+# Default operation timeout (eg. starting and stopping a container)
+timeout: 30s
+
 # This will be the label-prefix to look at settings on a container
 # usually won't need to change (only if running multiple instances)
 labelprefix: lazyloader

main.go

@@ -18,8 +18,10 @@ import (
 	"github.com/sirupsen/logrus"
 )
 
-var core *service.Core
-var discovery *containers.Discovery
+type controller struct {
+	core      *service.Core
+	discovery *containers.Discovery
+}
 
 func mustCreateDockerClient() *client.Client {
 	cli, err := client.NewClientWithOpts(client.FromEnv, client.WithAPIVersionNegotiation())
@@ -37,10 +39,10 @@ func main() {
 	}
 
 	dockerClient := mustCreateDockerClient()
-	discovery = containers.NewDiscovery(dockerClient)
+	discovery := containers.NewDiscovery(dockerClient)
 
 	var err error
-	core, err = service.New(dockerClient, discovery, config.Model.PollFreq)
+	core, err := service.New(dockerClient, discovery, config.Model.PollFreq)
 	if err != nil {
 		logrus.Fatal(err)
 	}
@@ -50,11 +52,16 @@ func main() {
 		core.StopAll()
 	}
 
+	controller := controller{
+		core,
+		discovery,
+	}
+
 	// Set up http server
 	subFs, _ := fs.Sub(httpAssets, "assets")
 	router := http.NewServeMux()
 	router.Handle(httpAssetPrefix, http.StripPrefix(httpAssetPrefix, http.FileServer(http.FS(subFs))))
-	router.HandleFunc("/", ContainerHandler)
+	router.HandleFunc("/", controller.ContainerHandler)
 
 	srv := &http.Server{
 		Addr:    config.Model.Listen,
@@ -79,7 +86,7 @@ func main() {
 	}
 }
 
-func ContainerHandler(w http.ResponseWriter, r *http.Request) {
+func (s *controller) ContainerHandler(w http.ResponseWriter, r *http.Request) {
 	host := r.Host
 	if host == "" {
 		w.WriteHeader(http.StatusNotFound)
@@ -87,11 +94,11 @@ func ContainerHandler(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 	if host == config.Model.StatusHost && config.Model.StatusHost != "" {
-		StatusHandler(w, r)
+		s.StatusHandler(w, r)
 		return
 	}
 
-	if sOpts, err := core.StartHost(host); err != nil {
+	if sOpts, err := s.core.StartHost(host); err != nil {
 		if errors.Is(err, containers.ErrNotFound) {
 			w.WriteHeader(http.StatusNotFound)
 			io.WriteString(w, "not found")
@@ -111,17 +118,17 @@ func ContainerHandler(w http.ResponseWriter, r *http.Request) {
 	}
 }
 
-func StatusHandler(w http.ResponseWriter, r *http.Request) {
+func (s *controller) StatusHandler(w http.ResponseWriter, r *http.Request) {
 	switch r.URL.Path {
 	case "/":
 		var stats runtime.MemStats
 		runtime.ReadMemStats(&stats)
 
-		qualifying, _ := discovery.QualifyingContainers(r.Context())
-		providers, _ := discovery.ProviderContainers(r.Context())
+		qualifying, _ := s.discovery.QualifyingContainers(r.Context())
+		providers, _ := s.discovery.ProviderContainers(r.Context())
 
 		statusPageTemplate.Execute(w, StatusPageModel{
-			Active:         core.ActiveContainers(),
+			Active:         s.core.ActiveContainers(),
 			Qualifying:     qualifying,
 			Providers:      providers,
 			RuntimeMetrics: fmt.Sprintf("Heap=%d, InUse=%d, Total=%d, Sys=%d, NumGC=%d", stats.HeapAlloc, stats.HeapInuse, stats.TotalAlloc, stats.Sys, stats.NumGC),

pkg/config/config.go

@@ -19,6 +19,7 @@ type ConfigModel struct {
 
 	StopDelay time.Duration // Amount of time to wait before stopping a container
 	PollFreq  time.Duration // How often to check for changes
+	Timeout   time.Duration // Default operation timeout (eg. starting/stopping a container)
 
 	Verbose bool // Debug-level logging
 

pkg/service/service.go

@@ -7,6 +7,7 @@ import (
 	"sort"
 	"sync"
 	"time"
+	"traefik-lazyload/pkg/config"
 	"traefik-lazyload/pkg/containers"
 
 	"github.com/docker/docker/api/types"
@@ -59,16 +60,18 @@ func (s *Core) StartHost(hostname string) (*ContainerState, error) {
 	s.mux.Lock()
 	defer s.mux.Unlock()
 
-	ctx := context.Background()
+	ctx, cancel := context.WithTimeout(context.Background(), config.Model.Timeout)
 
 	ct, err := s.discovery.FindContainerByHostname(ctx, hostname)
 	if err != nil {
 		logrus.Warnf("Unable to find container for host %s: %s", hostname, err)
+		cancel()
 		return nil, err
 	}
 
 	if ets, exists := s.active[ct.ID]; exists {
 		logrus.Debugf("Asked to start host, but we already think it's started: %s", ets.name)
+		cancel()
 		return ets, nil
 	}
 
@@ -80,6 +83,7 @@ func (s *Core) StartHost(hostname string) (*ContainerState, error) {
 
 	go func() {
 		defer func() {
+			cancel()
 			s.mux.Lock()
 			ets.pinned = false
 			ets.lastActivity = time.Now()
@@ -102,10 +106,13 @@ func (s *Core) StopAll() {
 	logrus.Info("Stopping all containers...")
 	for cid, ct := range s.active {
 		logrus.Infof("Stopping %s...", ct.name)
-		s.client.ContainerStop(ctx, cid, container.StopOptions{})
+		if err := s.client.ContainerStop(ctx, cid, container.StopOptions{}); err != nil {
+			logrus.Warnf("Error stopping %s: %v", ct.name, err)
+		} else {
 			delete(s.active, cid)
 		}
 	}
+}
 
 // Returns all actively managed containers
 func (s *Core) ActiveContainers() []*ContainerState {
@@ -136,20 +143,24 @@ func (s *Core) startContainerSync(ctx context.Context, ct *containers.Wrapper) e
 	return nil
 }
 
-func (s *Core) startDependencyFor(ctx context.Context, needs []string, forContainer string) {
+func (s *Core) startDependencyFor(ctx context.Context, needs []string, forContainer string) error {
 	for _, dep := range needs {
 		providers, err := s.discovery.FindDepProvider(ctx, dep)
 
 		if err != nil {
 			logrus.Errorf("Error finding dependency provider for %s: %v", dep, err)
+			return err
 		} else if len(providers) == 0 {
 			logrus.Warnf("Unable to find any container that provides %s for %s", dep, forContainer)
+			return ErrProviderNotFound
 		} else {
 			for _, provider := range providers {
 				if !provider.IsRunning() {
 					logrus.Infof("Starting dependency for %s: %s", forContainer, provider.NameID())
 
-					s.startContainerSync(ctx, &provider)
+					if err := s.startContainerSync(ctx, &provider); err != nil {
+						return err
+					}
 
 					delay, _ := provider.ConfigDuration("provides.delay", 2*time.Second)
 					logrus.Debugf("Delaying %s to start %s", delay.String(), dep)
@@ -158,10 +169,13 @@ func (s *Core) startDependencyFor(ctx context.Context, needs []string, forContai
 			}
 		}
 	}
+
+	return nil
 }
 
-func (s *Core) stopDependenciesFor(ctx context.Context, cid string, cts *ContainerState) {
+func (s *Core) stopDependenciesFor(ctx context.Context, cid string, cts *ContainerState) []error {
 	// Look at our needs, and see if anything else needs them; if not, shut down
+	var errs []error
 
 	deps := make(map[string]bool) // dep -> needed
 	for _, dep := range cts.needs {
@@ -182,6 +196,7 @@ func (s *Core) stopDependenciesFor(ctx context.Context, cid string, cts *Contain
 			containers, err := s.discovery.FindDepProvider(ctx, dep)
 			if err != nil {
 				logrus.Errorf("Unable to find dependency provider containers for %s: %v", dep, err)
+				errs = append(errs, err)
 			} else if len(containers) == 0 {
 				logrus.Warnf("Unable to find any containers for dependency %s", dep)
 			} else {
@@ -195,6 +210,7 @@ func (s *Core) stopDependenciesFor(ctx context.Context, cid string, cts *Contain
 		}
 	}
 
+	return errs
 }
 
 // Ticker loop that will check internal state against docker state (Call Poll)
@@ -216,7 +232,7 @@ func (s *Core) pollThread(rate time.Duration) {
 // stopping idle containers
 // Will normally happen in the background with the pollThread
 func (s *Core) Poll() {
-	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
+	ctx, cancel := context.WithTimeout(context.Background(), config.Model.Timeout)
 	defer cancel()
 
 	s.checkForNewContainersSync(ctx)