Add improved health check and expose it for docker (#373)

Author: andig

Dockerfile

@@ -43,5 +43,7 @@ EXPOSE 7090/udp
 # SMA Energy Manager
 EXPOSE 9522/udp
 
+HEALTHCHECK --interval=60s --start-period=60s --timeout=30s --retries=3 CMD [ "evcc", "health" ]
+
 ENTRYPOINT [ "/evcc/entrypoint.sh" ]
 CMD [ "evcc" ]

cmd/health.go

@@ -0,0 +1,59 @@
+// +build !windows
+
+package cmd
+
+import (
+	"fmt"
+	"net/http"
+	"os"
+	"time"
+
+	"github.com/andig/evcc/server"
+	"github.com/andig/evcc/util"
+	"github.com/spf13/cobra"
+	"github.com/spf13/viper"
+	"github.com/tv42/httpunix"
+)
+
+const serviceName = "evcc"
+
+// healthCmd represents the meter command
+var healthCmd = &cobra.Command{
+	Use:   "health",
+	Short: "Check application health",
+	Run:   runHealth,
+}
+
+func init() {
+	rootCmd.AddCommand(healthCmd)
+}
+
+func runHealth(cmd *cobra.Command, args []string) {
+	util.LogLevel(viper.GetString("log"), viper.GetStringMapString("levels"))
+	log.INFO.Printf("evcc %s (%s)", server.Version, server.Commit)
+
+	u := &httpunix.Transport{
+		DialTimeout:           100 * time.Millisecond,
+		RequestTimeout:        1 * time.Second,
+		ResponseHeaderTimeout: 1 * time.Second,
+	}
+
+	u.RegisterLocation(serviceName, server.SocketPath)
+
+	var client = http.Client{
+		Transport: u,
+	}
+
+	var ok bool
+	resp, err := client.Get(fmt.Sprintf("http+unix://%s/health", serviceName))
+
+	if err == nil && resp.StatusCode == http.StatusOK {
+		log.INFO.Printf("health check ok")
+		ok = true
+	}
+
+	if !ok {
+		log.ERROR.Printf("health check failed")
+		os.Exit(1)
+	}
+}

cmd/root.go

@@ -184,5 +184,8 @@ func run(cmd *cobra.Command, args []string) {
 	site.DumpConfig()
 	go site.Run(conf.Interval)
 
+	// uds health check listener
+	go server.HealthListener(site)
+
 	log.FATAL.Println(httpd.ListenAndServe())
 }

core/health.go

@@ -0,0 +1,49 @@
+package core
+
+import (
+	"sync/atomic"
+	"time"
+)
+
+// Health is a health checker that needs regular updates to stay healthy
+type Health struct {
+	locker  uint32 // mutex
+	updated time.Time
+	timeout time.Duration
+}
+
+// NewHealth creates new health checker
+func NewHealth(timeout time.Duration) (health *Health) {
+	return &Health{timeout: timeout}
+}
+
+// Healthy returns health status based on last update timestamp
+func (health *Health) Healthy() bool {
+	start := time.Now()
+
+	for time.Since(start) < time.Second {
+		if atomic.CompareAndSwapUint32(&health.locker, 0, 1) {
+			defer atomic.StoreUint32(&health.locker, 0)
+			return time.Since(health.updated) < health.timeout
+		}
+
+		time.Sleep(50 * time.Millisecond)
+	}
+
+	return false
+}
+
+// Update updates the health timer on each loadpoint update
+func (health *Health) Update() {
+	start := time.Now()
+
+	for time.Since(start) < time.Second {
+		if atomic.CompareAndSwapUint32(&health.locker, 0, 1) {
+			health.updated = time.Now()
+			atomic.StoreUint32(&health.locker, 0)
+			return
+		}
+
+		time.Sleep(50 * time.Millisecond)
+	}
+}

core/site.go

@@ -24,6 +24,8 @@ type Site struct {
 	uiChan       chan<- util.Param // client push messages
 	lpUpdateChan chan *LoadPoint
 
+	*Health
+
 	log *util.Logger
 
 	// configuration
@@ -92,6 +94,7 @@ func NewSiteFromConfig(
 func NewSite() *Site {
 	lp := &Site{
 		log:     util.NewLogger("core"),
+		Health:  NewHealth(60 * time.Second),
 		Voltage: 230, // V
 	}
 
@@ -122,7 +125,7 @@ type LoadpointConfiguration struct {
 	TargetSoC   int    `json:"targetSoC"`
 }
 
-// GetMode Gets loadpoint charge mode
+// GetMode gets loadpoint charge mode
 func (site *Site) GetMode() api.ChargeMode {
 	return site.loadpoints[0].GetMode()
 }
@@ -340,6 +343,7 @@ func (site *Site) update(lp Updater) {
 
 	if sitePower, err := site.sitePower(); err == nil {
 		lp.Update(sitePower)
+		site.Health.Update()
 	}
 }
 

docker/tmpl.Dockerfile

@@ -35,5 +35,7 @@ COPY docker/bin/* /evcc/
 
 EXPOSE 7070
 
+HEALTHCHECK --interval=60s --start-period=60s --timeout=30s --retries=3 CMD [ "evcc", "health" ]
+
 ENTRYPOINT [ "/evcc/entrypoint.sh" ]
 CMD [ "evcc" ]

go.mod

@@ -41,6 +41,7 @@ require (
 	github.com/spf13/viper v1.7.1
 	github.com/stretchr/testify v1.6.1 // indirect
 	github.com/tcnksm/go-latest v0.0.0-20170313132115-e3007ae9052e
+	github.com/tv42/httpunix v0.0.0-20191220191345-2ba4b9c3382c
 	github.com/volkszaehler/mbmd v0.0.0-20200831092453-b235d6a65b21
 	golang.org/x/net v0.0.0-20200707034311-ab3426394381
 	gopkg.in/check.v1 v1.0.0-20190902080502-41f04d3bba15 // indirect

go.sum

@@ -389,6 +389,8 @@ github.com/tebeka/strftime v0.1.3/go.mod h1:7wJm3dZlpr4l/oVK0t1HYIc4rMzQ2XJlOMIU
 github.com/technoweenie/multipartstreamer v1.0.1 h1:XRztA5MXiR1TIRHxH2uNxXxaIkKQDeX7m2XsSOlQEnM=
 github.com/technoweenie/multipartstreamer v1.0.1/go.mod h1:jNVxdtShOxzAsukZwTSw6MDx5eUJoiEBsSvzDU9uzog=
 github.com/tmc/grpc-websocket-proxy v0.0.0-20190109142713-0ad062ec5ee5/go.mod h1:ncp9v5uamzpCO7NfCPTXjqaC+bZgJeR0sMTm6dMHP7U=
+github.com/tv42/httpunix v0.0.0-20191220191345-2ba4b9c3382c h1:u6SKchux2yDvFQnDHS3lPnIRmfVJ5Sxy3ao2SIdysLQ=
+github.com/tv42/httpunix v0.0.0-20191220191345-2ba4b9c3382c/go.mod h1:hzIxponao9Kjc7aWznkXaL4U4TWaDSs8zcsY4Ka08nM=
 github.com/ugorji/go v1.1.4/go.mod h1:uQMGLiO92mf5W77hV/PUCpI3pbzQx3CRekS0kk+RGrc=
 github.com/valyala/bytebufferpool v1.0.0/go.mod h1:6bBcMArwyJ5K/AmCkWv1jt77kVWyCJ6HpOuEn7z0Csc=
 github.com/valyala/fasttemplate v1.0.1/go.mod h1:UQGH1tvbgY+Nz5t2n7tXsz52dQxojPUpymEIMZ47gx8=

server/http.go

@@ -32,6 +32,7 @@ type route struct {
 
 // site is the minimal interface for accessing site methods
 type site interface {
+	Healthy() bool
 	Configuration() core.SiteConfiguration
 	LoadPoints() []*core.LoadPoint
 	loadpoint
@@ -100,10 +101,15 @@ func jsonResponse(w http.ResponseWriter, r *http.Request, content interface{}) {
 }
 
 // HealthHandler returns current charge mode
-func HealthHandler() http.HandlerFunc {
+func HealthHandler(site site) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
-		res := struct{ OK bool }{OK: true}
-		jsonResponse(w, r, res)
+		if !site.Healthy() {
+			w.WriteHeader(http.StatusInternalServerError)
+			return
+		}
+
+		w.WriteHeader(http.StatusOK)
+		fmt.Fprintln(w, "OK")
 	}
 }
 
@@ -231,7 +237,7 @@ type HTTPd struct {
 // NewHTTPd creates HTTP server with configured routes for loadpoint
 func NewHTTPd(url string, site site, hub *SocketHub, cache *util.Cache) *HTTPd {
 	var routes = map[string]route{
-		"health":       {[]string{"GET"}, "/health", HealthHandler()},
+		"health":       {[]string{"GET"}, "/health", HealthHandler(site)},
 		"config":       {[]string{"GET"}, "/config", ConfigHandler(site)},
 		"templates":    {[]string{"GET"}, "/config/templates/{class:[a-z]+}", TemplatesHandler()},
 		"state":        {[]string{"GET"}, "/state", StateHandler(cache)},

server/uds.go

@@ -0,0 +1,41 @@
+// +build !windows
+
+package server
+
+import (
+	"net"
+	"net/http"
+	"os"
+)
+
+// SocketPath is the unix domain socket path
+const SocketPath = "/tmp/evcc"
+
+// remoteIfExists deletes file if it exists or fails
+func remoteIfExists(file string) {
+	_, err := os.Stat(file)
+	if err == nil {
+		err = os.Remove(file)
+	}
+
+	if err != nil && !os.IsNotExist(err) {
+		log.FATAL.Fatal(err)
+	}
+}
+
+// HealthListener attaches listener to unix domain socket and runs listener
+func HealthListener(site site) {
+	remoteIfExists(SocketPath)
+
+	l, err := net.Listen("unix", SocketPath)
+	if err != nil {
+		log.FATAL.Fatal(err)
+	}
+	defer l.Close()
+
+	mux := http.NewServeMux()
+	httpd := http.Server{Handler: mux}
+	mux.HandleFunc("/health", HealthHandler(site))
+
+	_ = httpd.Serve(l)
+}