fix: move health/readiness probes to its own port so they are not logged

Author: henrywhitaker3

chart/templates/deployment.yaml

@@ -53,14 +53,17 @@ spec:
             - name: metrics
               containerPort: {{ .Values.configuration.metrics.port }}
               protocol: TCP
+            - name: probes
+              containerPort: {{ .Values.configuration.probes_port }}
+              protocol: TCP
           livenessProbe:
             httpGet:
               path: /
-              port: http
+              port: probes
           readinessProbe:
             httpGet:
               path: /
-              port: http
+              port: probes
           resources:
             {{- toYaml .Values.resources | nindent 12 }}
       {{- with .Values.nodeSelector }}

chart/values.yaml

@@ -25,6 +25,9 @@ configuration:
   # Whether to print http logs to stdout
   http_logs: true
 
+  # The port the health/readiness probes listen on
+  probes_port: 9744
+
   # Whether to expose prompage metrics
   metrics:
     enabled: false

cmd/serve/serve.go

@@ -10,6 +10,7 @@ import (
 	"syscall"
 
 	"github.com/henrywhitaker3/prompage/internal/app"
+	"github.com/henrywhitaker3/prompage/internal/health"
 	"github.com/henrywhitaker3/prompage/internal/http"
 	"github.com/henrywhitaker3/prompage/internal/metrics"
 	"github.com/spf13/cobra"
@@ -34,6 +35,7 @@ func NewServeCommand(app *app.App) *cobra.Command {
 			http := http.NewHttp(app, cache)
 			metrics := metrics.NewServer(app.Config.Metrics.Port)
 			metrics.Init()
+			probes := health.NewHealth(app.Config.ProbesPort)
 
 			go cache.Work(ctx)
 			go func() {
@@ -54,9 +56,18 @@ func NewServeCommand(app *app.App) *cobra.Command {
 					}
 				}()
 			}
+			go func() {
+				if err := probes.Start(); err != nil {
+					if !errors.Is(err, stdhttp.ErrServerClosed) {
+						fmt.Println(fmt.Errorf("probes server failed: %v", err))
+						cancel()
+					}
+				}
+			}()
 
 			<-ctx.Done()
 			metrics.Stop(context.Background())
+			probes.Stop(context.Background())
 			return http.Stop(context.Background())
 		},
 	}

internal/config/config.go

@@ -56,10 +56,11 @@ type Datasource struct {
 }
 
 type Config struct {
-	Port     int       `yaml:"port"`
-	HttpLogs *bool     `yaml:"http_logs"`
-	Metrics  Metrics   `yaml:"metrics"`
-	Services []Service `yaml:"services"`
+	Port       int       `yaml:"port"`
+	ProbesPort int       `yaml:"probes_port"`
+	HttpLogs   *bool     `yaml:"http_logs"`
+	Metrics    Metrics   `yaml:"metrics"`
+	Services   []Service `yaml:"services"`
 
 	Datasources []Datasource `yaml:"datasources"`
 
@@ -101,6 +102,9 @@ func setDefaults(conf *Config) {
 	if conf.Metrics.Port == 0 {
 		conf.Metrics.Port = 9743
 	}
+	if conf.ProbesPort == 0 {
+		conf.ProbesPort = 9744
+	}
 	if conf.HttpLogs == nil {
 		en := true
 		conf.HttpLogs = &en

internal/health/health.go

@@ -0,0 +1,61 @@
+package health
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+
+	"github.com/labstack/echo/v4"
+)
+
+type Health struct {
+	e    *echo.Echo
+	port int
+
+	ready   bool
+	healthy bool
+}
+
+// TODO: add an actual health/readiness check so these aren't meaningless
+
+func NewHealth(port int) *Health {
+	e := echo.New()
+	e.HideBanner = true
+	h := &Health{
+		e:       e,
+		healthy: true,
+		ready:   true,
+		port:    port,
+	}
+
+	h.e.GET("/healthz", h.healthHandler())
+	h.e.GET("/readyz", h.readyHandler())
+
+	return h
+}
+
+func (h *Health) Start() error {
+	return h.e.Start(fmt.Sprintf(":%d", h.port))
+}
+
+func (h *Health) Stop(ctx context.Context) error {
+	return h.e.Shutdown(ctx)
+}
+
+func (h *Health) healthHandler() echo.HandlerFunc {
+	return func(c echo.Context) error {
+		if h.healthy {
+			return c.String(http.StatusOK, "HEALTHY")
+		}
+		return c.String(http.StatusServiceUnavailable, "UNHEALTHY")
+	}
+}
+
+func (h *Health) readyHandler() echo.HandlerFunc {
+	return func(c echo.Context) error {
+		if h.ready {
+			return c.String(http.StatusOK, "READY")
+		}
+		return c.String(http.StatusServiceUnavailable, "NOT READY")
+	}
+}