Docker healthcheck

.gitignore

@@ -10,3 +10,4 @@ backend/public/
 backend/Cold-Friendly-Feud
 backend/famf.db
 test-results/
+backend/.cache/

Dockerfile

@@ -6,6 +6,7 @@ WORKDIR /src
 RUN npm install
 
 FROM base AS dev
+RUN apk add curl
 COPY --from=builder /src/node_modules/ /src/node_modules/
 COPY . /src/
 WORKDIR /src

backend/Dockerfile

@@ -4,11 +4,14 @@ WORKDIR /src
 
 FROM base as dev
 ENV CGO_ENABLED 1
-RUN apk add gcc musl-dev sqlite
+ENV GOCACHE=/src/.cache/go-build
+ENV GOMODCACHE=/src/.cache/go-mod
+RUN apk add gcc musl-dev sqlite curl
 RUN go install github.com/air-verse/air@latest
 COPY --from=games . /src/games/
 COPY . .
-RUN go mod download
+RUN mkdir -p .cache/go-build .cache/go-mod && \
+    go mod download
 CMD ["air", "--build.cmd", "go build .", "--build.bin", "/src/Cold-Friendly-Feud"]
 
 FROM base AS builder

backend/api/health.go

@@ -0,0 +1,94 @@
+package api
+
+import (
+	"fmt"
+	"time"
+
+	"github.com/gorilla/websocket"
+)
+
+type HealthStatus struct {
+	Status  string        `json:"status"`
+	Details HealthDetails `json:"details"`
+}
+
+type HealthDetails struct {
+	WebSocket ComponentStatus `json:"websocket"`
+	Database  ComponentStatus `json:"database"`
+}
+
+type ComponentStatus struct {
+	Status    string    `json:"status"`
+	Timestamp time.Time `json:"timestamp,omitempty"`
+	Error     string    `json:"error,omitempty"`
+}
+
+func HealthTest(port string) (HealthStatus, error) {
+	status := HealthStatus{
+		Status: "up",
+		Details: HealthDetails{
+			WebSocket: checkWebSocket(port),
+			Database:  checkDatabase(),
+		},
+	}
+
+	if status.Details.WebSocket.Status == "down" || status.Details.Database.Status == "down" {
+		status.Status = "down"
+		return status, fmt.Errorf("one or more components are down")
+	}
+
+	return status, nil
+}
+
+func checkWebSocket(port string) ComponentStatus {
+	status := ComponentStatus{
+		Status:    "up",
+		Timestamp: time.Now().UTC(),
+	}
+
+	url := fmt.Sprintf("ws://localhost%s/api/ws", port)
+
+	// Set a shorter timeout for health checks
+	dialer := websocket.Dialer{
+		HandshakeTimeout: 5 * time.Second,
+	}
+
+	// Attempt connection
+	conn, _, err := dialer.Dial(url, nil)
+	if err != nil {
+		status.Status = "down"
+		status.Error = fmt.Sprintf("websocket connection failed: %v", err)
+		return status
+	}
+	defer conn.Close()
+
+	// Ping and wait for pong to verify connection is working
+	err = conn.WriteMessage(websocket.PingMessage, []byte{})
+	if err != nil {
+		status.Status = "down"
+		status.Error = fmt.Sprintf("failed to send ping: %v", err)
+	}
+
+	return status
+}
+
+func checkDatabase() ComponentStatus {
+	status := ComponentStatus{
+		Status:    "up",
+		Timestamp: time.Now().UTC(),
+	}
+
+	if store == nil {
+		status.Status = "down"
+		status.Error = "store not initialized"
+		return status
+	}
+
+	rooms := store.currentRooms()
+	if rooms == nil {
+		status.Status = "down"
+		status.Error = "failed to query rooms"
+	}
+
+	return status
+}

backend/main.go

@@ -1,6 +1,7 @@
 package main
 
 import (
+	"encoding/json"
 	"flag"
 	"log"
 	"net/http"
@@ -39,9 +40,18 @@ func main() {
 		api.ServeWs(httpWriter, httpRequest)
 	})
 
-	http.HandleFunc("/api/healthcheckz", func(httpWriter http.ResponseWriter, httpRequest *http.Request) {
-		httpWriter.Write([]byte("ok"))
-		httpWriter.WriteHeader(200)
+	http.HandleFunc("/api/health", func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+
+		status, err := api.HealthTest(cfg.addr)
+
+		if err != nil {
+			w.WriteHeader(503)
+		} else {
+			w.WriteHeader(200)
+		}
+
+		json.NewEncoder(w).Encode(status)
 	})
 
 	http.HandleFunc("/api/rooms/{roomCode}/logo", func(httpWriter http.ResponseWriter, httpRequest *http.Request) {

chart/templates/deployment.yaml

@@ -84,11 +84,11 @@ spec:
           {{- end }}
           livenessProbe:
             httpGet:
-              path: /api/healthcheckz
+              path: /api/health
               port: bhttp
           readinessProbe:
             httpGet:
-              path: /api/healthcheckz
+              path: /api/health
               port: bhttp
           {{- with .Values.backend.volumeMounts }}
           volumeMounts:

docker/docker-compose-dev-wsl.yaml

@@ -1,23 +1,46 @@
 services:
   frontend:
     image: ${docker_registry}/famf-web:dev
-    network_mode: "host"
+    network_mode: 'host'
     volumes:
       - ../:/src
     environment:
       - HOST=0.0.0.0
+    healthcheck:
+      # Verifies that the frontend dev server is responding on port 3000
+      # Start period is longer due to npm install
+      test: ['CMD', 'curl', '-f', 'http://localhost:3000']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 60s
   backend:
     image: ${docker_registry}/famf-server:dev
-    network_mode: "host"
+    network_mode: 'host'
     environment:
       - HOST=0.0.0.0
       - GAME_STORE=${game_store}
     volumes:
       - ../backend:/src
       - ../games/:/src/games/
+    # Verifies dedicated backend health endpoint returns 200
+    # Start period is longer due to go build
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:8080/api/health']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 60s
   proxy:
     image: nginx:1.27-alpine
-    network_mode: "host"
+    network_mode: 'host'
     volumes:
       - ./nginx/nginx.wsl.conf:/etc/nginx/nginx.conf
       - ../dev/cert/:/etc/nginx/cert/
+    healthcheck:
+      # -k flag allows self-signed certificates in development
+      test: ['CMD', 'curl', '-f', '-k', 'https://localhost:443']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 10s

docker/docker-compose-dev.yaml

@@ -5,6 +5,14 @@ services:
       - ../:/src
     ports:
       - 3000:3000
+    healthcheck:
+      # Verifies that the frontend dev server is responding on port 3000
+      # Start period is longer due to npm install
+      test: ['CMD', 'curl', '-f', 'http://localhost:3000']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 60s
   backend:
     image: ${docker_registry}/famf-server:dev
     ports:
@@ -15,10 +23,25 @@ services:
     volumes:
       - ../backend:/src
       - ../games/:/src/games/
+    # Verifies dedicated backend health endpoint returns 200
+    # Start period is longer due to go build
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:8080/api/health']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 60s
   proxy:
     image: nginx:1.27-alpine
     ports:
       - 443:443
     volumes:
       - ./nginx/nginx.conf:/etc/nginx/nginx.conf
       - ../dev/cert/:/etc/nginx/cert/
+    healthcheck:
+      # -k flag allows self-signed certificates in development
+      test: ['CMD', 'curl', '-f', '-k', 'https://localhost:443']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 10s

docker/docker-compose.yaml

@@ -1,9 +1,39 @@
 services:
   frontend:
-    image: ${docker_registry}/famf-web:latest
+    image: ${docker_registry}/famf-web:dev
     ports:
       - 3000:3000
+    healthcheck:
+      # Verifies that the frontend dev server is responding on port 3000
+      # Start period is longer due to npm install
+      test: ['CMD', 'curl', '-f', 'http://localhost:3000']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 60s
   backend:
-    image: ${docker_registry}/famf-server:latest
+    image: ${docker_registry}/famf-server:dev
+    ports:
+      - 8080:8080
     environment:
+      # One of memory, sqlite
       GAME_STORE: ${game_store}
+    # Verifies dedicated backend health endpoint returns 200
+    # Start period is longer due to go build
+    healthcheck:
+      test: ['CMD', 'curl', '-f', 'http://localhost:8080/api/health']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 60s
+  proxy:
+    image: nginx:1.27-alpine
+    ports:
+      - 443:443
+    healthcheck:
+      # -k flag allows self-signed certificates in development
+      test: ['CMD', 'curl', '-f', '-k', 'https://localhost:443']
+      interval: 5s
+      timeout: 3s
+      retries: 3
+      start_period: 10s

e2e/playwright-dev.config.js

@@ -48,15 +48,15 @@ module.exports = defineConfig({
       use: { ...devices['Desktop Chrome'] },
     },
 
-    {
-      name: 'firefox',
-      use: { ...devices['Desktop Firefox'] },
-    },
-
-    {
-      name: 'webkit',
-      use: { ...devices['Desktop Safari'] },
-    },
+    // {
+    //   name: 'firefox',
+    //   use: { ...devices['Desktop Firefox'] },
+    // },
+    //
+    // {
+    //   name: 'webkit',
+    //   use: { ...devices['Desktop Safari'] },
+    // },
 
     /* Test against mobile viewports. */
     // {

e2e/playwright-prod.config.js

@@ -48,15 +48,15 @@ module.exports = defineConfig({
       use: { ...devices['Desktop Chrome'] },
     },
 
-    {
-      name: 'firefox',
-      use: { ...devices['Desktop Firefox'] },
-    },
-
-    {
-      name: 'webkit',
-      use: { ...devices['Desktop Safari'] },
-    },
+    // {
+    //   name: 'firefox',
+    //   use: { ...devices['Desktop Firefox'] },
+    // },
+    //
+    // {
+    //   name: 'webkit',
+    //   use: { ...devices['Desktop Safari'] },
+    // },
 
     /* Test against mobile viewports. */
     // {

e2e/playwright.config.js

@@ -12,6 +12,8 @@ const { defineConfig, devices } = require('@playwright/test');
  */
 module.exports = defineConfig({
   testDir: './tests',
+  /* 60s timeout (initial build takes 60s) */
+  timeout: 60000,
   /* Run tests in files in parallel */
   fullyParallel: false,
   /* Fail the build on CI if you accidentally left test.only in the source code. */
@@ -44,19 +46,26 @@ module.exports = defineConfig({
   /* Configure projects for major browsers */
   projects: [
     {
-      name: 'chromium',
-      use: { ...devices['Desktop Chrome'] },
+      name: 'docker',
+      testMatch: /global\.setup\.ts/,
     },
-
     {
-      name: 'firefox',
-      use: { ...devices['Desktop Firefox'] },
+      name: 'chromium',
+      use: { ...devices['Desktop Chrome'] },
+      dependencies: ['docker'],
     },
 
-    {
-      name: 'webkit',
-      use: { ...devices['Desktop Safari'] },
-    },
+    // {
+    //   name: 'firefox',
+    //   use: { ...devices['Desktop Firefox'] },
+    //   dependencies: ['docker'],
+    // },
+    //
+    // {
+    //   name: 'webkit',
+    //   use: { ...devices['Desktop Safari'] },
+    //   dependencies: ['docker'],
+    // },
 
     /* Test against mobile viewports. */
     // {