[WTH-378] Caddy 설정 수정

README.md

@@ -119,7 +119,7 @@ EC2 (ARM64)
 무중단 배포를 위해 Blue-Green 방식을 사용합니다.
 
 1. 새 컨테이너(Blue/Green) 시작
-2. `/actuator/health` 헬스 체크 (20회, 3초 간격)
+2. `/actuator/health` 헬스 체크 (기본 120초 타임아웃, 3초 간격)
 3. Caddy upstream 전환
 4. 이전 컨테이너 종료
 

infra/dev/caddy/Caddyfile

@@ -1,10 +1,21 @@
 # {$DOMAIN} 은 Github Action에서 주입한다.
 {$DOMAIN} {
 
+    @sseStream path_regexp /stream$
+    @notSseStream {
+        not path_regexp /stream$
+    }
+
     # 응답 압축 활성화
     # zstd: 최신 브라우저에서 더 효율적인 압축
     # gzip: 대부분 클라이언트와 호환되는 기본 압축
-    encode zstd gzip
+    # SSE는 장시간 스트리밍 응답이므로 압축 대상에서 제외한다.
+    encode @notSseStream zstd gzip
+
+    header @sseStream {
+        Cache-Control "no-cache, no-transform"
+        X-Accel-Buffering "no"
+    }
 
     # 보안 관련 HTTP 헤더 설정
     header {

infra/dev/scripts/deploy.sh

@@ -5,6 +5,10 @@ set -euo pipefail
 : "${APP_IMAGE:?APP_IMAGE is required}"
 : "${DOMAIN:?DOMAIN is required}"
 : "${DEPLOY_DIR:=/opt/weeth/dev}"
+: "${HEALTH_CHECK_TIMEOUT_SECONDS:=120}"
+: "${HEALTH_CHECK_INTERVAL_SECONDS:=3}"
+: "${HEALTH_CHECK_CURL_CONNECT_TIMEOUT_SECONDS:=2}"
+: "${HEALTH_CHECK_CURL_MAX_TIME_SECONDS:=5}"
 
 cd "$DEPLOY_DIR"
 
@@ -37,22 +41,29 @@ echo "[deploy] image=$APP_IMAGE new_color=$NEW_COLOR old_color=$OLD_COLOR"
 docker compose --profile "$NEW_COLOR" -f docker-compose.yml pull "app-$NEW_COLOR"
 docker compose --profile "$NEW_COLOR" -f docker-compose.yml up -d "app-$NEW_COLOR"
 
-for i in {1..20}; do
-  if curl -fsS "http://127.0.0.1:${NEW_HEALTH_PORT}/actuator/health" >/dev/null; then
+health_check_deadline=$((SECONDS + HEALTH_CHECK_TIMEOUT_SECONDS))
+healthy=false
+
+while [ "$SECONDS" -le "$health_check_deadline" ]; do
+  if curl -fsS \
+    --connect-timeout "$HEALTH_CHECK_CURL_CONNECT_TIMEOUT_SECONDS" \
+    --max-time "$HEALTH_CHECK_CURL_MAX_TIME_SECONDS" \
+    "http://127.0.0.1:${NEW_HEALTH_PORT}/actuator/health" >/dev/null; then
     echo "[deploy] new app is healthy"
+    healthy=true
     break
   fi
 
-  if [ "$i" -eq 20 ]; then
-    echo "[deploy] health check failed, stopping new container"
-    docker compose --profile "$NEW_COLOR" -f docker-compose.yml stop "app-$NEW_COLOR" || true
-    docker compose --profile "$NEW_COLOR" -f docker-compose.yml rm -f "app-$NEW_COLOR" || true
-    exit 1
-  fi
-
-  sleep 3
+  sleep "$HEALTH_CHECK_INTERVAL_SECONDS"
 done
 
+if [ "$healthy" != true ]; then
+  echo "[deploy] health check failed after ${HEALTH_CHECK_TIMEOUT_SECONDS}s, stopping new container"
+  docker compose --profile "$NEW_COLOR" -f docker-compose.yml stop "app-$NEW_COLOR" || true
+  docker compose --profile "$NEW_COLOR" -f docker-compose.yml rm -f "app-$NEW_COLOR" || true
+  exit 1
+fi
+
 echo "reverse_proxy weeth-dev-app-${NEW_COLOR}:8080" > ./caddy/upstream.conf
 
 # 현재 Caddy 컨테이너의 DOMAIN과 비교하여 변경 시에만 재생성
@@ -62,10 +73,24 @@ if [ "$CURRENT_DOMAIN" != "$DOMAIN" ]; then
   echo "[deploy] domain changed, recreating caddy"
   docker compose up -d --force-recreate caddy
 elif docker compose ps caddy --format '{{.State}}' 2>/dev/null | grep -q running; then
-  if docker compose exec -T caddy caddy reload --config /etc/caddy/Caddyfile; then
+  reload_ok=false
+  if reload_output=$(docker compose exec -T caddy caddy reload --config /etc/caddy/Caddyfile 2>&1); then
+    # reload가 0을 리턴해도 실제 활성 config에 새 upstream이 반영됐는지 admin API로 검증한다
+    if docker compose exec -T caddy wget -qO- http://localhost:2019/config/ 2>/dev/null \
+        | grep -q "weeth-dev-app-${NEW_COLOR}:8080"; then
+      reload_ok=true
+    else
+      echo "[deploy] caddy reload returned ok but new upstream not active"
+    fi
+  else
+    echo "[deploy] caddy reload failed:"
+    echo "$reload_output"
+  fi
+
+  if [ "$reload_ok" = true ]; then
     echo "[deploy] caddy reloaded"
   else
-    echo "[deploy] caddy reload failed, restarting caddy"
+    echo "[deploy] falling back to restart"
     docker compose restart caddy
   fi
 else

infra/prod/caddy/Caddyfile

@@ -1,10 +1,21 @@
 # {$DOMAIN} 은 Github Action에서 주입한다.
 {$DOMAIN} {
 
+    @sseStream path_regexp /stream$
+    @notSseStream {
+        not path_regexp /stream$
+    }
+
     # 응답 압축 활성화
     # zstd: 최신 브라우저에서 더 효율적인 압축
     # gzip: 대부분 클라이언트와 호환되는 기본 압축
-    encode zstd gzip
+    # SSE는 장시간 스트리밍 응답이므로 압축 대상에서 제외한다.
+    encode @notSseStream zstd gzip
+
+    header @sseStream {
+        Cache-Control "no-cache, no-transform"
+        X-Accel-Buffering "no"
+    }
 
     # 보안 관련 HTTP 헤더 설정
     header {

infra/prod/scripts/deploy.sh

@@ -5,6 +5,10 @@ set -euo pipefail
 : "${APP_IMAGE:?APP_IMAGE is required}"
 : "${DOMAIN:?DOMAIN is required}"
 : "${DEPLOY_DIR:=/opt/weeth/prod}"
+: "${HEALTH_CHECK_TIMEOUT_SECONDS:=120}"
+: "${HEALTH_CHECK_INTERVAL_SECONDS:=3}"
+: "${HEALTH_CHECK_CURL_CONNECT_TIMEOUT_SECONDS:=2}"
+: "${HEALTH_CHECK_CURL_MAX_TIME_SECONDS:=5}"
 
 cd "$DEPLOY_DIR"
 
@@ -37,22 +41,29 @@ echo "[deploy] image=$APP_IMAGE new_color=$NEW_COLOR old_color=$OLD_COLOR"
 docker compose --profile "$NEW_COLOR" -f docker-compose.yml pull "app-$NEW_COLOR"
 docker compose --profile "$NEW_COLOR" -f docker-compose.yml up -d "app-$NEW_COLOR"
 
-for i in {1..20}; do
-  if curl -fsS "http://127.0.0.1:${NEW_HEALTH_PORT}/actuator/health" >/dev/null; then
+health_check_deadline=$((SECONDS + HEALTH_CHECK_TIMEOUT_SECONDS))
+healthy=false
+
+while [ "$SECONDS" -le "$health_check_deadline" ]; do
+  if curl -fsS \
+    --connect-timeout "$HEALTH_CHECK_CURL_CONNECT_TIMEOUT_SECONDS" \
+    --max-time "$HEALTH_CHECK_CURL_MAX_TIME_SECONDS" \
+    "http://127.0.0.1:${NEW_HEALTH_PORT}/actuator/health" >/dev/null; then
     echo "[deploy] new app is healthy"
+    healthy=true
     break
   fi
 
-  if [ "$i" -eq 20 ]; then
-    echo "[deploy] health check failed, stopping new container"
-    docker compose --profile "$NEW_COLOR" -f docker-compose.yml stop "app-$NEW_COLOR" || true
-    docker compose --profile "$NEW_COLOR" -f docker-compose.yml rm -f "app-$NEW_COLOR" || true
-    exit 1
-  fi
-
-  sleep 3
+  sleep "$HEALTH_CHECK_INTERVAL_SECONDS"
 done
 
+if [ "$healthy" != true ]; then
+  echo "[deploy] health check failed after ${HEALTH_CHECK_TIMEOUT_SECONDS}s, stopping new container"
+  docker compose --profile "$NEW_COLOR" -f docker-compose.yml stop "app-$NEW_COLOR" || true
+  docker compose --profile "$NEW_COLOR" -f docker-compose.yml rm -f "app-$NEW_COLOR" || true
+  exit 1
+fi
+
 echo "reverse_proxy weeth-prod-app-${NEW_COLOR}:8080" > ./caddy/upstream.conf
 
 # 현재 Caddy 컨테이너의 DOMAIN과 비교하여 변경 시에만 재생성
@@ -62,10 +73,24 @@ if [ "$CURRENT_DOMAIN" != "$DOMAIN" ]; then
   echo "[deploy] domain changed, recreating caddy"
   docker compose up -d --force-recreate caddy
 elif docker compose ps caddy --format '{{.State}}' 2>/dev/null | grep -q running; then
-  if docker compose exec -T caddy caddy reload --config /etc/caddy/Caddyfile; then
+  reload_ok=false
+  if reload_output=$(docker compose exec -T caddy caddy reload --config /etc/caddy/Caddyfile 2>&1); then
+    # reload가 0을 리턴해도 실제 활성 config에 새 upstream이 반영됐는지 admin API로 검증한다
+    if docker compose exec -T caddy wget -qO- http://localhost:2019/config/ 2>/dev/null \
+        | grep -q "weeth-prod-app-${NEW_COLOR}:8080"; then
+      reload_ok=true
+    else
+      echo "[deploy] caddy reload returned ok but new upstream not active"
+    fi
+  else
+    echo "[deploy] caddy reload failed:"
+    echo "$reload_output"
+  fi
+
+  if [ "$reload_ok" = true ]; then
     echo "[deploy] caddy reloaded"
   else
-    echo "[deploy] caddy reload failed, restarting caddy"
+    echo "[deploy] falling back to restart"
     docker compose restart caddy
   fi
 else