From 540bce911fa784bd5d6d6223ef8f03717d92af07 Mon Sep 17 00:00:00 2001
From: simoleo89 <simoleo89@gmail.com>
Date: Thu, 30 Jul 2026 18:40:42 +0200
Subject: [PATCH] fix(deploy): free PORT before PM2 start to clear EADDRINUSE
 orphans

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 .gitea/workflows/deploy.yaml | 28 ++++++++++++++++++++++++----
 1 file changed, 24 insertions(+), 4 deletions(-)

diff --git a/.gitea/workflows/deploy.yaml b/.gitea/workflows/deploy.yaml
index 60c17d9624..6fef1f7563 100644
--- a/.gitea/workflows/deploy.yaml
+++ b/.gitea/workflows/deploy.yaml
@@ -381,16 +381,35 @@ jobs:
 
           # Next.js prefers an already-set process PORT over .env. PM2 may still
           # have PORT=3000 from an older start, while .env (and nginx) expect 3002.
-          # Export PORT before reload so --update-env aligns the running process.
+          # Export PORT before start so the process matches health checks.
           DEPLOY_PORT="$(grep -E '^[[:space:]]*PORT=' "${LIVE}/.env" 2>/dev/null | tail -1 | cut -d= -f2- || true)"
           DEPLOY_PORT="$(printf '%s' "${DEPLOY_PORT}" | tr -cd '0-9')"
           DEPLOY_PORT="${DEPLOY_PORT:-3002}"
           export PORT="${DEPLOY_PORT}"
           echo "PM2/health PORT=${PORT}"
 
-          echo "Starting PM2 (zero-downtime reload)..."
+          free_tcp_port() {
+            local port="$1"
+            [ -n "${port}" ] || return 0
+            if command -v fuser >/dev/null 2>&1; then
+              fuser -k "${port}/tcp" 2>/dev/null || true
+            elif command -v lsof >/dev/null 2>&1; then
+              # Portable fallback when fuser is unavailable.
+              lsof -tiTCP:"${port}" -sTCP:LISTEN 2>/dev/null | xargs -r kill -9 2>/dev/null || true
+            fi
+          }
+
+          echo "Starting PM2 with a clean PORT=${PORT} listener..."
           cd "${LIVE}"
-          pm2 reload next --update-env || pm2 start pnpm --name next --update-env -- start
+          # Cutover already stopped the app, but failed deploys can leave orphans
+          # on 3002 (or an old PM2 env still bound to 3000).
+          pm2 stop next --kill-timeout 10000 2>/dev/null || true
+          free_tcp_port "${PORT}"
+          free_tcp_port 3000
+          sleep 1
+          pm2 delete next 2>/dev/null || true
+          PORT="${PORT}" pm2 start pnpm --name next -- start
+          pm2 save 2>/dev/null || true
 
           sleep 3
           if ! pm2 show next 2>/dev/null | grep -q 'online'; then
@@ -415,7 +434,8 @@ jobs:
           if [ "${HEALTH_OK}" != "1" ]; then
             echo "ERROR: Health check failed after deploy (${HEALTH_URL})" >&2
             echo "Last body: ${BODY:-<empty>}" >&2
-            echo "PM2 env PORT (should match .env):" >&2
+            echo "Listeners on PORT ${PORT}:" >&2
+            ss -tlnp 2>/dev/null | grep ":${PORT} " >&2 || netstat -tlnp 2>/dev/null | grep ":${PORT} " >&2 || true
             pm2 env 0 2>/dev/null | grep -E '^PORT=' >&2 || true
             pm2 logs next --lines 40 --nostream >&2 || true
             exit 1