From 8047b07e4769feae865b77f3bb3e88ee9b984030 Mon Sep 17 00:00:00 2001
From: GambleCodez Affiliates <thetylo88@gmail.com>
Date: Fri, 20 Feb 2026 19:02:54 -0600
Subject: [PATCH] Improve prod-run health output and service/logrotate
 resilience

---
 prod-run.sh | 148 ++++++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 119 insertions(+), 29 deletions(-)

diff --git a/prod-run.sh b/prod-run.sh
index 9cd7f24..6f0f2b7 100755
--- a/prod-run.sh
+++ b/prod-run.sh
@@ -7,9 +7,13 @@ INDEX_FILE="$PROJECT_DIR/index.js"
 ENV_FILE="$PROJECT_DIR/.env"
 LOG_DIR="$PROJECT_DIR/logs"
 BOT_LOG="$LOG_DIR/runewager.log"
+LEGACY_BOT_LOG="$PROJECT_DIR/runewager.log"
+CRASH_LOG="$LOG_DIR/crash.log"
+SYSTEMD_STDOUT_LOG="$LOG_DIR/runewager-1.log"
 SERVICE_FILE="/etc/systemd/system/runewager.service"
 LOGROTATE_FILE="/etc/logrotate.d/runewager"
 CRON_LINE='0 3 * * * /usr/sbin/logrotate -f /etc/logrotate.d/runewager'
+HEALTH_LOG_MAX_AGE_MINUTES=10
 
 say() {
   printf '[runewager] %s\n' "$*"
@@ -31,6 +35,66 @@ ensure_file() {
   [ -f "$1" ] || : > "$1"
 }
 
+ensure_log_paths() {
+  ensure_dir "$LOG_DIR"
+  ensure_file "$BOT_LOG"
+  ensure_file "$CRASH_LOG"
+  ensure_file "$SYSTEMD_STDOUT_LOG"
+
+  if [ ! -f "$LEGACY_BOT_LOG" ]; then
+    : > "$LEGACY_BOT_LOG"
+  fi
+}
+
+
+get_bot_pid() {
+  pgrep -f "node .*index\.js" | head -n 1 || true
+}
+
+log_age_seconds() {
+  local file="$1"
+  if [ ! -f "$file" ]; then
+    echo -1
+    return
+  fi
+  local now mtime
+  now="$(date +%s)"
+  mtime="$(stat -c %Y "$file" 2>/dev/null || echo 0)"
+  echo $((now - mtime))
+}
+
+print_health_summary() {
+  local pid
+  pid="$(get_bot_pid)"
+
+  if [ -n "$pid" ]; then
+    say "✅ Bot is running (pid: $pid)"
+  else
+    err "❌ Bot is not running"
+    return
+  fi
+
+  local active_log="$BOT_LOG"
+  if [ ! -s "$active_log" ] && [ -s "$LEGACY_BOT_LOG" ]; then
+    active_log="$LEGACY_BOT_LOG"
+  fi
+
+  local age
+  age="$(log_age_seconds "$active_log")"
+  if [ -s "$active_log" ] && [ "$age" -ge 0 ] && [ "$age" -le $((HEALTH_LOG_MAX_AGE_MINUTES * 60)) ]; then
+    say "✅ Health check passed: recent logs are present (${age}s old) at $active_log"
+    return
+  fi
+
+  if [ -s "$active_log" ] && [ "$age" -gt $((HEALTH_LOG_MAX_AGE_MINUTES * 60)) ]; then
+    warn "Bot process is running but logs are stale (${age}s old) at $active_log"
+  else
+    warn "Bot process is running but no recent logs were found"
+  fi
+
+  warn "If /start is unresponsive, restart with: pkill -f 'node .*index\.js' && bash $PROJECT_DIR/prod-run.sh"
+}
+
 fix_working_directory() {
   if [ "$(pwd)" != "$PROJECT_DIR" ]; then
     say "Switching working directory to $PROJECT_DIR"
@@ -77,9 +141,14 @@ fix_git_remote_and_branch() {
 }
 
 fix_node_modules() {
-  if [ ! -d "$PROJECT_DIR/node_modules" ] || [ ! -f "$PROJECT_DIR/node_modules/.package-lock.json" ]; then
-    say "Installing dependencies with npm ci --omit=dev"
-    npm ci --omit=dev
+  if [ ! -d "$PROJECT_DIR/node_modules" ]; then
+    if [ -f "$PROJECT_DIR/package-lock.json" ]; then
+      say "Installing dependencies with npm ci --omit=dev"
+      npm ci --omit=dev
+    else
+      say "Installing dependencies with npm install --omit=dev (package-lock.json missing)"
+      npm install --omit=dev
+    fi
   fi
 }
 
@@ -246,9 +315,16 @@ fix_require_and_import_hints() {
 }
 
 ensure_logrotate_config() {
+  local log_owner="runewager"
+  local log_group="runewager"
+  if ! id -u "$log_owner" >/dev/null 2>&1; then
+    log_owner="$(id -un 2>/dev/null || echo root)"
+    log_group="$(id -gn 2>/dev/null || echo root)"
+  fi
+
   local conf
-  conf=$(cat <<'CONF'
-/var/www/html/Runewager/logs/runewager*.log /var/www/html/Runewager/logs/crash.log {
+  conf=$(cat <<CONF
+$LOG_DIR/runewager*.log $LOG_DIR/crash.log {
     daily
     size 5M
     rotate 7
@@ -257,7 +333,7 @@ ensure_logrotate_config() {
     compress
     delaycompress
     copytruncate
-    create 0640 runewager runewager
+    create 0640 $log_owner $log_group
 }
 CONF
 )
@@ -271,9 +347,16 @@ CONF
     fi
   fi
 
+  if [ -f "$LOGROTATE_FILE" ] && ! grep -Fq "$LOG_DIR/runewager*.log" "$LOGROTATE_FILE"; then
+    warn "Replacing $LOGROTATE_FILE to use detected log directory"
+    printf '%s\n' "$conf" > "$LOGROTATE_FILE"
+  fi
+
   if [ -f "$LOGROTATE_FILE" ] && command -v logrotate >/dev/null 2>&1; then
-    if ! logrotate -d "$LOGROTATE_FILE" >/dev/null 2>&1; then
+    local logrotate_output=''
+    if ! logrotate_output="$(logrotate -d "$LOGROTATE_FILE" 2>&1)"; then
       warn "logrotate dry-run failed for $LOGROTATE_FILE"
+      warn "$logrotate_output"
     else
       say "logrotate config validation passed"
     fi
@@ -295,8 +378,16 @@ ensure_cron_fallback() {
 }
 
 ensure_systemd_service() {
+  local service_user="runewager"
+  local service_group="runewager"
+  if ! id -u "$service_user" >/dev/null 2>&1; then
+    service_user="$(id -un 2>/dev/null || echo root)"
+    service_group="$(id -gn 2>/dev/null || echo root)"
+    warn "runewager user missing; using $service_user:$service_group for systemd service"
+  fi
+
   local conf
-  conf=$(cat <<'CONF'
+  conf=$(cat <<CONF
 [Unit]
 Description=Runewager bot
 After=network-online.target
@@ -304,17 +395,17 @@ Wants=network-online.target
 
 [Service]
 Type=simple
-WorkingDirectory=/var/www/html/Runewager
-ExecStart=/var/www/html/Runewager/prod-run.sh
-User=runewager
-Group=runewager
-EnvironmentFile=/var/www/html/Runewager/.env
+WorkingDirectory=$PROJECT_DIR
+ExecStart=$PROJECT_DIR/prod-run.sh
+User=$service_user
+Group=$service_group
+EnvironmentFile=$PROJECT_DIR/.env
 Environment=NODE_ENV=production
 Restart=always
 RestartSec=5
 KillSignal=SIGTERM
 TimeoutStopSec=20
-StandardOutput=append:/var/www/html/Runewager/logs/runewager-1.log
+StandardOutput=append:$SYSTEMD_STDOUT_LOG
 StandardError=inherit
 
 [Install]
@@ -334,7 +425,12 @@ CONF
   fi
 
   local required=(
-    'EnvironmentFile=/var/www/html/Runewager/.env'
+    "WorkingDirectory=$PROJECT_DIR"
+    "ExecStart=$PROJECT_DIR/prod-run.sh"
+    "User=$service_user"
+    "Group=$service_group"
+    "EnvironmentFile=$PROJECT_DIR/.env"
+    "StandardOutput=append:$SYSTEMD_STDOUT_LOG"
     'Environment=NODE_ENV=production'
     'Restart=always'
     'RestartSec=5'
@@ -367,6 +463,9 @@ print_human_errors() {
 
   local blob
   blob="$(tail -n 200 "$BOT_LOG" 2>/dev/null || true)"
+  if [ -z "$blob" ] && [ -f "$LEGACY_BOT_LOG" ]; then
+    blob="$(tail -n 200 "$LEGACY_BOT_LOG" 2>/dev/null || true)"
+  fi
 
   say "Human-readable diagnosis"
   printf '%s\n' "$blob" | grep -qi 'Cannot find module' && say "Root cause: A Node module is missing or a require path is wrong. Run npm ci and verify relative import paths."
@@ -382,21 +481,13 @@ print_human_errors() {
 }
 
 is_bot_running() {
-  pgrep -f "node .*index\.js" >/dev/null 2>&1
+  [ -n "$(get_bot_pid)" ]
 }
 
 check_bot_status_twice() {
-  if is_bot_running; then
-    say "Bot running"
-  else
-    say "Bot NOT running"
-  fi
+  print_health_summary
   sleep 15
-  if is_bot_running; then
-    say "Bot running"
-  else
-    say "Bot NOT running"
-  fi
+  print_health_summary
 }
 
 start_background_if_needed() {
@@ -433,10 +524,9 @@ main() {
     exit 1
   fi
 
-  ensure_dir "$LOG_DIR"
+  ensure_log_paths
   ensure_dir "$PROJECT_DIR/data"
   ensure_dir "$PROJECT_DIR/data/backups"
-  ensure_file "$BOT_LOG"
 
   fix_merge_conflicts
   fix_lockfiles
@@ -453,10 +543,10 @@ main() {
   ensure_cron_fallback
   ensure_systemd_service
 
-  print_human_errors
   check_bot_status_twice
   start_background_if_needed
   check_bot_status_twice
+  print_human_errors
 }
 
 main "$@"