#!/bin/bash
# 服务器卡顿定位 — 一次跑完，输出「卡在哪一层」结论
# 用法: bash deploy/profile-server-lag.sh
# 只读诊断，不改任何配置

set -e
DJANGO_DIR="/opt/1panel/apps/openresty/nginx/www/sites/43.142.166.152.443/django"
NGINX_LOG="/opt/1panel/apps/openresty/nginx/log/access.log"
cd "$DJANGO_DIR"

echo "=============================================="
echo "  星阙服务器卡顿定位 — $(date '+%Y-%m-%d %H:%M:%S')"
echo "=============================================="

# ---------- A. 进程层：Gunicorn 是否在排队 ----------
echo ""
echo ">>> [A] Gunicorn 进程 / 端口"
systemctl is-active gunicorn-dianjing.service 2>/dev/null || echo "systemd: inactive"
GC=$(pgrep -cf 'gunicorn.*a_long_dianjing' 2>/dev/null || echo 0)
echo "gunicorn 进程数: $GC (正常=5: 1master+4worker)"
ss -ltn 2>/dev/null | grep ':8001' || netstat -ltn 2>/dev/null | grep ':8001' || echo "8001 未监听"

echo ""
echo ">>> [A2] 本机直连 Gunicorn（绕过 Nginx）"
for path in "/" "/hqhd/"; do
  T=$(curl -s -o /dev/null -w '%{time_total}' --connect-timeout 3 "http://127.0.0.1:8001${path}" 2>/dev/null || echo "FAIL")
  echo "  127.0.0.1:8001${path} -> ${T}s"
done

# ---------- B. 数据库层：握手 + 单条 SQL ----------
echo ""
echo ">>> [B] MySQL 延迟（从本机 Django 目录测）"
python3 - <<'PY'
import time, sys
try:
    import app_secrets
    import pymysql
except Exception as e:
    print(f"  无法导入: {e}")
    sys.exit(0)

host, port = app_secrets.DATABASE_HOST, int(app_secrets.DATABASE_PORT)
user, pwd, db = app_secrets.DATABASE_USER, app_secrets.DATABASE_PASSWORD, app_secrets.DATABASE_NAME
print(f"  当前配置: {host}:{port} / {db}")

def bench(label, h, p):
    try:
        t0 = time.perf_counter()
        conn = pymysql.connect(host=h, port=p, user=user, password=pwd, database=db, connect_timeout=5)
        t_conn = (time.perf_counter() - t0) * 1000
        t1 = time.perf_counter()
        with conn.cursor() as c:
            c.execute("SELECT COUNT(*) FROM information_schema.tables WHERE table_schema=%s", (db,))
            c.fetchone()
        t_q = (time.perf_counter() - t1) * 1000
        conn.close()
        print(f"  [{label}] {h}:{p}  新建连接={t_conn:.0f}ms  单条SQL={t_q:.0f}ms")
        return t_conn, t_q
    except Exception as e:
        print(f"  [{label}] {h}:{p}  失败: {e}")
        return None, None

cur_c, cur_q = bench("生产库", host, port)
loc_c, loc_q = bench("本机Docker", "172.17.0.15", 3306)

print("")
if cur_q and cur_q > 30:
    print("  >>> 结论B: 数据库在远程，每条SQL ~50ms+，接口跑10+条SQL就会慢200ms+")
if loc_q and loc_q < 10 and cur_q and cur_q > 30:
    print("  >>> 修复: app_secrets.py 改 DATABASE_HOST=172.17.0.15 PORT=3306 后 restart gunicorn")
PY

# ---------- C. Django 层：模拟 dddhq 打了多少条 SQL ----------
echo ""
echo ">>> [C] /dingdan/ddhq 一次请求打几条 SQL、耗时多少"
python3 - <<'PY'
import os, sys, time
os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'a_long_dianjing.settings')
sys.path.insert(0, '.')

try:
    import django
    django.setup()
    from django.db import connection, reset_queries
    from django.conf import settings
    settings.DEBUG = True  # 仅本脚本进程内开启 query log

    reset_queries()
    t0 = time.perf_counter()

    # 不跑完整 view（要 token），只测典型 ORM 链：订单列表 count + page
    from orders.models import Order
    from django.db.models import Q
    qs = Order.query.filter(Q(Status__in=[1, 7])).order_by('-CreateTime')
    total = qs.count()
    page = list(qs[:5].values('OrderID', 'Status', 'CreateTime'))

    elapsed = (time.perf_counter() - t0) * 1000
    n = len(connection.queries)
    sql_ms = sum(float(q.get('time', 0)) for q in connection.queries) * 1000

    print(f"  仅订单 count+取5条: SQL条数={n}  Django记录SQL耗时={sql_ms:.0f}ms  总耗时={elapsed:.0f}ms  订单总数={total}")
    print(f"  （完整 /ddhq 还有会员/标签/商家等查询，实际 SQL 通常 15~30 条）")
    if n > 0:
        print("  最慢的3条SQL:")
        for q in sorted(connection.queries, key=lambda x: float(x.get('time', 0)), reverse=True)[:3]:
            sql = q['sql'][:120].replace('\n', ' ')
            print(f"    {float(q['time'])*1000:.0f}ms  {sql}...")
except Exception as e:
    print(f"  Django 测速失败: {e}")
    import traceback; traceback.print_exc()
PY

# ---------- D. Nginx 层：最近真实接口 rt= 分布 ----------
echo ""
echo ">>> [D] Nginx 最近 hqhd 接口 rt=（真实用户体感）"
if [ -f "$NGINX_LOG" ]; then
  echo "  最近20条 hqhd:"
  grep 'hqhd' "$NGINX_LOG" | tail -20 | while read -r line; do
    RT=$(echo "$line" | grep -oP 'rt=\K[0-9.]+' || echo "?")
    URI=$(echo "$line" | grep -oP '"(GET|POST) \K[^ ]+' || echo "?")
    echo "    rt=${RT}s  $URI"
  done
  echo ""
  echo "  各接口 rt 中位数（最近500条 hqhd）:"
  grep 'hqhd' "$NGINX_LOG" | tail -500 | grep -oP 'rt=\K[0-9.]+' | sort -n | awk '
    { a[NR]=$1; sum+=$1 }
    END {
      if(NR==0){ print "    无数据"; exit }
      mid=(NR%2)?a[(NR+1)/2]:(a[NR/2]+a[NR/2+1])/2
      printf "    样本=%d  中位rt=%.3fs  平均rt=%.3fs  最大rt=%.3fs\n", NR, mid, sum/NR, a[NR]
    }'
else
  echo "  找不到 $NGINX_LOG"
fi

# ---------- E. 系统层：CPU/内存/磁盘/journal ----------
echo ""
echo ">>> [E] 系统资源"
echo "  load: $(cat /proc/loadavg 2>/dev/null || uptime)"
free -h 2>/dev/null | head -2 || true
journalctl --disk-usage 2>/dev/null | head -1 || true

# ---------- F. 并发层：谁在狂刷 ----------
echo ""
echo ">>> [F] 最近1分钟请求最多的 IP+接口（是否有人刷接口拖慢）"
if [ -f "$NGINX_LOG" ]; then
  tail -3000 "$NGINX_LOG" | grep 'hqhd' | awk '{
    for(i=1;i<=NF;i++) if($i ~ /^[0-9]+\.[0-9]+\.[0-9]+\.[0-9]+$/) ip=$i
    if(match($0, /"(GET|POST) ([^ ]+)/, m)) uri=m[2]
    if(ip && uri) c[ip" "uri]++
  } END {
    n=0
    for(k in c) arr[n++]=c[k]" "k
    asort(arr)
    for(i=n;i>0 && i>n-10;i--) print "  "arr[i]
  }' 2>/dev/null || echo "  (需 gawk 支持 match 第三参数，可忽略)"
fi

echo ""
echo "=============================================="
echo "  怎么看结果："
echo "  - A 里 curl >1s 且 B 里 SQL>30ms → 卡在数据库远程连接"
echo "  - A 里 curl 快但 D 里 rt 慢 → 卡在 Nginx/上游配置或并发"
echo "  - C 里 SQL条数>20 → 卡在接口 SQL 太多（后端代码）"
echo "  - E 里 journal 很大/load 很高 → 卡在系统资源"
echo "  - F 里同一 IP 每秒同一接口 → 卡在被人刷接口"
echo "=============================================="