服务器性能优化技巧

# 1. 系统整体：负载、内存、swap
uptime; free -h; vmstat 1 5

# 2. 磁盘 IO：%util 接近 100% 说明磁盘是瓶颈
iostat -x 1 5

# 3. 网络：看 TIME_WAIT、ESTABLISHED 数量与重传率
ss -s; netstat -n | awk '/^tcp/{++S[$NF]} END{for(k in S) print k, S[k]}'

# 4. 文件描述符：当前使用量与上限
cat /proc/sys/fs/file-nr; ulimit -n

# 5. Web 层：QPS 与 P99 耗时
tail -n 10000 /var/log/nginx/access.log | awk '{print $NF}' | sort -n \
  | awk '{a[NR]=$1} END{printf "P50=%.3fs P95=%.3fs P99=%.3fs\n", a[int(NR*0.5)], a[int(NR*0.95)], a[int(NR*0.99)]}'

# 一次性写入所有内核参数（Ubuntu 24.04 / Rocky Linux 9 通用）
sudo tee /etc/sysctl.d/99-performance.conf >/dev/null <<'EOF'
# 网络：连接队列与 TIME_WAIT
net.core.somaxconn = 65535
net.core.netdev_max_backlog = 250000
net.ipv4.tcp_max_syn_backlog = 65535
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_fin_timeout = 15
net.ipv4.ip_local_port_range = 1024 65535
net.ipv4.tcp_keepalive_time = 600
net.ipv4.tcp_max_tw_buckets = 262144

# 文件系统：描述符上限
fs.file-max = 2000000
fs.nr_open = 2000000

# 内存：少用 swap，控制脏页刷盘
vm.swappiness = 10
vm.dirty_ratio = 20
vm.dirty_background_ratio = 5

# Redis 服务器专用（非 Redis 机器请注释掉）
vm.overcommit_memory = 1
EOF

# 立即生效，并验证
sudo sysctl --system
sysctl net.core.somaxconn fs.file-max vm.swappiness

# 查看当前限制
ulimit -n          # soft 限制
ulimit -Hn         # hard 限制

# 永久生效：注意 fs.nr_open 必须 ≥ 这里设置的 hard 值，否则会导致无法登录
sudo tee /etc/security/limits.d/99-performance.conf >/dev/null <<'EOF'
*       soft    nofile    100000
*       hard    nofile    100000
*       soft    nproc     65535
*       hard    nproc     65535
root    soft    nofile    100000
root    hard    nofile    100000
EOF

# systemd 管理的服务不读 limits.conf，必须单独改服务的 LimitNOFILE
sudo systemctl edit nginx
# 在打开的编辑器中写入：
# [Service]
# LimitNOFILE=100000
sudo systemctl daemon-reload && sudo systemctl restart nginx

# 验证某个进程实际生效的限制（把 PID 换成实际进程号）
cat /proc/$(pidof nginx | awk '{print $1}')/limits | grep "open files"

# /etc/nginx/nginx.conf 核心优化片段
user  nginx;
worker_processes auto;
worker_rlimit_nofile 100000;
error_log /var/log/nginx/error.log warn;
pid /run/nginx.pid;

events {
    use epoll;
    worker_connections 65535;
    multi_accept on;
}

http {
    include       /etc/nginx/mime.types;
    default_type  application/octet-stream;

    # 日志格式增加 $request_time 与 $upstream_response_time，便于算 P99
    log_format main '$remote_addr - $remote_user [$time_local] "$request" '
                    '$status $body_bytes_sent "$http_referer" '
                    '"$http_user_agent" rt=$request_time urt=$upstream_response_time';
    access_log /var/log/nginx/access.log main buffer=32k flush=5s;

    # 传输优化：零拷贝 + 长连接复用
    sendfile        on;
    tcp_nopush      on;
    tcp_nodelay     on;
    keepalive_timeout  30;
    keepalive_requests 10000;

    # 压缩：只压文本类，图片视频本身已压缩不要再压
    gzip on;
    gzip_comp_level 5;
    gzip_min_length 1024;
    gzip_buffers 16 8k;
    gzip_types text/plain text/css application/json application/javascript
               application/xml text/xml image/svg+xml;
    gzip_vary on;
    gzip_disable "msie6";

    # 文件句柄缓存，静态资源多的站点收益明显
    open_file_cache          max=100000 inactive=60s;
    open_file_cache_valid    60s;
    open_file_cache_min_uses 2;
    open_file_cache_errors   on;

    # 客户端限制
    client_max_body_size     64m;
    client_body_buffer_size  256k;
    client_header_timeout    15s;
    client_body_timeout      30s;
    send_timeout             30s;
    reset_timedout_connection on;

    include /etc/nginx/conf.d/*.conf;
}

# 改完务必校验并热加载，reload 不会断连接
sudo nginx -t && sudo systemctl reload nginx

# 验证 worker 进程数与连接上限是否生效
ps -ef | grep -c "[n]ginx: worker"
curl -s http://127.0.0.1/nginx-status

# /etc/mysql/mysql.conf.d/mysqld.cnf（MySQL 8.4 / MariaDB 11.4 适用）
[mysqld]
# ---- 内存：8 核 32G 独享数据库示例 ----
innodb_buffer_pool_size        = 20G
innodb_buffer_pool_instances   = 8
innodb_log_file_size           = 2G
innodb_log_buffer_size         = 64M

# ---- IO 与持久化 ----
innodb_flush_method            = O_DIRECT
innodb_io_capacity             = 5000
innodb_io_capacity_max         = 10000
innodb_flush_log_at_trx_commit = 2
innodb_flush_neighbors         = 0          # SSD 必设为 0

# ---- 连接 ----
max_connections                = 1000
thread_cache_size              = 100
table_open_cache               = 4000
wait_timeout                   = 300
interactive_timeout            = 300

# ---- 临时表与排序 ----
tmp_table_size                 = 256M
max_heap_table_size            = 256M
sort_buffer_size               = 4M
join_buffer_size               = 4M
read_buffer_size               = 2M

# ---- 慢查询（必开） ----
slow_query_log                 = ON
slow_query_log_file            = /var/log/mysql/slow.log
long_query_time                = 1
log_queries_not_using_indexes  = OFF
min_examined_row_limit         = 100

character_set_server           = utf8mb4
collation_server               = utf8mb4_0900_ai_ci

# 应用配置并验证
sudo systemctl restart mysql
mysql -u root -p -e "SHOW VARIABLES LIKE 'innodb_buffer_pool_size'; SHOW VARIABLES LIKE 'max_connections';"

# 关键验证：InnoDB 缓冲池命中率应 > 99%
mysql -u root -p -e "
SELECT ROUND((1 - (SELECT VARIABLE_VALUE FROM performance_schema.global_status
                   WHERE VARIABLE_NAME='Innodb_buffer_pool_reads')
             / (SELECT VARIABLE_VALUE FROM performance_schema.global_status
                WHERE VARIABLE_NAME='Innodb_buffer_pool_read_requests')) * 100, 2) AS hit_rate_percent;"

# 用 mysqldumpslow 找出最该优化的慢查询
mysqldumpslow -s t -t 10 /var/log/mysql/slow.log

# 第一步：测出单个 PHP-FPM 子进程平均占用内存（单位 MB）
ps -eo size,rss,comm --sort -rss | grep php-fpm | awk '{s+=$2; n++} END{printf "总内存 %.0f MB, 进程数 %d, 平均 %.1f MB\n", s/1024, n, s/n/1024}'

# 第二步：套用公式（示例：16G 内存，系统与其它服务占 4G，单进程 40MB）
#   max_children = (16*1024 - 4*1024) / 40 ≈ 307，取保守值 250

# /etc/php/8.3/fpm/pool.d/www.conf
[www]
user = www-data
group = www-data

# 用 dynamic 动态模式（小内存机器用 ondemand）
pm = dynamic
pm.max_children      = 250     ; 按上面公式算出，硬上限
pm.start_servers     = 20      ; 启动时进程数，建议 = CPU 核数 × 4
pm.min_spare_servers = 10      ; 最小空闲进程
pm.max_spare_servers = 40      ; 最大空闲进程，超过会被回收
pm.max_requests      = 2000    ; 单进程处理多少请求后重启，防内存泄漏

; 慢日志：超过 5 秒的脚本记录下来，是定位性能问题的利器
slowlog = /var/log/php-fpm/www-slow.log
request_slowlog_timeout = 5s

; 状态页，配合 Nginx 查看实时排队情况
pm.status_path = /fpm-status

; 防超大脚本
php_admin_value[memory_limit] = 256M

# 重启并验证
sudo php-fpm8.3 -t && sudo systemctl restart php8.3-fpm

# 观察是否出现排队（listen queue > 0 说明 max_children 不够）
curl -s http://127.0.0.1/fpm-status
# 关键字段：listen queue（排队数）、active processes（活跃进程）、max children reached（达到上限次数）

# /etc/redis/redis.conf 关键片段
maxmemory 8gb
maxmemory-policy allkeys-lru
maxmemory-samples 5

# 持久化：纯缓存可关闭，关键数据开启 AOF
save 900 1
appendonly no

# 网络
tcp-backlog 511
timeout 0
tcp-keepalive 300

# 异步释放，避免删大 key 时卡住
lazyfree-lazy-eviction yes
lazyfree-lazy-expire yes
lazyfree-lazy-server-del yes
replica-lazy-flush yes

# 危险命令重命名，防止误操作
rename-command FLUSHALL ""
rename-command FLUSHDB  ""
rename-command KEYS     ""

# 应用并验证
sudo systemctl restart redis-server
redis-cli -a 'RedisPass2026' INFO memory | grep -E "used_memory_human|maxmemory_human|maxmemory_policy"
redis-cli -a 'RedisPass2026' INFO stats  | grep -E "keyspace_hits|keyspace_misses"

# 计算命中率：应 > 95%，低于 90% 说明缓存设计有问题
# 命中率 = hits / (hits + misses)