使用 nginx+gunicorn+django 的间歇性 502s

Posted

技术标签:

【中文标题】使用 nginx+gunicorn+django 的间歇性 502s【英文标题】:Intermittent 502s with nginx+gunicorn+django 【发布时间】:2012-08-20 06:49:45 【问题描述】:

在过去的几周里,我们收到了越来越多的 502 错误。目前我们的堆栈是 nginx + gunicron + django 在一个 m1.large EC2 实例上,由一个小型 RDS 实例支持。

随着请求负载的增加,它们似乎变得更加频繁。我会在使用浏览器时看到随机的 502,但是我们命中 api (Tasty Pie) 的命令行脚本通常会在第二次或第三次请求时失败。但是,如果我在脚本发出请求之前将睡眠功能添加到脚本中,那么该请求就可以了,但接下来会出现 502。请注意,我们将摘要身份验证与请求库和睡眠包装器一起使用——因此是 401、200 模式。

为了使调试更加棘手,当 Gunicorn 使用 --debug 选项运行时,问题会自行解决。如果我删除 --debug 选项但将我的 Gunicorn 工作人员明确限制为 1,则错误仍然存​​在。

我的 nginx.conf:

user www-data;
worker_processes 4;
pid /var/run/nginx.pid;

events 
    worker_connections 768;
    # multi_accept on;


http 

    ##
    # Basic Settings
    ##

    sendfile on;
    tcp_nopush on;
    tcp_nodelay on;
    keepalive_timeout 65;
    types_hash_max_size 2048;
    # server_tokens off;

    # server_names_hash_bucket_size 64;
    # server_name_in_redirect off;

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

    ##
    # Logging Settings
    ##

    access_log /var/log/nginx/access.log;
    error_log /var/log/nginx/error.log;

    ##
    # Gzip Settings
    ##

    gzip on;
    gzip_disable "msie6";

        gzip_proxied any;
        gzip_types application/x-ghi-packedschemafeatures-v1
        gzip_http_version 1.1;
        gzip_comp_level 1;
        gzip_min_length 500;

        proxy_buffering on;
        proxy_http_version 1.1;

    ##
    # Virtual Host Configs
    ##

    include /etc/nginx/conf.d/*.conf;
    include /etc/nginx/sites-enabled/*;

虚拟主机文件:

    server 
      listen 80;
      server_name pipeline.ourdomain.com;
      location / 
        rewrite ^ https://$server_name$request_uri permanent;
      


server 
      listen 443;
      server_name pipeline.ourdomain.com;
      ssl on;
      ssl_protocols SSLv3 TLSv1;
      ssl_ciphers ALL:-ADH:+HIGH:+MEDIUM:-LOW:-SSLv2:-EXP;
      ssl_session_cache shared:SSL:10m;
      ssl_certificate     /etc/ssl/certs/ourdomain.com.combined.crt;
      ssl_certificate_key /etc/ssl/private/ourdomain.com.key;

      root /var/www/;
      location /static/ 
               alias /var/www/production/pipeline/public/;
      

      location / 
             proxy_pass_header Server;
             proxy_set_header Host $http_host;
             proxy_redirect off;
             proxy_set_header X-Real-IP $remote_addr;
             proxy_set_header X-Scheme $scheme;
             proxy_set_header X-Forwarded-Protocol https;
             proxy_connect_timeout 240;
             proxy_read_timeout 280;
             proxy_pass http://localhost:8000/;
      
      error_page 500 502 503 504 /static/50x.html;




独角兽指挥部

#!/bin/bash
set -e
LOGFILE=/var/log/gunicorn/ea_pipeline.log
LOGDIR=$(dirname $LOGFILE)

SETTINGS=production_settings
# user/group to run as
USER=ubuntu
GROUP=ubuntu
DJANGO_PATH=$(dirname $(readlink -f $0))/../
cd $DJANGO_PATH
echo $(pwd)
. ../env/bin/activate
test -d $LOGDIR || mkdir -p $LOGDIR
exec ../env/bin/gunicorn_django \
     --user=$USER --group=$GROUP --log-level=debug \
     --preload \
     --workers=4 \
     --timeout=90 \
     --settings=$SETTINGS \
     --limit-request-line=8190 \
     --limit-request-field_size 0 \
     --pythonpath=$DJANGO_PATH \
     --log-file=$LOGFILE production_settings.py 2>>$LOGFILE

访问日志示例:

67.134.170.194 - - [24/Aug/2012:00:28:17 +0000] "GET /api/v1/storage/ HTTP/1.1" 401 5 "-" "python-requests/0.13.8 CPython/2.7.3 Linux/3.2.0-29-generic"
67.134.170.194 - - [24/Aug/2012:00:28:18 +0000] "GET /api/v1/storage/ HTTP/1.1" 200 326 "-" "python-requests/0.13.8 CPython/2.7.3 Linux/3.2.0-29-generic"
67.134.170.194 - - [24/Aug/2012:00:28:18 +0000] "GET /api/v1/customer/?client_id=lamb_01 HTTP/1.1" 502 18 "-" "python-requests/0.13.8 CPython/2.7.3 Linux/3.2.0-29-generic"
67.134.170.194 - - [24/Aug/2012:00:29:41 +0000] "GET /api/v1/storage/ HTTP/1.1" 502 18 "-" "python-requests/0.13.8 CPython/2.7.3 Linux/3.2.0-29-generic"

Nginx 错误日志:

2012/08/24 00:28:18 [error] 16490#0: *3 connect() failed (111: Connection refused) while connecting to upstream, client: 67.134.170.194, server: pipeline.ourdomain.com, request: "GET /api/v1/customer/?client_id=lamb_01 HTTP/1.1", upstream: "http://127.0.0.1:8000/api/v1/customer/?client_id=lamb_01", host: "pipeline.ourdomain.com"
2012/08/24 00:29:41 [error] 16490#0: *7 connect() failed (111: Connection refused) while connecting to upstream, client: 67.134.170.194, server: pipeline.ourdomain.com, request: "GET /api/v1/storage/ HTTP/1.1", upstream: "http://127.0.0.1:8000/api/v1/storage/", host: "pipeline.ourdomain.com"

Gunicorn 日志样本:

2012-08-24 17:03:13 [8716] [INFO] Starting gunicorn 0.14.3
2012-08-24 17:03:13 [8716] [DEBUG] Arbiter booted
2012-08-24 17:03:13 [8716] [INFO] Listening at: http://127.0.0.1:8000 (8716)
2012-08-24 17:03:13 [8716] [INFO] Using worker: sync
2012-08-24 17:03:13 [8735] [INFO] Booting worker with pid: 8735
2012-08-24 17:03:13 [8736] [INFO] Booting worker with pid: 8736
2012-08-24 17:03:13 [8737] [INFO] Booting worker with pid: 8737
2012-08-24 17:03:13 [8738] [INFO] Booting worker with pid: 8738
2012-08-24 17:03:21 [8738] [DEBUG] GET /api/v1/storage/
Assertion failed: ok (mailbox.cpp:84)
2012-08-24 17:03:21 [8738] [INFO] Parent changed, shutting down: <Worker 8738>
2012-08-24 17:03:21 [8738] [INFO] Worker exiting (pid: 8738)
Error in sys.exitfunc:
2012-08-24 17:03:21 [8737] [DEBUG] GET /api/v1/storage/
2012-08-24 17:03:22 [8838] [INFO] Starting gunicorn 0.14.3
2012-08-24 17:03:22 [8838] [ERROR] Connection in use: ('127.0.0.1', 8000)
2012-08-24 17:03:22 [8838] [ERROR] Retrying in 1 second.
2012-08-24 17:03:22 [8737] [INFO] Parent changed, shutting down: <Worker 8737>
2012-08-24 17:03:22 [8737] [INFO] Worker exiting (pid: 8737)
Error in sys.exitfunc:
2012-08-24 17:03:22 [8736] [DEBUG] GET /api/v1/customer/
2012-08-24 17:03:23 [8736] [INFO] Parent changed, shutting down: <Worker 8736>
2012-08-24 17:03:23 [8736] [INFO] Worker exiting (pid: 8736)
Error in sys.exitfunc:
2012-08-24 17:03:23 [8838] [ERROR] Connection in use: ('127.0.0.1', 8000)
2012-08-24 17:03:23 [8838] [ERROR] Retrying in 1 second.
2012-08-24 17:03:24 [8735] [DEBUG] GET /api/v1/upload_action/
2012-08-24 17:03:24 [8838] [ERROR] Connection in use: ('127.0.0.1', 8000)
2012-08-24 17:03:24 [8838] [ERROR] Retrying in 1 second.
2012-08-24 17:03:24 [8735] [INFO] Parent changed, shutting down: <Worker 8735>
2012-08-24 17:03:24 [8735] [INFO] Worker exiting (pid: 8735)
Error in sys.exitfunc:
2012-08-24 17:03:25 [8838] [DEBUG] Arbiter booted
2012-08-24 17:03:25 [8838] [INFO] Listening at: http://127.0.0.1:8000 (8838)
2012-08-24 17:03:25 [8838] [INFO] Using worker: sync
2012-08-24 17:03:25 [8907] [INFO] Booting worker with pid: 8907
2012-08-24 17:03:25 [8908] [INFO] Booting worker with pid: 8908
2012-08-24 17:03:25 [8909] [INFO] Booting worker with pid: 8909
2012-08-24 17:03:25 [8910] [INFO] Booting worker with pid: 8910

【问题讨论】:

@VBart Nginx 正在记录到 access.log 但在您的配置中:error_log /var/log/nginx/error.log;。而且,nginx错误日志的格式不一样,不是你这里展示的那种。 @VBart 是的。在上面添加。 好的,现在很清楚 nginx 无法连接到 gunicorn。 gunicorn 日志中的这些错误似乎是罪魁祸首。也许您应该尝试向 gunicorn 开发人员报告错误。 @VBart 感谢您的信息。我不知道是 gunicorn 还是 nginx 有问题。我正在设置 uWSGI 并且情况似乎更好。 【参考方案1】:

这是一个非常古老的帖子。但是我在 NGinx+Gunicorn+Flask 设置中遇到了完全相同的问题。在每 300 个左右的请求中,我也会遇到与您相同的日志的 502 错误。将 gunicorn 工作类型更改为异步为我解决了这个问题(我选择了 gthread)。希望这个答案对某人有所帮助。

如何更改设置:http://docs.gunicorn.org/en/stable/settings.html#worker-class

如何选择您的工人类型: http://docs.gunicorn.org/en/latest/design.html#choosing-a-worker-type

这里有一个很好的解释: How many concurrent requests does a single Flask process receive?

【讨论】:

以上是关于使用 nginx+gunicorn+django 的间歇性 502s的主要内容,如果未能解决你的问题,请参考以下文章

使用 Django、Nginx、Gunicorn 和主管进行空日志记录

django-gunicorn-nginx:502 网关错误

Django 不使用 NGINX + GUNICORN 提供静态文件

使用 nginx 和 gunicorn 运行多个 django 项目

如何使用 gunicorn 为 django 配置 nginx?

使用Nginx + Gunicorn + Django 方式部署django程序