基于Raindrop网,分享项目的组网架构和部署。
项目组网架构项目访问分为两个流,通过nginx分两个端口暴露给外部使用:数据流:
用户访问Raindrop网站。控制流:
管理人员通过supervisor监控、管理服务器进程。
图中除了将程序部署在ECS上外,还使用了OSS(对象存储服务,可以理解成一个nosql数据库),主要是为了存放一些静态文件,提高访问速度。阿里的OSS还可以与CDN一起使用,同样可以提高访问速度。
通过nginx对外暴露两个端口,如上所述,80端口供用户访问网站,另一个端口供管理人员使用。80端口:
根据请求的url配置了方向代理,分别导向client(angular)和server(flask).
其中server通过gunicorn部署在[socket]localhost:10000
上
配置如下:
server {
listen 80 default_server;
# set client body size to 4M (add by dh) #
client_max_body_size 4M;
gzip_types text/plain text/css application/json application/Javascript text/xml application/xml application/xml+rss text/Javascript;
# product
root /home/raindrop/www/client/dist;
# Add index.php to the list if you are using PHP
index index.html index.htm index.nginx-debian.html;
server_name _;
location / {
# First attempt to serve request as file, then
# as directory, then fall back to displaying a 404.
try_files $uri $uri/ =404;
} # access flask static folder
location /static/ {
# product
root /home/raindrop/www/server/app;
}
location /api/ {
proxy_pass http://localhost:10000/api/;
proxy_redirect off;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}
# Error pages
error_page 413 @413_json;
location @413_json {
default_type application/json;
return 200 '{"msg": "Request Entity Too Large(max=4M)"}';
}
}
gunicorn作为wsgi容器,用来执行flask server。
gunicorn可以使用异步socket:gevent,其本质是基于greenlet实现协程的第三方库,改善io阻塞问题,通过简单的配置就能使程序获得极高的并发处理能力。
注意:使用gunicorn != 使用gevent,如要开启gevent socket,启动gunicorn时需要增加–work-class参数,如下:
gunicorn –workers=4 –worker-class socketio.sgunicorn.GeventSocketIOWorker -b localhost:10000 wsgi:app
除了gunicorn,也可以选用uwsgi,但是有两点限制需要注意:
1.如果使用了Flask-socketio,请不要使用uwsgi,原因
Note regarding uWSGI: While this server has support for gevent and WebSocket,
there is no way to use the custom event loop needed by gevent-socketio,
so there is no directly available method for hosting Flask-SocketIO applications on it.
If you figure out how to do this please let me know!
2.如果使用异步WSGI Server,请勿使用uwsgi,原因:慎用异步 WSGI Server 运行 Flask 应用
Flask(Werkzeug)的Local Thread无法与uwsgi基于uGreen的微线程兼容,引起Local Thread工作混乱。
在程序运行过程中会有一些比较耗时但并非紧急的工作,这些任务可以使用异步任务来处理,提高server的响应速度。
celery – Distributed Task Queue是一个第三方库,提供异步队列和任务处理功能。
Raindrop使用celery配合实现feed功能。通过flask进行配置,使用redis作为异步队列来存储任务,并将处理结果(feed activity)存储在redis中。
如上所述,我们需要在服务器上运行gunicorn和celery两个进程。
显然,我们需要一个monitor来帮我们管理这两个进程,避免进程崩溃不能及时拉起,阻塞业务。
supervisor:Supervisor process control system for UNIX是一个开源monitor程序,并且内置了web管理功能,可以远程监控、重启进程。配置如下:
[inet_http_server]
# web 管理端口,通过nginx暴露给远端用户
port=127.0.0.1:51000
[program:raindrop]
# 程序启动命令
command = gunicorn --workers=4 --worker-class socketio.sgunicorn.GeventSocketIOWorker -b localhost:10000 wsgi:app
directory = /home/raindrop/www/server
user = dh
stopwaitsecs=60
stdout_logfile = /var/log/raindrop/supervisor-raindrop.log
redirect_stderr = true
[program:celery]
command = celery -P gevent -A wsgi.celery worker
directory = /home/raindrop/www/server
user = dh
stopwaitsecs=60
stdout_logfile = /var/log/raindrop/supervisor-celery.log
redirect_stderr = true
mysql:
网站主要数据存储在关系型数据库中redis:
缓存mysql数据 + celery异步队列
使用阿里云的服务器ECS:
部署服务端程序OSS:
存储前端静态文件(速度很快,对前端体验改善很大,采用angular框架强烈推荐使用)
nginx:
HTTP Serverangularjs:
client框架flask:
server框架gunicorn:
web 容器celery:
异步任务处理supervisor:
monitorredis:
数据缓存 + 任务队列mysql:
数据库
ubuntu12.04 64:
操作系统virtualenv
: python虚拟环境(隔离项目开发环境,不用担心包冲突了)vagrant
: 基于virtualbox的本地虚拟环境(环境可以导入导出,团队开发必备)gulp
: angular工程打包工具pip
: python包管理工具fabric
: 基于python的远程脚本(自动化部署神器)
基于Ubuntu 12.0.4,其它系统安装命令(apt-get 等)请自行修改。
如果是第一次部署,需要初始化ECS服务器,安装基本工具:nginx, supervisor, redis, mysql, pip, virtualenv
打包项目代码
发布静态文件到OSS服务器
发布项目代码到ECS服务器,安装server依赖的包
修改mysql, nginx, supervisor
配置文件
拉起所需进程
然后各种apt-get install, scp, tar, cp, mv
,不得不说,这是一个烦人且毫无技术含量的工作,干过几次后基本就可以摔键盘了。
不过,有繁琐的地方,一定有自动化。其实完成上面这些工作,三条命令足以:
fab init_env
fab build
fab deploy
这要感谢Fabric:Simple, Pythonic remote execution and deployment项目,封装了非常简洁的远程操作命令。
使用Fabric,只需要编写一个fabile.py脚本,在启动定义init_env, build, deploy
三个任务:
# -*- coding: utf-8 -*-
import os, re, hashlib
from termcolor import colored
from datetime import datetime
from fabric.api import *
from fabric.contrib.files import exists
class FabricException(Exception):
pass
env.abort_exception = FabricException
# 服务器地址,可以有多个,依次部署:
env.hosts = [
'user@120.1.1.1'
]
env.passwords = {
'user@120.1.1.1:22':'123456'
}
# sudo用户为root:
env.sudo_user = 'root'
# mysql
db_user = 'root'
db_password = '123456'
_TAR_FILE = 'raindrop.tar.gz'
_REMOTE_TMP_DIR = '/tmp'
_REMOTE_BASE_DIR = '/home/raindrop'
_ALIYUN_OSS = {
'endpoint' : 'oss-cn-qingdao.aliyuncs.com',
'bucket' : 'yourbucketname',
'accessKeyId' : 'youraccessKeyId' ,
'accessKeySecret': 'youraccessKeySecret'
}
def build():
'''
必须先打包编译client,再打包整个项目
'''
with lcd(os.path.join(os.path.abspath('.'), 'client')):
local('gulp build') # 上传静态文件到oss服务器,并修改index.html中对静态文件的引用
with lcd(os.path.join(os.path.abspath('.'), 'client/dist')):
with lcd('scripts'):
for file in _list_dir('./'):
if oss_put_object_from_file(file, local('pwd', capture=True) + '/' + file):
_cdnify('../index.html', file)
with lcd('styles'):
for file in _list_dir('./'):
if oss_put_object_from_file(file, local('pwd', capture=True) + '/' + file):
_cdnify('../index.html', file)
# 注意在oss上配置跨域规则,否则fonts文件无法加载
# !!修改fonts文件夹请放开此段程序!!
# with lcd('fonts'):
# for file in _list_dir('./'):
# oss_put_object_from_file('fonts/%s' % file, local('pwd', capture=True) + '/' + file)
with lcd(os.path.join(os.path.abspath('.'), 'server')):
local('pip freeze > requirements/common.txt')
excludes = ['oss', 'distribute']
[local('sed -i -r -e \'/^.*' + exclude + '.*$/d\' "requirements/common.txt"') for exclude in excludes]
local('python setup.py sdist')
# e.g command: fab deploy:'',Fasle
# 注意命令两个参数间不要加空格
def deploy(archive='', needPut='True'):
if archive is '':
filename = '%s.tar.gz' % local('python setup.py --fullname', capture=True).strip()
archive = 'dist/%s' % filename
else:
filename = archive.split('/')[-1] tmp_tar = '%s/%s' % (_REMOTE_TMP_DIR, filename)
if eval(needPut):
# 删除已有的tar文件:
run('rm -f %s' % tmp_tar)
# 上传新的tar文件:
put(archive, _REMOTE_TMP_DIR)
# 创建新目录:
newdir = 'raindrop-%s' % datetime.now().strftime('%y-%m-%d_%H.%M.%S')
with cd(_REMOTE_BASE_DIR):
sudo('mkdir %s' % newdir)
# 重置项目软链接:
with cd(_REMOTE_BASE_DIR):
# 解压到新目录:
with cd(newdir):
sudo('tar -xzvf %s --strip-compOnents=1' % tmp_tar)
# 保存上传文件
if exists('www/server/app/static/upload/images/', use_sudo=True):
sudo('cp www/server/app/static/upload/images/ %s/server/app/static/upload/ -r' % newdir)
sudo('rm -f www')
sudo('ln -s %s www' % newdir)
with cd(_REMOTE_BASE_DIR):
with prefix('source %s/env/local/bin/activate' % _REMOTE_BASE_DIR):
sudo('pip install -r www/server/requirements/common.txt')
# 启动服务
with cd('www'):
# mysql
sudo('cp etc/my.cnf /etc/mysql/')
sudo('restart mysql')
# monitor
sudo('cp etc/rd_super.conf /etc/supervisor/conf.d/')
sudo('supervisorctl stop celery')
sudo('supervisorctl stop raindrop')
sudo('supervisorctl reload')
sudo('supervisorctl start celery')
sudo('supervisorctl start raindrop')
# nginx
sudo('cp etc/rd_nginx.conf /etc/nginx/sites-available/')
# ln -f —-如果要建立的链接名已经存在,则删除之
sudo('ln -sf /etc/nginx/sites-available/rd_nginx.conf /etc/nginx/sites-enabled/default')
sudo('nginx -s reload')
def init_env():
sudo('aptitude update')
sudo('aptitude safe-upgrade')
# sudo('apt-get install nginx')
sudo('aptitude install python-software-properties')
sudo('add-apt-repository ppa:nginx/stable')
sudo('aptitude update')
sudo('apt-get install nginx')
sudo('apt-get install supervisor')
sudo('apt-get install redis-server')
sudo('apt-get install mysql-server')
sudo('apt-get install python-pip python-dev build-essential')
sudo('pip install virtualenv')
run('mkdir /var/log/raindrop -p')
with cd ('/home/raindrop'):
sudo('virtualenv env')
def oss_put_object_from_file(key, file_path):
from oss.oss_api import *
oss = OssAPI(_ALIYUN_OSS['endpoint'], _ALIYUN_OSS['accessKeyId'], _ALIYUN_OSS['accessKeySecret'])
res = oss.put_object_from_file(_ALIYUN_OSS['bucket'], key, file_path)
return res.status == 200 and True or False
def _expand_path(path):
print path
return '"$(echo %s)"' % path
def sed(filename, before, after, limit='', backup='.bak', flags=''):
# Characters to be escaped in both
for char in "/'":
before = before.replace(char, r'\%s' % char)
after = after.replace(char, r'\%s' % char)
# Characters to be escaped in replacement only (they're useful in regexen
# in the 'before' part)
for char in "()":
after = after.replace(char, r'\%s' % char)
if limit:
limit = r'/%s/ ' % limit
cOntext= {
'script': r"'%ss/%s/%s/%sg'" % (limit, before, after, flags),
'filename': _expand_path(filename),
'backup': backup
}
# Test the OS because of differences between sed versions
with hide('running', 'stdout'):
platform = local("uname")
if platform in ('NetBSD', 'OpenBSD', 'QNX'):
# Attempt to protect against failures/collisions
hasher = hashlib.sha1()
hasher.update(env.host_string)
hasher.update(filename)
context['tmp'] = "/tmp/%s" % hasher.hexdigest()
# Use temp file to work around lack of -i
expr = r"""cp -p %(filename)s %(tmp)s \
&& sed -r -e %(script)s %(filename)s > %(tmp)s \
&& cp -p %(filename)s %(filename)s%(backup)s \
&& mv %(tmp)s %(filename)s"""
else:
context['extended_regex'] = '-E' if platform == 'Darwin' else '-r'
expr = r"sed -i%(backup)s %(extended_regex)s -e %(script)s %(filename)s"
command = expr % context
return local(command)
def _cdnify(index_file, cdn_file):
sed(index_file, '([^<]*(src|href)=")[^<]*' + cdn_file + '"', '\\1http://' + _ALIYUN_OSS['bucket'] + '.' + _ALIYUN_OSS['endpoint'] + '/' + cdn_file + '"')
def _list_dir(dir=None, access=lcd, excute=local):
"""docstring for list_dir"""
if dir is None:
return []
with access(dir):
string = excute("for i in *; do echo $i; done", capture=True)
files = string.replace("\r","").split("\n")
return files
通过fabric自动化部署有两点需要注意:
1.安装mysql时,设置的密码不能生效,需要登到服务器上手动设置一下:
mysql -u root
use mysql;
update user set password=PASSWORD('123456') where User='root';
flush privileges;
quit;
2.服务器上的用户需要自己手动创建。
在build任务中,使用gulp build打包client的angular代码。
这里先不多说,有需要请参考github项目generator-gulp-angular
setup是python自己的打包工具。
build任务中,最后使用python setup.py sdist命令把整个工程打包成一个tar.gz文件,上传到服务器。
使用这个工具需要编写如下两个文件:setup.py:
打包脚本
#!/usr/bin/env python
from setuptools import setup, find_packages
import server
try:
long_description = open('README.md').read()
except:
long_description = server.__description__
REQUIREMENTS = []
exclude_lib = ['oss', 'distribute']
for lib in open("server/requirements/common.txt").readlines():
for exclude in exclude_lib:
if lib.lower() not in exclude:
REQUIREMENTS.append(lib)
setup(
name='raindrop',
url='https://www.yudianer.com',
version=server.__version__,
author=server.__author__,
author_email=server.__email__,
description=server.__description__,
long_description=long_description,
license=server.__license__,
packages=find_packages(),
zip_safe=False,
platforms='any',
install_requires=REQUIREMENTS
)
MANIFEST.in:
指定打包哪些文件夹
recursive-include etc *
recursive-include client/dist *
recursive-include server/app *
recursive-include server/requirements *
recursive-exclude server *.pyc
prune server/app/static/upload/images
prune server/env
prune server/tests
本文由raindrop网码农撰写。欢迎转载,但请注明出处。