由于在测试环境上用docker部署了多个应用,而且他们的端口有的相同,有的又不相同,数量也比较多,在使用jenkins发版本的时候,不好配置,于是想要写一个脚本,能在docker 容器创建、停止的时候,自动生成nginx反向代理,然后reload nginx
我的原则是尽量简单,轻量,内存占用少
目标很明确,只要能监听到docker的容器启动/停止事件,即可
网上查了一下可以用docker events来监听docker事件,试了一下,发现基本可以满足,于是用python写了一段程序,用来监听docker事件
python
#!/usr/bin/python
# coding: utf8
import os
import json
import re
import subprocess
def override(path, text):
if not os.path.exists(path) and os.path.exists(path+"_temp"):
os.rename(path+"_temp",path)
fw = open(path+"_temp", 'wb')
fw.write(text)
fw.close()
if os.path.exists(path):
os.remove(path)
os.rename(path+"_temp", path)
def read(path):
try:
fr = open(path, "rb")
except IOError:
print "The file don't exist, Please double check!"
return
lines = fr.readlines()
ret = ''
for line in lines:
ret += line
return ret
def read_jsonfile(path):
return json.loads(read(path))
def cmd(command):
return os.popen(command).read()
def get_name(container):
return cmd("docker inspect -f '{{.Name}}' " + container).replace("/", "").replace('\n', '')
def get_ip(container):
return cmd("docker inspect -f '{{.NetworkSettings.IPAddress}}' " + container).replace('\n', '')
def get_port(container):
return cmd("docker inspect -f '{{.Config.ExposedPorts}}' " + container).replace('/tcp:{}]', '').replace('map[', '').replace('\n', '')
def get_info(container):
filename = "/var/lib/docker/containers/" + container + "/config.v2.json"
config = read_jsonfile(filename)
name = config['Name'].replace("/", "")
port = config['Config']['ExposedPorts'].keys()[0].replace('/tcp', '')
ip = cmd("docker inspect -f '{{.NetworkSettings.IPAddress}}' " + name)
# ip = config['NetworkSettings']['Networks']['bridge']['IPAddress']
ret = {'name': name, 'port': port, 'ip': ip}
return ret
tpl = """
server {
listen 80;
server_name $name.test.com;
location / {
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header Host $http_host;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_pass http://$ip:$port;
}
}
"""
def generate_conf():
print "generate_conf"
out = cmd("docker ps | grep -v CONTAINER | awk '{print $1}'")
containers = out.split("\n")
servers = ''
hosts = ''
for con in containers:
if con != '':
name = get_name(con)
ip = get_ip(con)
port = get_port(con)
print ip, port
if len(port) >= 2:
servers += tpl.replace("$name", name).replace("$ip", ip).replace("$port", port)
hosts += "11.12.13.14 " + name + ".test.com\n"
override('/usr/local/openresty/nginx/conf/vhost.conf', servers)
override('/usr/local/openresty/nginx/html/vhost.html', "<pre>" + hosts + "</pre>")
def reload_nginx():
print "reload nginx"
cmd('nginx -s reload')
def auto_reload():
generate_conf()
reload_nginx()
print " ==================== docker events ==================== "
# auto_reload()
proc = subprocess.Popen(["docker", "events"],
# shell=True, # windows: true, linux: false
stdout=subprocess.PIPE)
while 1:
out = proc.stdout.readline()
event = re.sub('\(|\)', "", out).split(" ")
if out.find('container stop') != -1:
auto_reload()
print ' container stop '
elif out.find('container start') != -1:
auto_reload()
print ' start container '
if out == '':
print "out "
break
启动命令:
nohup ./docker.py > /dev/null 2>&1 &
程序会在后台运行,断开ssh也不会结束
主要就是生成一个 conf 文件,这个文件要在nginx.conf里面引入,然后每次有容器启动/停止都生成这个文件,然后重启nginx,我这了还把容器名加上一个域名,组合成了一个子域名,然后把对应的映射关系生成了一个html文件,通过浏览器可以访问这个文件,然后把对应的代码 复制到本机的 hosts 文件里面,可以实现通过域名访问应用,当然只是开发测试的时候会这么做,但是也足够了。