基于docker搭建测试环境(二)
简介（已过时）
前文基于docker搭建测试环境展示了docker ci环境的搭建过程，本文将进一步丰富细节，主要讲述“将web项目编译，创建镜像以及将镜像部署到某个主机上，创建容器并运行”的脚本实现。
流程
脚本语言的选择
- linux shell,操作swarm和registry，使用remote api
- linux shell,操作swarm使用
docker -H xxx command，操作registry使用remote api - python脚本,操作swarm使用docker-py，操作registry使用remote api
作者一开始采用第一种方案实现了一个shell脚本，但在使用过程中发现如下问题
-
处理请求返回的响应值比较困难，即便在linux中安装jq等工具，代码还是比较复杂。
- 有些请求不返回任何结果，表示正常运行
- 有些请求返回一个id，表示正常运行
- 有些请求返回一个json字符串，需要解析并获取其中的key值
-
异常处理代码非常丑陋，有时甚至无法及时发现异常（即实际运行失败，在jenkins仍显示build成功）。
使用python后，python的语言处理能力自不用多说，还可以使用try except语句块优雅的处理异常。
脚本思路
设置环境变量
脚本中用到的一些环境变量需要预先设置（jenkins需要环境变量相关插件）
DOCKER_ADDRESS=192.168.56.154:2375
REGISTRY_ADDRESS=192.168.56.154:5000
# 需要将容器的8080端口映射为主机的哪个端口
PORT=3000
一些约定
- web项目编译成war包后，基于
REGISTRY_ADDRESS/tomcat7镜像生成REGISTRY_ADDRESS/JOB_NAME镜像。 - 容器在某个主机运行时，容器的name是JOB_NAME（JOB_NAME表示jenkins job的名字，可从jenkins自带的环境变量中获取）
- 容器对外暴漏8080和8000端口，8080在主机上的映射端口由PORT变量指定，8000端口作为调试端口，其映射端口由容器自动指定
- web项目在容器的
/logs目录下记录日志，/logs目录映射到主机的/logs目录
脚本实现
from docker import AutoVersionClient
import os
import sys
# python 3.x
# import http.client
# python 2.x
import httplib
PORT = os.getenv('PORT')
JOB_NAME = os.getenv('JOB_NAME')
DOCKER_ADDRESS = os.getenv('DOCKER_ADDRESS')
REGISTRY_ADDRESS = os.getenv('REGISTRY_ADDRESS')
WORKSPACE = os.getenv('WORKSPACE')
DOCKERFILE = WORKSPACE + '/Dockerfile'
IMAGE_NAME = JOB_NAME
FULL_IMAGE_NAME=REGISTRY_ADDRESS + '/' + IMAGE_NAME
# http://docker-py.readthedocs.org/en/latest/api/#containers
jenkinsCli = AutoVersionClient(base_url='localhost:2375',timeout=10)
swarmCli = AutoVersionClient(base_url=DOCKER_ADDRESS,timeout=10)
# registryConn = http.client.HTTPConnection(REGISTRY_ADDRESS)
registryConn = httplib.HTTPConnection(REGISTRY_ADDRESS)
def remove_image_in_registry():
try:
registryConn.request('GET','v1/repositories/' + IMAGE_NAME + '/tags/latest')
response = registryConn.getresponse()
print("query image %s in registry,http status %d" %(FULL_IMAGE_NAME,response.status))
# if image is in registry
if response.status / 100 == 2 :
print("registry image %s id : %s" %(FULL_IMAGE_NAME,response.read()))
registryConn.request('DELETE','v1/repositories/'+ IMAGE_NAME + '/tags/latest')
r = registryConn.getresponse()
print("delete image %s in registry,http status %d" %(FULL_IMAGE_NAME,r.status))
else :
print("query image %s in registry fail,resason %s" %(FULL_IMAGE_NAME,response.reason))
except Exception as inst:
print(type(inst)) # the exception instance
print(inst.args) # arguments stored in .args
print(inst) # __str__ allows args to be printed directly
pass
finally:
registryConn.close()
def remove_container_with_imageName():
response = swarmCli.containers(all=True)
container_num = len(response)
if container_num > 0 :
i=0
while i < container_num :
if response[i]['Image'] == FULL_IMAGE_NAME :
print('remove container :' + response[i]['Names'][0])
print('remove container id :' + response[i]['Id'])
swarmCli.remove_container(response[i]['Id'],force=True)
i=i+1
def clear():
# delete image in jenkins
try:
print("remove image : %s in jenkins" %(FULL_IMAGE_NAME))
jenkinsCli.remove_image(FULL_IMAGE_NAME,force=True,noprune=False)
except Exception as inst:
print(type(inst)) # the exception instance
print(inst.args) # arguments stored in .args
print(inst) # __str__ allows args to be printed directly
pass
# remove image in registry
remove_image_in_registry()
# remove container in swarm
remove_container_with_imageName()
# remove image in swarm
try:
print("remove image : %s in swarm" %(FULL_IMAGE_NAME))
swarmCli.remove_image(FULL_IMAGE_NAME,force=True,noprune=False)
except Exception as inst:
print(type(inst)) # the exception instance
print(inst.args) # arguments stored in .args
print(inst) # __str__ allows args to be printed directly
pass
def build():
# remove dockerfile
if(os.path.isfile(DOCKERFILE)):
os.remove(DOCKERFILE)
file = open(DOCKERFILE,mode='a+')
file.write("FROM %s \n"%(REGISTRY_ADDRESS + '/tomcat7'))
file.write('ADD target/*.war $TOMCAT_HOME/webapps/ \n')
file.write('CMD bash start.sh')
file.close()
print('build dockerfile')
response = [line for line in jenkinsCli.build(path=WORKSPACE, rm=True, tag=FULL_IMAGE_NAME)]
response
def push():
print("push image : %s from jenkins to registry" %(FULL_IMAGE_NAME))
response = [line for line in jenkinsCli.push(FULL_IMAGE_NAME, stream=True)]
response
def run():
# pull image
for line in swarmCli.pull(FULL_IMAGE_NAME, stream=True):
print(line)
# create container
config = swarmCli.create_host_config(binds=['/logs:/logs'],port_bindings={8080:PORT,8000:None},publish_all_ports=True)
container = swarmCli.create_container(image=FULL_IMAGE_NAME,name=IMAGE_NAME,ports=[8080,8000],volumes=['/logs'],host_config=config)
print(container)
# start container
response = swarmCli.start(container=container.get('Id'))
print(response)
def main():
try:
clear()
# build dockerfile
build()
# push image from jenkins to registry
push()
# pull image,create container,start container
run()
sys.exit(0)
except Exception as inst:
print(type(inst)) # the exception instance
print(inst.args) # arguments stored in .args
print(inst) # __str__ allows args to be printed directly
sys.exit(1)
finally:
jenkinsCli.close()
swarmCli.close()
main()
回滚
基于docker ci的回滚。其实就是docker镜像保存最后一次正确的历史版本（需要将hudson build序号跟镜像的tag联系起来），一旦最新代码运行有问题，就找到最后一次正确的docker镜像，部署该镜像即可。
小结
脚本的shell版本和python版本可在git@github.com:qiankunli/dockerci.git中下载。
代码中，为了可读性加了许多注释，代码本身还是比较简单的。
作者第一次使用python写程序，有问题的地方，欢迎大家指正。
留下评论