在python脚本中设置环境变量，并运行相关应用

1. 问题

在自动化应用的时候 ，有时候环境变量与运行需要不一致。这时候有两种选择：

1. 改变节点环境变量，使得其和运行需求保持一致；
2. 在自动化脚本中设置环境变量，其范围只在脚本运行环境中有效。

显然，当需要运行多个自动化脚本，每个的需求不一致的时候，选择2是更好的设计。

2. 解决方案和案例

假设我们有两个java的运行版本，一个是本节点上的java，版本7；另一个是云盘上共享的版本8。我们需要运行elasticsearch 6.2.3，它需要java8。而我们不希望更新节点上的java：

• java 8路径：/shared/java8/bin
• elasticsearch 6.2.3路径：/opt/elasticsearch6.2.3/bin
•  elasticsearch用户：elasticsearch

import subprocess, re, os, socket, resource, pwd, crypt, logging, sys, platform

def demote(user_uid, user_gid):
    def result():
        os.setgid(user_gid)
        os.setuid(user_uid)
    return result

class setuputil:
    @staticmethod
    def run_command(args, username=None, usercwd=None):
        try:
            if username:
                pw_record = pwd.getpwnam(username)
                env = os.environ.copy()
                env['HOME']  = pw_record.pw_dir
                env['LOGNAME']  = username
                env['USER']  = username
                if usercwd:
                    env['PWD']  = usercwd
                p = subprocess.Popen(args, preexec_fn=demote(pw_record.pw_uid, pw_record.pw_gid),
                                     cwd=usercwd, env=env, close_fds=True)
            else:
                p = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.PIPE, close_fds = True)

            outputs = p.communicate()
            for out in outputs:
                if out:
                    return (p.returncode, out)
        except OSError as os_e:
            return (os_e.errno, os_e.strerror)
        except ValueError as val_e:
            return (val_e.errno, val_e.strerror)
        else:
            return (p.returncode, outputs)

    @staticmethod
    def java_version():
        java_ver_pattern = '\"(\d+\.\d+).*\"'
        java_cmd = "java"
        if 'JAVA_HOME' in os.environ:
            java_cmd = os.environ['JAVA_HOME'] + "/bin/java"
        version = subprocess.check_output([java_cmd, '-version'], stderr=subprocess.STDOUT)
        if version:
            return re.search(java_ver_pattern, version).groups()[0]
        else:
            return None

    @staticmethod
    def start_app(es_exec, pid_file, es_user):
        rtcode, msg = setuputil.run_command([es_exec, '-d', '-p', pid_file], es_user)
        if rtcode:
            return (rtcode, msg)
        else:
            return (0, "OK")

MIN_JAVA_VERSION = 1.8

ELASTIC_USER = 'elastic'
ELASTIC_PASSWORD = 'elastic'
ELASTIC_PID = '/tmp/elasticpid'

def main():
    # set up java home
    os.environ['JAVA_HOME'] = '/shared/java8'

    j_ver = setuputil.java_version()
    if not j_ver or float(j_ver) < MIN_JAVA_VERSION:
        logging.error("Java isn't installed or java version is not up to date: %s", j_ver if j_ver else "None")
        return (1, "Java isn't installed or java version is not up to date: %s"%(j_ver if j_ver else "None"))

    rtcode, msg = setuputil.start_app('/opt/elasticsearch6.2.3/bin/elasticsearch', ELASTIC_PID, ELASTIC_USER)
    if rtcode:
        logging.error("Failed to start elasticsearch: %d, %s", rtcode, msg)
        return (rtcode, msg)
    else:
        logging.debug("Start elasticsearch: %s", msg)
        return (0, None)

if __name__ == '__main__':
    main()

3. 注意的问题

有几个需要注意的问题：

• 环境变量的适用范围(scope)
  • 一般只在运行的当前shell有效
  • 就上例而言，JAVA_HOME只在运行脚本的进程里是有效
• 在哪里设置环境变量？
  • 需要在运行命令之前设置　　
  • 就上例而言，JAVA_HOME需要在运行java版本检查，和运行elasticsearch之前设置