如何将Docker环境变量传递给Telegraf inputs.exec调用的脚本？含报错排查

问题背景 #

构建运行Telegraf inputs.exec插件的Docker镜像，插件调用exec_oracle_python.sh脚本读取Dockerfile定义的username、password、sid环境变量，启动容器时触发exec format error，且环境变量无法正常传递给脚本；但在Python脚本中硬编码变量时Telegraf运行正常。

报错信息 #

$ sudo docker run wf-telegraf-python:test6
2022-07-18T19:11:36Z I! Using config file: /etc/telegraf/telegraf.conf
2022-07-18T19:11:36Z I! Starting Telegraf 1.22.0
2022-07-18T19:11:36Z I! Loaded inputs: exec
2022-07-18T19:11:36Z I! Loaded aggregators: 
2022-07-18T19:11:36Z I! Loaded processors: 
2022-07-18T19:11:36Z I! Loaded outputs: wavefront
2022-07-18T19:11:36Z I! Tags enabled: host=72d8b202e548
2022-07-18T19:11:36Z I! [agent] Config: Interval:1m0s, Quiet:false, Hostname:"72d8b202e548", Flush Interval:10s
2022-07-18T19:11:36Z E! [inputs.exec] Error in plugin: exec: fork/exec /exec_oracle_python.sh: exec format error for command '/exec_oracle_python.sh': 

原始配置文件 #

Dockerfile

FROM <redacted>/telegraf:latest

ENV username=\"user\"
ENV password=\"123\"
ENV sid=\"orcl\"

COPY telegraf.conf /etc/telegraf/
COPY /telegraf.d/10-output.conf /etc/telegraf/telegraf.d/
COPY entrypoint.sh /entrypoint.sh
COPY py_script.py /py_script.py 
COPY exec_oracle_python.sh /exec_oracle_python.sh

RUN DEBIAN_FRONTEND=noninteractive apt-get update && \
    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends python3 python3-pip python3-setuptools python3-wheel && \
    rm -rf /var/lib/apt/lists/*

RUN pip3 install requests pyyaml

CMD [\"/bin/sh\", \"-c\", \"echo ${username}\"]
CMD [\"--config-directory\",\"/etc/telegraf/telegraf.d\"]
ENTRYPOINT [ \"/entrypoint.sh\" ]

telegraf.conf（输入插件部分）

[[inputs.exec]]
commands = [\"/exec_oracle_python.sh\"]
timeout = \"5s\"
name_suffix = \"_mycollector\"
data_format = \"wavefront\"

exec_oracle_python.sh

python3 py_script.py -u $username -p $password -s $sid

py_script.py

import time
import argparse

parser = argparse.ArgumentParser()
parser.add_argument('-u', '--user', help="Pass the username with SELECT_CATALOG_ROLE role granted", required=True)
parser.add_argument('-p', '--passwd', required=True)
parser.add_argument('-s', '--sid', help="SID to connect to", required=True)
args = parser.parse_args()

servcount = 10
t = int(time.time())
mydata = "error.count %d %d source=dev env=staging user=%s password=%s sid=%s" % (servcount, t, args.user, args.passwd, args.sid)
print(mydata)

entrypoint.sh

#!/bin/bash
set -e

if [ "${1:0:1}" = '-' ]; then
    set -- telegraf "$@"
fi

if [ $EUID -ne 0 ]; then
    exec "$@"
else
    setcap cap_net_raw,cap_net_bind_service+ep /usr/bin/telegraf || echo "Failed to set additional capabilities on /usr/bin/telegraf"
    exec setpriv --reuid telegraf --init-groups "$@"
fi

解决方案 #

1. 修复exec格式错误 #

报错核心原因是exec_oracle_python.sh缺少脚本解释器声明（shebang），系统无法识别执行方式，同时要确保脚本拥有可执行权限：

• 修改exec_oracle_python.sh，在第一行添加shebang，并修正Python脚本的绝对路径：

  #!/bin/bash
  python3 /py_script.py -u $username -p $password -s $sid
  

• 在Dockerfile中添加权限设置命令：

  RUN chmod +x /exec_oracle_python.sh /entrypoint.sh
  

2. 正确传递Docker环境变量 #

修正Dockerfile环境变量定义

原始ENV变量多了额外引号，会导致变量值携带无效引号，需删除：

ENV username=user
ENV password=123
ENV sid=orcl

解决Dockerfile重复CMD问题

Dockerfile中最后一个CMD会覆盖之前的配置，保留Telegraf启动所需的CMD即可：

# 删除多余的CMD行，只保留
CMD ["--config-directory","/etc/telegraf/telegraf.d"]

确保Telegraf继承环境变量

Telegraf默认会继承容器的环境变量，entrypoint.sh切换到telegraf用户时也会保留这些变量。如果仍有问题，可以直接在Telegraf配置中通过bash执行命令，强制解析环境变量：

[[inputs.exec]]
commands = ["/bin/bash -c 'python3 /py_script.py -u $username -p $password -s $sid'"]
timeout = "5s"
name_suffix = "_mycollector"
data_format = "wavefront"

内容的提问来源于stack exchange，提问作者nadir781