docker-library / docker-library/python
Debbuging a `-slim` image?
还没有人认领这个 Issue。
- 主要语言
- Dockerfile
- 星标
- 2.8k
- 派生
- 1.1k
- 平均合并
- 3 天 1 小时
- 30 天内合并 PR
- 1
描述
We are hitting a rare deadlock in production that can't be reproduced using debug images.
Only way to debug it is to attach to the production image as the problem happens.
Problem is that production are based on the -slim images.
apt install gdb and python3-dbg is not enough because the python running is built from source and not aligned with what is available in apt.
I've tried to start the corresponding non-slim image but their binaries don't seem to align enough to make gdb happy:
Start python inside -slim image
➡️ My end goal is to run gdb py-bt on this process:
docker run -ti python:3.11-slim-bookworm python3 -c "import time; time.sleep(1000)"
Inside this image there is no chance to debug at all due to missing gdb and debug symbols. This is expected.
Build a debugger image from the corresponding non-slim image.
FROM python:3.11-bookworm
RUN --mount=type=cache,target=/var/cache/apt --mount=type=cache,target=/var/lib/apt \
apt update -y && \
apt install -y gdb
docker build -t mydebugpython .
Use the newly built debug image to attach into the first running container
Find the container id and use that to attach to the same pid-namespace.
docker run \
--cap-add SYS_PTRACE \
--pid container:15aab7ea5f57 \
--privileged \
-ti \
--entrypoint bash docker.io/library/mydebugpython
root@bb90d04f8722:/# gdb -q python --pid 1 -ex "py-bt"
Reading symbols from python...
Attaching to program: /usr/local/bin/python, process 1
warning: Build ID mismatch between current exec-file /usr/local/bin/python
and automatically determined exec-file /usr/local/bin/python3.11
exec-file-mismatch handling is currently "ask"
Load new symbol table from "/usr/local/bin/python3.11"? (y or n) y
Reading symbols from /usr/local/bin/python3.11...
Reading symbols from target:/usr/local/bin/../lib/libpython3.11.so.1.0...
(No debugging symbols found in target:/usr/local/bin/../lib/libpython3.11.so.1.0)
Reading symbols from target:/lib/x86_64-linux-gnu/libc.so.6...
Reading symbols from /usr/lib/debug/.build-id/82/ce4e6e4ef08fa58a3535f7437bd3e592db5ac0.debug...
Reading symbols from target:/lib/x86_64-linux-gnu/libm.so.6...
Reading symbols from /usr/lib/debug/.build-id/ea/87e1b3daf095cd53f1f99ab34a88827eccce80.debug...
Reading symbols from target:/lib64/ld-linux-x86-64.so.2...
Reading symbols from /usr/lib/debug/.build-id/38/e7d4a67acf053c794b3b8094e6900b5163f37d.debug...
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
__GI___clock_nanosleep (clock_id=1, flags=1, req=0x7ffd9d9ec458, rem=0x0) at ../sysdeps/unix/sysv/linux/clock_nanosleep.c:71
71 ../sysdeps/unix/sysv/linux/clock_nanosleep.c: No such file or directory.
Traceback (most recent call first):
(unable to read python frame information) # <-------------- :(
So it seems like the difference of slim and non-slim isn't only the presence of debug symbols or not, but also the binary is built differently?
Is this possible to solve somehow? My knowledge of how python is built and how to operate gdb are only moderate.
贡献指南
这个仓库没有索引到贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
调研方向
未指定任何 repository 文件或测试。首先比较 python:3.11-slim-bookworm 和 python:3.11-bookworm 镜像的构建,然后重现 gdb attach 命令,并检查可执行文件与调试符号之间的不匹配。完成的标准是确定 py-bt 是否可以针对 slim 镜像工作,并记录一种可复现的调试方法或限制。
由索引模型根据 Issue 内容生成。
评估
- 技术栈
- docker, python
- 领域
- devops, tooling
- Issue 类型
- 缺陷
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 活跃度
- 停滞
- 描述清晰度
- 基本清楚
- 新手友好度
- 25/100