使用dev镜像执行docker exec pytest时提示找不到可执行文件的问题

问题场景 #

使用rapidsai/rapidsai-dev:22.04-cuda11.5-devel-ubuntu20.04-py3.9镜像构建容器后，在CI中执行以下命令时出现错误：

CONTAINER=$(docker run -d gpu-test)
docker exec $CONTAINER pytest

错误信息：

OCI runtime exec failed: exec failed: unable to start container process: exec: "pytest": executable file not found in $PATH: unknown

但改用Dockerfile中注释的rapidsai/rapidsai:22.04-cuda11.5-runtime-ubuntu20.04-py3.9镜像构建容器则无此问题，当前Dockerfile内容如下：

# Pulls the basic Image from NVIDIA repository
FROM rapidsai/rapidsai-dev:22.04-cuda11.5-devel-ubuntu20.04-py3.9
# FROM rapidsai/rapidsai:22.04-cuda11.5-runtime-ubuntu20.04-py3.9

# Updates OS libraries
RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
    --fix-missing git python3-setuptools python3-pip build-essential libcurl4-gnutls-dev \
    zlib1g-dev rsync vim nano cmake tabix

# Install libraries needed in the examples
RUN /opt/conda/envs/rapids/bin/pip install \
     scanpy==1.9.1 wget pytabix dash-daq \
     dash-html-components dash-bootstrap-components dash-core-components \
     utils pytest

RUN /opt/conda/envs/rapids/bin/pip install \
    torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116

WORKDIR /workspace
ENV HOME /workspace

原因分析 #

• 你通过/opt/conda/envs/rapids/bin/pip安装的pytest，实际位于/opt/conda/envs/rapids/bin/pytest路径下
• rapidsai-dev镜像的默认环境变量PATH未包含/opt/conda/envs/rapids/bin目录，导致直接执行pytest时系统找不到该命令
• 而rapidsai镜像默认激活了rapids conda环境，其PATH已包含上述目录，因此能正常找到pytest

解决方案 #

方案1：直接调用pytest的完整路径 #

修改CI中的执行命令，指定pytest的绝对路径：

docker exec $CONTAINER /opt/conda/envs/rapids/bin/pytest

方案2：在Dockerfile中添加conda环境路径到PATH #

在Dockerfile末尾添加ENV指令，将conda环境的bin目录加入全局PATH：

WORKDIR /workspace
ENV HOME /workspace
# 新增以下行
ENV PATH="/opt/conda/envs/rapids/bin:$PATH"

重新构建镜像后，直接执行docker exec $CONTAINER pytest即可正常运行。

方案3：执行命令时先激活conda环境 #

通过bash执行命令，先激活rapids环境再运行pytest：

docker exec $CONTAINER bash -c "source /opt/conda/bin/activate rapids && pytest"

内容的提问来源于stack exchange，提问作者fabio.geraci