多看官网是没错的。 https://docs.docker.com/engine/reference/builder/#usage

. 和 * 在dockerfile 中区别

目录结构

[root@k8s-master model]# tree
.
├── aa
│   ├── _aaa.md
│   └── bb
│       └── cc
│           └── dd
│               └── ee
│                   └── _ee.md
├── Dockerfile
├── Dockerfile_cp_.
├── Dockerfile_cp_2
├── graph.pbtxt
├── instances_train.json
├── publish.log
└── task_0
    └── mask_rcnn_weight_best.h5
6 directories, 9 files

. 是当前 整个目录（包含目录本身）

[root@k8s-master model]# docker build -t docker_cp_test:cp_. -f Dockerfile_cp_. .
Sending build context to Docker daemon    284MB
Step 1/2 : FROM busybox:1.30
1.30: Pulling from library/busybox
53071b97a884: Pull complete 
Digest: sha256:4b6ad3a68d34da29bf7c8ccb5d355ba8b4babcad1f99798204e7abb43e54ee3d
Status: Downloaded newer image for busybox:1.30
 ---> 64f5d945efcc
Step 2/2 : COPY . /home/test/model/
 ---> 3907e815300f
Successfully built 3907e815300f
Successfully tagged docker_cp_test:cp_.
[root@k8s-master model]# docker run --rm -it docker_cp_test:cp_. ls /home/test/model
Dockerfile            aa                    publish.log
Dockerfile_cp_.       graph.pbtxt           task_0
Dockerfile_cp_2       instances_train.json

COPY * 是目录下所有 （不包含目录本身）

https://docs.docker.com/engine/reference/builder/#copy
Ifis a directory, the entire contents of the directory are copied, including filesystem metadata.
The directory itself is not copied, just its contents.

但是看到 有 ls /home/test/model/bb/cc/dd/ee/_ee.md，这个 还是没问题的。 A context is processed recursively.
多看官网是没错的。

[root@k8s-master model]# cat Dockerfile_cp_2
FROM busybox:1.30
COPY * /home/test/model/
[root@k8s-master model]# docker build -t docker_cp_test:cp_2 -f Dockerfile_cp_2 .
Sending build context to Docker daemon    284MB
Step 1/2 : FROM busybox:1.30
 ---> 64f5d945efcc
Step 2/2 : COPY * /home/test/model/
 ---> e6910b62dab3
Successfully built e6910b62dab3
Successfully tagged docker_cp_test:cp_2
[root@k8s-master model]# docker run --rm -it docker_cp_test:cp_2 ls /home/test/model
Dockerfile                _aaa.md                   instances_train.json
Dockerfile_cp_.           bb                        mask_rcnn_weight_best.h5
Dockerfile_cp_2           graph.pbtxt               publish.log
[root@k8s-master model]# docker run --rm -it docker_cp_test:cp_2 ls /home/test/model/bb/cc/dd/ee/_ee.md
/home/test/model/bb/cc/dd/ee/_ee.md

dockerfile RUN 的 命令中 有 “”

To use a different shell, other than ‘/bin/sh’, use the_exec_form passing in the desired shell. For example:
RUN ["/bin/bash", "-c", "echo hello"]

Note
The_exec_form is parsed as a JSON array, which means that you must use double-quotes (“) around words not single-quotes (‘).

比如 我想执行

sed -i "/format/a \        '\/transfer_learning'," test_factory.py
sed -i "/format/d" test_factory.py

所以我就在 dockerfile 里写

FROM {{ base_image }}
RUN ["/bin/bash", "-c", "sed -i \"/format/a \\        '\/trained_model_services',\" /opt/letrain-run-script/deploy_service/test_factory.py "]
RUN ["/bin/bash", "-c", "sed -i \"/format/d\" /opt/letrain-run-script/deploy_service/test_factory.py "]
COPY . {{ model_copy_dest }}
ENTRYPOINT ["python", "/opt/letrain-run-script/deploy_service/deploy_service.py",\
    "--scenario={{ tarined_model_scenario }}",\
    "--deploy_path={{ tarined_model_location }}", "--port=80",\
    "--model_name={{ tarined_model_name }}"]
ENV    PATH=$PATH":/usr/local/cuda/bin"
ENV    LETRAIN_PATH="/opt/letrain"
ENV    export LETRAIN_PATH

before

from resoure import DeployServiceRunner
def create_app():
    import falcon
    import os
    api = falcon.API()
    scenario = os.environ.get('LETRAIN_SCENARIO', '')
    model_path = os.environ.get('LETRAIN_MODEL_PATH', '')
    model_name = os.environ.get('LETRAIN_MODEL_NAME', '')
    service_uuid = os.environ.get('SERVICE_UUID', '')
    api.add_route(
        '/internal_api/service/{0}'.format(service_uuid),
        DeployServiceRunner(scenario, model_path, model_name)
    )
    return api

after

from resoure import DeployServiceRunner
def create_app():
    import falcon
    import os
    api = falcon.API()
    scenario = os.environ.get('LETRAIN_SCENARIO', '')
    model_path = os.environ.get('LETRAIN_MODEL_PATH', '')
    model_name = os.environ.get('LETRAIN_MODEL_NAME', '')
    service_uuid = os.environ.get('SERVICE_UUID', '')
    api.add_route(
        '/transfer_learning',
        DeployServiceRunner(scenario, model_path, model_name)
    )
    return api