Docker Support Tensorflow start
Tensoflow 容器启动
启动Tensorflow容器还是比较简单的，根据docker hub提供的方法进行容器启动即可
1.搜索镜像
root@es2:~# docker search tensorflow
NAME DESCRIPTION STARS OFFICIAL AUTOMATED
tensorflow/tensorflow Official docker images for deep learning f... 1001
jupyter/tensorflow-notebook Jupyter Notebook Scientific Python Stack w... 74
xblaster/tensorflow-jupyter Dockerized Jupyter with tensorflow 50 [OK]
romilly/rpi-docker-tensorflow Tensorflow and Jupyter running in docker c... 19
floydhub/tensorflow tensorflow 12 [OK]
bitnami/tensorflow-serving Bitnami Docker Image for TensorFlow Serving 11 [OK]
tensorflow/serving Official images for TensorFlow Serving (ht... 8
2.下载镜像
root@es2:~# docker pull tensorflow/tensorflow
Using default tag: latest
latest: Pulling from tensorflow/tensorflow
b234f539f7a1: Extracting [===========================> ] 23.4 MB/43.12 MB
55172d420b43: Download complete
5ba5bbeb6b91: Download complete
43ae2841ad7a: Download complete
f6c9c6de4190: Download complete
5624105f79a2: Downloading [======> ] 17.12 MB/142.4 MB
1cbdffd12405: Download complete
3.启动容器
docker run -d --name tensorflow_test -p 18888:8888 tensorflow/tensorflow
参数解析：
-d 后台运行
--name 给容器起个名字
-p 映射到本地的端口（本地端口:容器端口）
登录jupyter
1.找到登录token
jupyter登录需要传入token，获取方式如下：
docker ps
root@es2:~# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
17d804ae12cc tensorflow/tensorflow "/run_jupyter.sh -..." 58 seconds ago Up 58 seconds 6006/tcp, 0.0.0.0:18888->8888/tcp tensorflow_test
d939ac70635a google/cadvisor:latest "/usr/bin/cadvisor..." 5 days ago Up 5 days 0.0.0.0:8080->8080/tcp cadvisor
b9f4b7b1295c registry:2 "/entrypoint.sh /e..." 6 days ago Up 6 days 0.0.0.0:5000->5000/tcp registry2
找到要启动容器的ID
docker logs 17d804ae12cc
root@es2:~# docker logs 17d804ae12cc
[I 08:13:35.956 NotebookApp] Writing notebook server cookie secret to /root/.local/share/jupyter/runtime/notebook_cookie_secret
[W 08:13:35.972 NotebookApp] WARNING: The notebook server is listening on all IP addresses and not using encryption. This is not recommended.
[I 08:13:35.980 NotebookApp] Serving notebooks from local directory: /notebooks
[I 08:13:35.980 NotebookApp] 0 active kernels
[I 08:13:35.980 NotebookApp] The Jupyter Notebook is running at:
[I 08:13:35.980 NotebookApp] http://17d804ae12cc:8888/?token=4e392a15850c82958dd68a7e2eca205ae112740ce2ceb437
[I 08:13:35.980 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).
[C 08:13:35.980 NotebookApp]
Copy/paste this URL into your browser when you connect for the first time,
to login with a token:
http://17d804ae12cc:8888/?token=4e392a15850c82958dd68a7e2eca205ae112740ce2ceb437&token=4e392a15850c82958dd68a7e2eca205ae112740ce2ceb437
找到这条日志后，然后替换ip与端口
http://宿主机：映射端口/?token=4e392a15850c82958dd68a7e2eca205ae112740ce2ceb437
Jupyter.png
容器安装Python3 Anaconda3
如果需要环境中支持Anaconda需要手动去容器安装
1.登录容器
docker exec -it 17d804ae12cc bash
root@es2:~# docker exec -it 17d804ae12cc bash
root@17d804ae12cc:/notebooks#
2.更新apt-get安装wget
apt-get update
root@17d804ae12cc:/notebooks# apt-get update
Get:1 http://archive.ubuntu.com/ubuntu xenial InRelease [247 kB]
Get:2 http://security.ubuntu.com/ubuntu xenial-security InRelease [107 kB]
Get:3 http://archive.ubuntu.com/ubuntu xenial-updates InRelease [109 kB]
...
apt-get install wget
root@17d804ae12cc:/notebooks# apt-get install wget
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following NEW packages will be installed:
wget
...
3.下载Anaconda3
root@17d804ae12cc:/notebooks# wget https://repo.anaconda.com/archive/Anaconda3-5.2.0-Linux-x86_64.sh
--2018-08-07 08:35:17-- https://repo.anaconda.com/archive/Anaconda3-5.2.0-Linux-x86_64.sh
Resolving repo.anaconda.com (repo.anaconda.com)... 104.17.111.77, 104.17.110.77, 104.17.107.77, ...
Connecting to repo.anaconda.com (repo.anaconda.com)|104.17.111.77|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 651745206 (622M) [application/x-sh]
Saving to: 'Anaconda3-5.2.0-Linux-x86_64.sh'
Anaconda3-5.2.0-Linux-x86_64.sh 15%[===========>
...
Saving to: 'Anaconda3-5.2.0-Linux-x86_64.sh'
Anaconda3-5.2.0-Linux-x86_64.sh 100%[==================================================================================>] 621.55M 12.3MB/s in 49s
2018-08-07 08:36:07 (12.6 MB/s) - 'Anaconda3-5.2.0-Linux-x86_64.sh' saved [651745206/651745206]
4.运行安装
bash Anaconda3-5.2.0-Linux-x86_64.sh
root@17d804ae12cc:/notebooks# bash Anaconda3-5.2.0-Linux-x86_64.sh
Welcome to Anaconda3 5.2.0
In order to continue the installation process, please review the license
agreement.
Please, press ENTER to continue
>>>
5.jupyter notebook加载pytho3内核
ipython kernel install --name python3
root@17d804ae12cc:/notebooks# ipython kernel install --name python3
Installed kernelspec python3 in /usr/local/share/jupyter/kernels/python3
保存镜像
安装了Anaconda3的容器，想在别的机器上使用，而不想再次安装的话，可以将已经安装的容器保存成镜像
1.创建镜像
docker commit -a "test" -m "install Anaconda3-5.2.0" 17d804ae12cc tensorflow:1.0.0
如此就把容器保存成了加载中的镜像，其中TAG为了区分同名镜像的不同版本
语法 docker commit [OPTIONS] CONTAINER [REPOSITORY[:TAG]]
说明：
-a:提交的镜像作者
-c:使用Dockerfile指令来创建镜像
-m:提交时的说明文字
-p：在commit时，将容器暂停
root@es2:~# docker commit -a "test" -m "install Anaconda3-5.2.0" 17d804ae12cc tensorflow:1.0.0
sha256:8ff9c17845442d36f0fe43fbc65fa6cda6aef7a7bd5d69455eaf232b9fc6ceaf
- 查看镜像
root@es2:~# docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
tensorflow 1.0.0 8ff9c1784544 2 minutes ago 5.22 GB
es2:5000/itzzy/eureka-server-1.0.0 latest 3c8d61cfde11 10 days ago 723 MB
itmuch/eureka-server-1.0.0 latest 3c8d61cfde11 10 days ago 723 MB
itzzy/eureka-server-1.0.0 latest 3c8d61cfde11 10 days ago 723 MB
localhost:5000/itzzy/eureka-server-1.0.0 latest 3c8d61cfde11 10 days ago 723 MB
ubuntu 16.04 7aa3602ab41e 11 days ago 115 MB
tensorflow/tensorflow latest caab7ec02690 3 weeks ago 1.25 GB
2.保存镜像
docker save -o tensorflow_Anaconda3-5.2.0.tar tensorflow:1.0.0
可以看到生成了tar文件
-rw------- 1 root root 5340353536 Aug 7 17:19 tensorflow_Anaconda3-5.2.0.tar
3.其他主机加载镜像
只需要把你保存的镜像传过去load 一下就好了
docker load -i tensorflow_Anaconda3-5.2.0.tar
