说明

即使该系统有用户zfbroottestubuntu等,下面介绍的步骤只影响本用户,既不需要root权限,也不会对其他用户造成影响(开机自启的service文件需要root用户编辑和设置开机自启,之后就不需要操作了)

1. 创建虚拟环境jupyter

# 安装venv
sudo apt-get install python3-venv
# 创建虚拟环境,名称为jupyter
python3 -m venv jupyter

2. 安装nodejs(用于jupyterlab安装扩展)

# 下载nvm用于管理npm、nodejs环境
wget -qO- https://raw.githubusercontent.com/nvm-sh/nvm/v0.35.3/install.sh | bash
# 重新启动即可使用nvm命令
# nvm ls-remote          列出nodejs所有可用版本
# nvm install 10.10.0    安装nodejs 10.10.0版本
# 安装nodejs最新版本
nvm install node

把nvm环境bin文件夹放入PATH,即在~/.bashrc添加一行内容,必须把自己路径放在前面,避免先搜索到/usr/local/bin目录:

export PATH=/home/zfb/.nvm/versions/node/v14.5.0/bin:${PATH}

3. 安装pip包

# 激活虚拟环境jupyter
source jupyter/bin/activate
# 在虚拟环境jupyter中安装jupyterlab和nodejs
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple jupyterlab npm nodejs

4. 使用jupyterlab

先把python虚拟环境jupyterbin文件夹放入PATH,即在~/.bashrc添加一行内容,必须把自己路径放在前面,避免先搜索到/usr/local/bin目录:

export PATH=/home/zfb/jupyter/bin:${PATH}

在命令行输入jupyter lab即可在本地端口打开(不需要激活虚拟环境),可以通过命令which jupyter得到/home/zfb/jupyter/bin/jupyter结果
在jupyterlab运行期间,可以通过命令jupyter notebook list查看当前运行的jupyter实例
列出当前已安装的扩展:jupyter labextension list
卸载某个扩展:jupyter labextension uninstall my-extension-name
查看jupyter的kernel:jupyter kernelspec list
注意:http://127.0.0.1:8888/lab是jupyterlab的地址;http://127.0.0.1:8888/tree是传统jupyter notebook的地址

5. 配置jupyterlab

在终端输入以下命令生成加密秘钥:

# 激活虚拟环境jupyter
source jupyter/bin/activate
# 密码设置为123456,此命令输出密码的sha1结果,用于下一步配置文件token
python -c "from notebook.auth import passwd;print(passwd('123456'))"

在命令行输入jupyter lab --generate-config,则会生成文件/home/zfb/.jupyter/jupyter_notebook_config.py,打开该文件,修改以下内容:

c.NotebookApp.allow_remote_access = True
c.NotebookApp.ip = '0.0.0.0'
c.NotebookApp.notebook_dir = '/home/zfb/jp_data/'
c.NotebookApp.open_browser = False
c.NotebookApp.password = 'sha1:10d130e9bad7:b73d9821f96ccc4f42b2071b5dc46f2357373da3'
c.NotebookApp.port = 8888

安装扩展时如果找不到node,那么需要确保它在PATH,然后手动启动jupyter lab,不要使用service启动即可在浏览器点击install安装

6. 开机自启jupyter

切换root用户(zfb用户不能执行sudo命令),创建文件jupyter-zfb.service,内容如下:

[Unit]
Description=Auto start jupyter lab Service for web
After=network.target

[Service]
Type=simple
# Type=forking
# PIDFile=/var/pid/master.pid
# 如果是在为其他用户配置jupyterlab,这里填对应的用户名
User=zfb
Restart=on-failure
RestartSec=10s
WorkingDirectory=/home/zfb/jupyter
ExecStart=/home/zfb/jupyter/bin/jupyter lab
# ExecReload=/home/zfb/jupyter/bin/jupyter lab

[Install]
WantedBy=multi-user.target

然后依次执行下面命令:

# 复制jupyter-zfb.service文件到指定目录
sudo cp ./jupyter-zfb.service /etc/systemd/system/
# 设置jupyter-zfb开机自启
systemctl enable jupyter-zfb.service
# 重载service文件
sudo systemctl daemon-reload
# 查看所有的开机自启项
systemctl list-unit-files --type=service|grep enabled
# 手动开启jupyter-zfb服务
service jupyter-zfb start
# 查看jupyter-zfb服务的运行状态
service jupyter-zfb status
# 停止jupyter-zfb服务
service jupyter-zfb stop

查看服务状态的输出如下:

root1@my-Server:~$ service jupyter-zfb status
● jupyter-zfb.service - Auto start jupyter lab Service for web
   Loaded: loaded (/etc/systemd/system/jupyter-zfb.service; enabled; vendor preset: enabled)
   Active: active (running) since Sun 2020-07-19 23:59:44 CST; 3s ago
 Main PID: 19426 (jupyter-lab)
    Tasks: 1 (limit: 7372)
   CGroup: /system.slice/jupyter-zfb.service
           └─19426 /home/zfb/jupyter/bin/python3 /home/zfb/jupyter/bin/jupyter-lab

Jul 19 23:59:44 my-Server systemd[1]: Started Auto start jupyter lab Service for web.
Jul 19 23:59:44 my-Server jupyter[19426]: [I 23:59:44.704 LabApp] JupyterLab extension loaded from /home/zfb/
Jul 19 23:59:44 my-Server jupyter[19426]: [I 23:59:44.704 LabApp] JupyterLab application directory is /home/z
Jul 19 23:59:44 my-Server jupyter[19426]: [I 23:59:44.706 LabApp] Serving notebooks from local directory: /ho
Jul 19 23:59:44 my-Server jupyter[19426]: [I 23:59:44.706 LabApp] The Jupyter Notebook is running at:
Jul 19 23:59:44 my-Server jupyter[19426]: [I 23:59:44.706 LabApp] http://my-Server:8888/
Jul 19 23:59:44 my-Server jupyter[19426]: [I 23:59:44.706 LabApp] Use Control-C to stop this server and shut
root1@my-Server:~$

问题:service运行,则一旦安装扩展之后重新打开,扩展处就显示500 Internal Server Error;但是直接运行在控制台无问题;nohup jupyter lab &也无问题;screen也无问题

6. 开机自启和nohup运行

创建文件startjupyterlab.sh并分配执行权限:

#!/bin/bash
# 后台运行,重定向错误日志,导出pid到文件
# nohup会免疫HUP信号,>>表示追加模式
/usr/bin/nohup /home/zfb/jupyter/bin/jupyter lab >> /home/zfb/jupyter/log/jupyterlab.log 2>&1 & echo $! > /home/zfb/jupyter/run_jupyter.pid

ubuntu 18.04不再使用inited管理系统,改用systemd,原本简单方便的/etc/rc.local文件已经没有了。systemd默认读取/etc/systemd/system/下的配置文件,该目录下的文件会链接/lib/systemd/system/下的文件,一般系统安装完/lib/systemd/system/下会有rc-local.service文件,即我们需要的配置文件,里面有写到rc.local的启动顺序和行为,文件内容如下cat /lib/systemd/system/rc-local.service

#  SPDX-License-Identifier: LGPL-2.1+
#
#  This file is part of systemd.
#
#  systemd is free software; you can redistribute it and/or modify it
#  under the terms of the GNU Lesser General Public License as published by
#  the Free Software Foundation; either version 2.1 of the License, or
#  (at your option) any later version.

# This unit gets pulled automatically into multi-user.target by
# systemd-rc-local-generator if /etc/rc.local is executable.
[Unit]
Description=/etc/rc.local Compatibility
Documentation=man:systemd-rc-local-generator(8)
ConditionFileIsExecutable=/etc/rc.local
After=network.target

[Service]
Type=forking
ExecStart=/etc/rc.local start
TimeoutSec=0
RemainAfterExit=yes
GuessMainPID=no

systemctl status rc-local可以查看当前是否有rc-local这个服务,如果没有则需要创建ln -fs /lib/systemd/system/rc-local.service /etc/systemd/system/rc-local.service。设置开机启动并运行服务可以看到如下输出:

zfb@my-Server:~$ service rc-local status
● rc-local.service - /etc/rc.local Compatibility
   Loaded: loaded (/lib/systemd/system/rc-local.service; static; vendor preset: enabled)
  Drop-In: /lib/systemd/system/rc-local.service.d
           └─debian.conf
   Active: inactive (dead)
Condition: start condition failed at Mon 2020-07-20 14:39:15 CST; 2s ago
           └─ ConditionFileIsExecutable=/etc/rc.local was not met
     Docs: man:systemd-rc-local-generator(8)
zfb@ny-Server:~$

然后执行以下操作:

# 创建文件
sudo vim /etc/rc.local
# 添加内容
#  #!/bin/bash
#
#  su - zfb -c "/bin/bash /home/zfb/startjupyterlab.sh"

# 添加执行权限
sudo chmod +x /etc/rc.local

运行service rc-local start即可启动服务,service rc-local status查看运行状态
日志分割:然后创建文件/etc/logrotate.d/jupyter-zfb

su zfb zfb
/home/zfb/jupyter/log/jupyterlab.log{
    weekly
    minsize 10M
    rotate 10
    missingok
    dateext
    notifempty
    sharedscripts
    postrotate
        if [ -f /home/zfb/jupyter/run_jupyter.pid ]; then
            /bin/kill -9 `cat /home/zfb/jupyter/run_jupyter.pid`
        fi
        /usr/bin/nohup /home/zfb/jupyter/bin/jupyter lab >> /home/zfb/jupyter/log/jupyterlab.log 2>&1 & echo $! > /home/zfb/jupyter/run_jupyter.pid
    endscript
}

执行命令logrotate -dvf /etc/logrotate.d/jupyter-zfb可以查看每次轮询的输出

  • d表示只是显示,并不实际执行
  • v表示显示详细信息
  • f表示即使不满足条件也强制执行一次

7. 添加其他python环境的kernel

在不激活任何环境的终端,创建新的虚拟环境py36(最后把它添加到jupyter的kernel)

# 创建新的虚拟环境py36
python3 -m venv py36
# 激活新虚拟环境py36
source py36/bin/activate
# 为新环境安装需要的库
# pip install -i https://pypi.tuna.tsinghua.edu.cn/simple
# 为虚拟环境安装kernel
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple ipykernel
# 将此虚拟环境配置到jupyter的kernel中,此命令返回
# Installed kernelspec kernel_py36 in /home/zfb/.local/share/jupyter/kernels/kernel_py36
# 若不指定--user,则会提示权限不足,因为默认安装到/usr/local/share/jupyter
python -m ipykernel install --name kernel_py36 --user
# 启动jupyterlab,此时可以看到已经有两个kernel可供切换(jupyter、kernel_py36)
jupyter lab

删除某个kernel:jupyter kernelspec remove kernel_py36

8. 添加matlab的kernel

激活虚拟环境jupyter(jupyterlab被安装在此虚拟环境),然后安装matlab_kernal,再切换到matlab的安装目录extern/engines/python/,运行setup.py文件,具体步骤的命令如下:

# 激活虚拟环境jupyter
source jupyter/bin/activate
# 在虚拟环境jupyter安装matlab_kernel
pip install matlab_kernel
# 若不指定--user,则会提示权限不足
python -m matlab_kernel install --user
# 切换到matlab安装目录的extern/engines/python/,然后运行命令
python setup.py install
# --build-base="/home/zfb/build" install --prefix="/home/zfb/jupyter/lib/python3.6/site-packages"
# 此时运行jupyter kernelspec list即可看到如下输出
# Available kernels:
#   matlab     /home/zfb/jupyter/share/jupyter/kernels/matlab
#   python3    /home/zfb/jupyter/share/jupyter/kernels/python3

保证最后/home/zfb/jupyter/lib/python3.6/site-packages/文件夹下有matlab文件夹和matlab_kernel文件夹:

matlab
├── engine
│   ├── _arch.txt
│   ├── basefuture.py
│   ├── engineerror.py
│   ├── enginehelper.py
│   ├── enginesession.py
│   ├── fevalfuture.py
│   ├── futureresult.py
│   ├── __init__.py
│   ├── matlabengine.py
│   ├── matlabfuture.py
│   └── __pycache__
│       ├── basefuture.cpython-36.pyc
│       ├── engineerror.cpython-36.pyc
│       ├── enginehelper.cpython-36.pyc
│       ├── enginesession.cpython-36.pyc
│       ├── fevalfuture.cpython-36.pyc
│       ├── futureresult.cpython-36.pyc
│       ├── __init__.cpython-36.pyc
│       ├── matlabengine.cpython-36.pyc
│       └── matlabfuture.cpython-36.pyc
├── __init__.py
├── _internal
│   ├── __init__.py
│   ├── mlarray_sequence.py
│   ├── mlarray_utils.py
│   └── __pycache__
│       ├── __init__.cpython-36.pyc
│       ├── mlarray_sequence.cpython-36.pyc
│       └── mlarray_utils.cpython-36.pyc
├── mlarray.py
├── mlexceptions.py
└── __pycache__
    ├── __init__.cpython-36.pyc
    ├── mlarray.cpython-36.pyc
    └── mlexceptions.cpython-36.pyc
5 directories, 31 files


matlab_kernel
├── check.py
├── __init__.py
├── kernel.json
├── kernel.py
├── __main__.py
├── matlab
│   ├── engine
│   │   ├── _arch.txt
│   │   ├── basefuture.py
│   │   ├── engineerror.py
│   │   ├── enginehelper.py
│   │   ├── enginesession.py
│   │   ├── fevalfuture.py
│   │   ├── futureresult.py
│   │   ├── __init__.py
│   │   ├── matlabengine.py
│   │   ├── matlabfuture.py
│   │   └── __pycache__
│   │       ├── basefuture.cpython-36.pyc
│   │       ├── engineerror.cpython-36.pyc
│   │       ├── enginehelper.cpython-36.pyc
│   │       ├── enginesession.cpython-36.pyc
│   │       ├── fevalfuture.cpython-36.pyc
│   │       ├── futureresult.cpython-36.pyc
│   │       ├── __init__.cpython-36.pyc
│   │       ├── matlabengine.cpython-36.pyc
│   │       └── matlabfuture.cpython-36.pyc
│   ├── __init__.py
│   ├── _internal
│   │   ├── __init__.py
│   │   ├── mlarray_sequence.py
│   │   ├── mlarray_utils.py
│   │   └── __pycache__
│   │       ├── __init__.cpython-36.pyc
│   │       ├── mlarray_sequence.cpython-36.pyc
│   │       └── mlarray_utils.cpython-36.pyc
│   ├── mlarray.py
│   ├── mlexceptions.py
│   └── __pycache__
│       ├── __init__.cpython-36.pyc
│       ├── mlarray.cpython-36.pyc
│       └── mlexceptions.cpython-36.pyc
├── matlabengineforpython-R2020a-py3.6.egg-info
└── __pycache__
    ├── check.cpython-36.pyc
    ├── __init__.cpython-36.pyc
    ├── kernel.cpython-36.pyc
    └── __main__.cpython-36.pyc

7 directories, 41 files

可以参考链接1链接2

9. 使用frp内网穿透

腾讯云主机的frps.ini添加一行:

# 不需要和frpc.ini一致
vhost_http_port = 8888

运行jupyterlab的服务器的frpc.ini添加一个部分:

[web]
type = http
local_port = 8888
custom_domains = lab.example.cn

如果要使用frp内网穿透的同时又给它设置域名,则域名解析记录添加一条名称为lab的A记录到腾讯云主机的IP(frps),在腾讯云主机再添加一个nginx项:

    server{
        listen 80;
        # 如果需要ssl,参考https://blog.whuzfb.cn/blog/2020/07/07/web_https/
        # listen 443 ssl;
        # include ssl/whuzfb.cn.ssl.conf;
        # 此时支持http与https
        server_name lab.example.cn;
        access_log /home/ubuntu/frp_linux_amd64/log/access_jupyter.log;
        error_log /home/ubuntu/frp_linux_amd64/log/error_jupyter.log;
        # 防止jupyter保存文件时413 Request Entity Too Large
        # client_max_body_size 50m; 0表示关闭检测
        client_max_body_size 0;
        location /{
            proxy_set_header Host $host;
            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
            proxy_redirect off;
            proxy_buffering off;
            proxy_pass http://127.0.0.1:8888;
        }

        location ~* /(api/kernels/[^/]+/(channels|iopub|shell|stdin)|terminals/websocket)/? {
            proxy_pass http://127.0.0.1:8888;
            proxy_set_header X-Real-IP $remote_addr;
            proxy_set_header Host $host;
            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
            # WebSocket support
            proxy_http_version 1.1;
            proxy_set_header Upgrade $http_upgrade;
            proxy_set_header Connection "upgrade";
        }
        # -------  旧方法:还是有部分报错/api/kernels err_too_many_redirects  ---------
        # # 必须有,否则请求/api/kernels/ 的状态码都是400
        # location /api/kernels/ {
        #     proxy_pass            http://127.0.0.1:8888;
        #     proxy_set_header      Host $host;
        #     # websocket support
        #     proxy_http_version    1.1;
        #     proxy_set_header      Upgrade "websocket";
        #     proxy_set_header      Connection "Upgrade";
        #     proxy_read_timeout    86400;
        # }
        # # 必须有,否则请求/terminals/ 的状态码都是400
        # location /terminals/ {
        #     proxy_pass            http://127.0.0.1:8888;
        #     proxy_set_header      Host $host;
        #     # websocket support
        #     proxy_http_version    1.1;
        #     proxy_set_header      Upgrade "websocket";
        #     proxy_set_header      Connection "Upgrade";
        #     proxy_read_timeout    86400;
        # }
    }

10. VSCode连接jupyter

由于jupyterlab可以运行在本地指定端口,所以可以通过IP和端口在客户自己浏览器进行远程开发(保证远程服务器的jupyter lab开机自启服务),这在局域网内很方便,但是对于没有公网IP的话,就无法使用此功能
好在VSCode可以直接打开远程jupyter,具体操作如下

  • 在客户本地机器安装Remote Development三件套插件,然后选择Remote-SSH: Connect to host,可以在本地提前创建配置文件(C:\Users\zfb\.ssh\config或者C:\ProgramData\ssh\ssh_config),内容类似:
# 第一个远程机器
Host mylab
    HostName 54.33.135.211
    Port 22
    User ubuntu
  • 根据提示输入远程服务器的密码即可连接成功,然后在远程服务器安装PythonPylanceIntelliCode这三个插件,打开远程服务器的文件夹,创建一个扩展名为ipynb的文件,然后VSCode会自动提示选择Python版本(既可以选择系统的,也可以根据路径选择某个虚拟环境里面的),接着VSCode会自动连接Kernel,用户可以在右上角查看当前Kernel的状态或者切换Kernel

11. ssh连接jupyter在本地打开

在浏览器使用远程ip:port的方法,则服务器必须有公网,而且还费流量,另一种方法,ssh连接,然后端口映射
服务器1:处于内网,已安装frpc,用户名为zfb,已安装配置好jupyterlab,运行在8888端口
云主机2:处于公网,ip为56.78.12.34,已安装frps,用户名为ubuntu,仅用于服务器的内网穿透,端口7001为服务器1提供ssh转发
执行以下命令,把用户3的电脑的本地端口8080绑定到服务器1的端口8888:
ssh -p 7001 -NL localhost:8080:localhost:8888 [email protected]
此时在用户3的本机打开网址http://127.0.0.1:8080即可访问服务器1的jupyterlab

12. matplotlib安装

首先在虚拟环境jupyter安装matplotlib库和ipympl库,后者用于显示可交互图形

# 激活虚拟环境jupyter
source jupyter/bin/activate
# 在虚拟环境jupyter安装matlab_kernel
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple matplotlib ipympl

重新打开浏览器会提示rebuild,点击确定。等待build成功然后点击reload即可正常使用此插件,如下代码

%matplotlib widget
import pandas as pd
import numpy as np
import matplotlib
from matplotlib import pyplot as plt

ts = pd.Series(np.random.randn(1000), index=pd.date_range('1/1/2000', periods=1000))
ts = ts.cumsum()

df = pd.DataFrame(np.random.randn(1000, 4), index=ts.index,
                  columns=['A', 'B', 'C', 'D'])
df = df.cumsum()
df.plot()
plt.legend(loc='best')
plt.title('我是中文')

如果中文乱码,则纠正中文乱码

13. 使用plotly显示python程序绘制的图片

使用方法见官网,python的使用不需要key和用户名,直接用就行

14. 使用plotly显示matlab的图片

详细使用方法见官网教程。注册plotly的chart-studio账号,然后在个人账户的setting点击api keys,选择Regenerate key,记住这个key和自己的用户名。然后下载压缩包并解压,打开matlab,输入

>> cd ~/plotly-graphing-library-for-matlab-master/
>> plotlysetup('DemoAccount', 'lr1c44zw81')  % 回车,剩下的内容都是自动执行
Adding Plotly to MATLAB toolbox directory ...  Done
Welcome to Plotly! If you are new to Plotly please enter: >> plotlyhelp to get started!

此时会创建文件~/.plotly/.credentials,里面已经保存用户名和key(注意该用户需要有toolbox的写入权限)
然后在jupyterlab写:

[X,Y,Z] = peaks;
contour(X,Y,Z,20);
% 个人用户还是用离线模式吧,否则只能创建100个图,还必须是公开分享
getplotlyoffline('https://cdn.plot.ly/plotly-latest.min.js')
fig2plotly(gcf, 'offline', true)

该命令会在当前目录生成一个html文件,双击打开即可
注意: 如果发现在其他目录无法使用fig2plotly函数,则可能是上一步骤,将plotly添加到Matlab工具箱出现了问题。可以自己手动将其复制到指定工具箱路径,或者直接把plotly-graphing-library-for-matlab-master文件夹的绝对路径添加到Matlab PATH

15. 使用plotly绘制matlab的包含ColorBar的图片

如果正在使用新版Matlab(R2019a以后),在.m文件中如果使用colorbar函数,则在调用plotly时候可能会遇到报错

Insufficient number of outputs from right hand side of equal sign to satisfy assignment.

Error in findColorbarAxis (line 8)
colorbarAxis = obj.State.Axis(colorbarAxisIndex).Handle;

Error in plotlyfig/update (line 557)
                colorbarAxis = findColorbarAxis(obj, handle(cols(c)));

Error in plotlyfig (line 208)
                obj.update;

Error in fig2plotly (line 44)
p = plotlyfig(varargin{:});

参考链接,于是打开文件findColorBarAxis.m

# 若Matlab的Plotly工具箱安装位置为/home/Polyspace/R2020a/toolbox/plotly
sudo vi /home/Polyspace/R2020a/toolbox/plotly/plotlyfig_auz/helpers/findColorBarAxis.m

整个文件内容替换为如下:

function colorbarAxis = findColorbarAxis(obj,colorbarHandle)
if isHG2
    colorbarAxisIndex = find(arrayfun(@(x)(isequal(getappdata(x.Handle,'ColorbarPeerHandle'),colorbarHandle)),obj.State.Axis));
    % If the above returns empty then we are on a more recent Matlab
    % release where the appdata entry is called LayoutPeers
    if isempty(colorbarAxisIndex)
        colorbarAxisIndex = find(arrayfun(@(x)(isequal(getappdata(x.Handle,'LayoutPeers'),colorbarHandle)),obj.State.Axis));
    end
else
    colorbarAxisIndex = find(arrayfun(@(x)(isequal(getappdata(x.Handle,'LegendColorbarInnerList'),colorbarHandle) + ...
        isequal(getappdata(x.Handle,'LegendColorbarOuterList'),colorbarHandle)),obj.State.Axis));
end
colorbarAxis = obj.State.Axis(colorbarAxisIndex).Handle;
end
01-06 21:37