shadowsocks-manager

A web-based Shadowsocks management tool.

Features:

Central user management
Heartbeat on Shadowsocks ports(users)
Shadowsocks multi-user API
Shadowsocks node cluster
Statistic for network traffic usage
Scheduled jobs
name.com API
Auto-creating DNS records
Production deployment ready
How’s the Shadowsocks supported:
- libev edition:
  - Full functional.
  - No builtin service manager, you need to install it and start the service by yourself.
- python edition:
  - Lacks the collection of traffic statistics.
  - Lacks the ability to test user port creation status.
  - ~~Pre-installed, and have a builtin service manager.~~
  - No builtin service manager, you need to install it and start the service by yourself.

Code in Python, base on Django, Django REST framework, Celery, and SQLite.

The development status can be found at: project home.

Screenshots

Shadowsocks Node List: Home › Shadowsocks › Shadowsocks Nodes

Add Shadowsocks Node: Home › Shadowsocks › Shadowsocks Nodes

Add Shadowsocks Account: Home › Shadowsocks › Shadowsocks Accounts

1. Requirements

Python 2.7, Python 3.x
Django 1.11.x, Django 3.x
Docker
Shadowsocks-libev 3.2.0 for Linux (multi-user API is required)

2. Install

This project is a part of an entire VPN solution, which includes the Shadowsocks server and Shadowsocks manager. The Shadowsocks server serves the traffic, the Shadowsocks manager serves the users and the traffic statistics. The solution is designed to be deployed in the AWS cloud. If you are looking for such a solution, you can refer to the repo aws-cfn-vpn. With aws-cfn-vpn, you can deploy the entire solution with a few commands.

2.1. Dependencies

Assume you have installed the Docker on your host.

2.2. Manual installation

# create a docker network
docker network create ssm-network

# run memcached, used by django cache
docker run -d --network ssm-network --name ssm-memcached memcached

# run rabbitmq, used by celery
docker run -d --network ssm-network --name ssm-rabbitmq rabbitmq

# create a directory to store the data, it will be mounted to the shadowsocks-manager container
mkdir -p ~/ssm-volume

# run the shadowsocks-manager
docker run -d -p 80:80 -v ~/ssm-volume:/var/local/ssm \
    --network ssm-network --name ssm alexzhangs/shadowsocks-manager \
    -e SSM_SECRET_KEY=yourkey -e SSM_DEBUG=False \
    -e SSM_MEMCACHED_HOST=ssm-memcached -e SSM_RABBITMQ_HOST=ssm-rabbitmq \
    -u admin -p yourpassword -M admin@yourdomain.com

2.3. Install with script

git clone https://github.com/alexzhangs/shadowsocks-manager
bash shadowsocks-manager/install.sh -u admin -p yourpassword -M admin@yourdomain.com

2.4. Verify the installation

If all go smoothly, the shadowsocks-manager services should have been all started. Open the web admin console in a web browser, and log on with the admin user.

Use:

http://<your_server_ip>/admin
or 
http://localhost/admin

If goes well, then congratulations! The installation has succeeded.

3. Using shadowsocks-manager

Shadowsocks server

First, you need to have a Shadowsocks server with the multi-user API enabled.

Install it with docker on the same server with shadowsocks-manager.

 # run shadowsocks-libev
 MGR_PORT=6001
 SS_PORTS=8381-8480
 ENCRYPT=aes-256-cfb
 docker run -d -p $SS_PORTS:$SS_PORTS/UDP -p $SS_PORTS:$SS_PORTS \
     --network ssm-network --name ssm-ss-libev shadowsocks/shadowsocks-libev:edge \
     ss-manager --manager-address 0.0.0.0:$MGR_PORT \
     --executable /usr/local/bin/ss-server -m $ENCRYPT -s 0.0.0.0 -u

 # Use below command to get the private IP address of the shadowsocks-libev container for later configuration.
 docker inspect ssm-ss-libev | grep IPAddress

Add Shadowsocks server to shadowsocks-manager

Add the Shadowsocks server as a Node of shadowsocks-manager from web admin console: Home › Shadowsocks › Shadowsocks Nodes.
Create users(ports) and assign Shadowsocks Node

Create users from web admin console: Home › Shadowsocks › Shadowsocks Accounts and assign the existing nodes to them.

After a few seconds, the created user ports should be available to your Shadowsocks client.

4. Sendmail (Optional)

sendmail is used to send account notification Email, it should be configured on the same server with shadowsocks-manager.

About how to configure sendmail client to use AWS SES as SMTP server on AWS EC2 instance, refer to repo aws-ec2-ses.

On macOS, refer to repo macos-aws-ses.

NOTE: This dependency needs the manual setup anyway, it is not handled by any installation script.

5. Differences from the alternation: shadowsocks/shadowsocks-manager

This repo Do’s:

Serve as a nonprofit business model.
Have central user management for multi nodes.
Collect traffic statistic that can be viewed by account, node, and period.
Show the existence and accessibility of ports in the admin.
Handle the DNS records if using Name.com as nameserver.

This repo Don’ts:

Handle self-serviced user registration.
Handle bill or payment.
Need to run an additional agent on each Shadowsocks server.

6. Some differences between the Shadowsocks Python edition (2.8.2) and libev edition

Version status for the Shadowsocks Python edition:

pypi: 2.8.2
github: 3.0.0

Although the Shadowsocks Python edition supports the multi-user API, but it doesn’t fit this project, here’s why:

The python edition code and doc seem to be out of maintenance due to some reason. If you really need this you probably need to fork it and make your own.
They are having different service process names and CLI interfaces which introduces the complexity of installation.
The Python edition lacks the list commands. A pull request was opened years ago but never merged.
The Python edition’s stat command has a very different way to use, I didn’t figure the usage syntax out by looking into the code.
The Python edition’s ping command returns a simple string pong rather than a list of ports.
The Python edition’s ping command has to be sent as the syntax: ping:{} in order to work if tested with nc. It caused by the tailing newline: ping\n.

So either you get some change on your own or stick with the libev edition.

7. Known Issues

DNS records matching for Node may not be accurate on macOS. For unknown reason sometimes DNS query returns only one IP address while multiple IP addresses were configured for the domain.

The Shadowsocks Python edition’s ssserver won’t start on macOS. The error message is like:

 $ ssserver -k passw0rd
 WARNING: /Users/***/.pyenv/versions/3.12.0/envs/ssm-3.12/bin/python3.12 is loading libcrypto in an unsafe way
 Abort trap: 6

Solution: Link the homebrew openssl library to the system library.

 sudo ln -s /opt/homebrew/opt/openssl/lib/libcrypto.dylib /usr/local/lib/
 sudo ln -s /opt/homebrew/opt/openssl/lib/libssl.dylib /usr/local/lib/

Install the project by source with pip under Python 2.7 get error:

 python --version
 Python 2.7.18

 pip install .

Error message:

 ...
 Collecting pyyaml
 Downloading PyYAML-5.4.1.tar.gz (175 kB)
     |████████████████████████████████| 175 kB 392 kB/s 
 Installing build dependencies ... done
 Getting requirements to build wheel ... error
 ERROR: Command errored out with exit status 1:
 ...
     raise AttributeError, attr
 AttributeError: cython_sources
 ...

Solution:

 pip install setuptools wheel
 pip install --no-build-isolation .

8. Development

The development of this project requires Python 3.x.

However, the installation of the project is compatible with both Python 2.7 and 3.x. To keep the compatibility is difficult, but it’s kept due to the historical reason. The following files are kept only for installing the source distribution of the PyPI package under Python 2.7:

setup.py
setup.cfg

8.1. Development Environment Setup

Install the dependencies

 # run memcached, used by django cache
 docker run -d -p 11211:11211 --name ssm-dev-memcached memcached

 # run rabbitmq, used by celery
 docker run -d -p 5672:5672 --name ssm-dev-rabbitmq rabbitmq

 # run shadowsocks-libev, simulate localhost node
 MGR_PORT=6001 SS_PORTS=8381-8479 ENCRYPT=aes-256-cfb
 docker run -d -p 127.0.0.1:$MGR_PORT:$MGR_PORT/UDP \
     -p 127.0.0.1:$SS_PORTS:$SS_PORTS/UDP -p 127.0.0.1:$SS_PORTS:$SS_PORTS \
     --name ssm-dev-ss-libev-localhost shadowsocks/shadowsocks-libev:edge \
     ss-manager --manager-address 0.0.0.0:$MGR_PORT \
     --executable /usr/local/bin/ss-server -m $ENCRYPT -s 0.0.0.0 -u

 # get the private IP address of the host, double check the result, it might not be correct
 PRIVATE_IP=$(ipconfig getifaddr en0 2>/dev/null || hostname -i | awk '{print $1}' 2>/dev/null)
 echo "PRIVATE_IP=$PRIVATE_IP"

 # run shadowsocks-libev, simulate private IP node
 MGR_PORT=6002 SS_PORTS=8381-8479 ENCRYPT=aes-256-cfb
 docker run -d -p $PRIVATE_IP:$MGR_PORT:$MGR_PORT/UDP \
     -p $PRIVATE_IP:$SS_PORTS:$SS_PORTS/UDP -p $PRIVATE_IP:$SS_PORTS:$SS_PORTS \
     --name ssm-dev-ss-libev-private shadowsocks/shadowsocks-libev:edge \
     ss-manager --manager-address 0.0.0.0:$MGR_PORT \
     --executable /usr/local/bin/ss-server -m $ENCRYPT -s 0.0.0.0 -u
        
 # run shadowsocks-libev, simulate public IP node
 MGR_PORT=6003 SS_PORTS=8480 ENCRYPT=aes-256-cfb
 docker run -d -p $PRIVATE_IP:$MGR_PORT:$MGR_PORT/UDP \
     -p $PRIVATE_IP:$SS_PORTS:$SS_PORTS/UDP -p $PRIVATE_IP:$SS_PORTS:$SS_PORTS \
     --name ssm-dev-ss-libev-public shadowsocks/shadowsocks-libev:edge \
     ss-manager --manager-address 0.0.0.0:$MGR_PORT \
     --executable /usr/local/bin/ss-server -m $ENCRYPT -s 0.0.0.0 -u

Link the project code in your workspace to the Python environment
```
 cd shadowsocks-manager
 pip install -e .
```
Set the SSM_DATA_HOME environment variable (optional, default is Django root directory)
```
 export SSM_DATA_HOME=~/.ssm-dev-data
```

Configure the shadowsocks-manager

 ssm-setup -c -m -l -u admin -p yourpassword

Run the development server

Make sure the memcached and rabbitmq are running.
```
 ssm-dev-start
```
Stop the development server
```
 ssm-dev-stop
```
Test the Django code

Make sure the memcached and rabbitmq are running, and also the 3 shadowsocks-libev node are running.
```
 ssm-test -t

 # or use the django test command directly for more options
 ssm-manage test --no-input -v 2
```
Test the Django code with coverage
```
 pip install coverage
 ssm-test -c
```
Upload the coverage report to codecov

Make sure the CODECOV_TOKEN is exported in the environment before uploading.
```
 pip install codecov-cli
 ssm-test -u
```
Test the Github workflows locally
```
 brew install act
 act -j test
```

Build the PyPI package

 pip install build

 # build source and binary distribution, equivalent to `python setup.py sdist bdist_wheel`
 # universal wheel is enabled in the pyproject.toml to make the wheel compatible with both Python 2 and 3
 python -m build

Upload the PyPI package

Set the ~/.pypirc file with the API token from the TestPyPI and PyPI before uploading.

 pip install twine

 # upload the package to TestPyPI
 python -m twine upload --repository testpypi dist/*

 # upload the package to PyPI
 python -m twine upload dist/*

Test the PyPI package

 # install the package from TestPyPI
 # --no-deps is used to skip installing dependencies for the TestPyPI environment
 pip install -i https://test.pypi.org/simple/ --no-deps shadowsocks-manager

 # install the package from PyPI
 # --no-binary is used to force building the package from the source
 # --use-pep517 is used together to make sure the PEP 517 is tested
 pip install --no-binary shadowsocks-manager --use-pep517 shadowsocks-manager

Build the docker image

 docker build -t alexzhangs/shadowsocks-manager .

8.2. CI/CD

Github Actions is currently used for the CI/CD. Travis CI is removed due to the limitation of the free plan.

The CI/CD workflows are defined in the .github/workflows directory.

ci-unittest.yml: Run the unit tests.
ci-testpypi.yml: Build and upload the package to TestPyPI.
ci-pypi.yml: Build and upload the package to PyPI. It can be triggered by the tag: ci-pypi or ci-pypi-(major|minor|patch|suffx).
ci-docker.yml: Build and push the docker image to Docker Hub. It can be triggered by the Github release.

9. Troubleshooting

Check the logs

 # supervisor
 cat /var/log/supervisor/supervisord.log

 # uWSGI
 cat /var/log/ssm-uwsgi.log

 # Celery
 cat /var/log/ssm-cerlery*

Check the services

 # nginx
 service nginx {status|start|stop|reload}

 # supervisor
 service supervisord {status|start|stop|restart}
 supervisorctl reload
 supervisorctl start all

 # uWSGI
 supervisorctl start ssm-uwsgi

 # Celery
 supervisorctl start ssm-celery-worker
 supervisorctl start ssm-celery-beat

 # Docker
 docker ps

Check the listening ports and processes (Linux)

 # TCP
 netstat -tanp

 # UDP
 netstat -uanp

Check the listening ports (MacOS)

 # TCP
 netstat -anp tcp

 # UDP
 netstat -anp udp

 # find the process by port
 lsof -i :80 -P