Many modbus devices support only one or very few clients. This proxy acts as a bridge between the client and the modbus device. It can be seen as a layer 7 reverse proxy. This allows multiple clients to communicate with the same modbus device.
When multiple clients are connected, cross messages are avoided by serializing communication on a first come first served REQ/REP basis.
From within your favorite python 3 environment type:
$ pip install modbus-proxy
Note: On some systems pip
points to a python 2 installation.
You might need to use pip3
command instead.
Additionally, if you want logging configuration:
- YAML:
pip install modbus-proxy[yaml]
(see below) - TOML:
pip install modbus-proxy[toml]
(see below)
First, you will need write a configuration file where you specify for each modbus device you which to control:
- modbus connection (the modbus device url)
- listen interface (to which url your clients should connect)
Configuration files can be written in YAML (.yml or .yaml) or TOML (.toml).
Suppose you have a PLC modbus device listening on plc1.acme.org:502 and you want your clients to connect to your machine on port 9000. A YAML configuration would look like this:
devices:
- modbus:
url: plc1.acme.org:502 # device url (mandatory)
timeout: 10 # communication timeout (s) (optional, default: 10)
connection_time: 0.1 # delay after connection (s) (optional, default: 0)
listen:
bind: 0:9000 # listening address (mandatory)
unit_id_remapping: # remap/forward unit IDs (optional, empty by default)
1: 0
Assuming you saved this file as modbus-config.yml
, start the server with:
$ modbus-proxy -c ./modbus-config.yml
Now, instead of connecting your client(s) to plc1.acme.org:502
you just need to
tell them to connect to *machine*:9000
(where machine is the host where
modbus-proxy is running).
Note that the server is capable of handling multiple modbus devices. Here is a configuration example for 2 devices:
devices:
- modbus:
url: plc1.acme.org:502
listen:
bind: 0:9000
- modbus:
url: plc2.acme.org:502
listen:
bind: 0:9001
If you have a single modbus device, you can avoid writting a configuration file by providing all arguments in the command line:
modbus-proxy -b tcp://0:9000 --modbus tcp://plc1.acme.org:502
(hint: run modbus-proxy --help
to see all available options)
You can also forward one unit ID to another whilst proxying. This is handy if the target modbus server has a unit on an index that is not supported by one of your clients.
devices:
- modbus: ... # see above.
listen: ... # see above.
unit_id_remapping:
1: 0
The above forwards requests to unit ID 1 to your modbus-proxy server to unit ID 0 on the actual modbus server.
Note that the reverse also applies: if you forward unit ID 1 to unit ID 0, all responses coming from unit 0 will look as if they are coming from 1, so this may pose problems if you want to use unit ID 0 for some clients and unit ID 1 for others (use unit ID 1 for all in that case).
To run the examples you will need to have
umodbus installed (do it
with pip install umodbus
).
Start the simple_tcp_server.py
(this will simulate an actual modbus hardware):
$ python examples/simple_tcp_server.py -b :5020
You can run the example client just to be sure direct communication works:
$ python examples/simple_tcp_client.py -a 0:5020
holding registers: [1, 2, 3, 4]
Now for the real test:
Start a modbus-proxy bridge server with:
$ modbus-proxy -b tcp://:9000 --modbus tcp://:5020
Finally run a the example client but now address the proxy instead of the server (notice we are now using port 9000 and not 5020):
$ python examples/simple_tcp_client.py -a 0:9000
holding registers: [1, 2, 3, 4]
- move the config file to a location you can remember, for example: to
/usr/lib/mproxy-conf.yaml
- go to
/etc/systemd/system/
- use nano or any other text editor of your choice to create a service file
mproxy.service
- the file should contain the following information:
[Unit]
Description=Modbus-Proxy
After=network.target
[Service]
Type=simple
Restart=always
ExecStart = modbus-proxy -c ./usr/lib/mproxy-conf.yaml
[Install]
WantedBy=multi-user.target
run systemctl daemon-reload
systemctl enable mproxy.service
systemctl start mproxy.service
The file names given here are examples, you can choose other names, if you wish.
This project ships with a basic Dockerfile which you can use as a base to launch modbus-proxy inside a docker container.
First, build the docker image with:
$ docker build -t modbus-proxy .
To bridge a single modbus device without needing a configuration file is straight forward:
$ docker run -d -p 5020:502 modbus-proxy -b tcp://0:502 --modbus tcp://plc1.acme.org:502
Now you should be able to access your modbus device through the modbus-proxy by
connecting your client(s) to <your-hostname/ip>:5020
.
If, instead, you want to use a configuration file, you must mount the file so it is visible by the container.
Assuming you have prepared a conf.yml
in the current directory:
devices:
- modbus:
url: plc1.acme.org:502
listen:
bind: 0:502
Here is an example of how to run the container:
docker run -p 5020:502 -v $PWD/conf.yml:/config/modbus-proxy.yml modbus-proxy
By default the Dockerfile will run modbus-proxy -c /config/modbus-proxy.yml
so
if your mounting that volume you don't need to pass any arguments.
Note that for each modbus device you add in the configuration file you need
to publish the corresponding bind port on the host
(-p <host port>:<container port>
argument).
Logging configuration can be added to the configuration file by adding a new logging
keyword.
The logging configuration will be passed to logging.config.dictConfig() so the file contents must obey the Configuration dictionary schema.
Here is a YAML example:
devices:
- modbus:
url: plc1.acme.org:502
listen:
bind: 0:9000
logging:
version: 1
formatters:
standard:
format: "%(asctime)s %(levelname)8s %(name)s: %(message)s"
handlers:
console:
class: logging.StreamHandler
formatter: standard
root:
handlers: ['console']
level: DEBUG
Logging configuration file.
If a relative path is given, it is relative to the current working directory.
If a .conf
or .ini
file is given, it is passed directly to
logging.config.fileConfig() so the file contents must
obey the
Configuration file format.
A simple logging configuration (also available at log.conf) which mimics the default configuration looks like this:
[formatters]
keys=standard
[handlers]
keys=console
[loggers]
keys=root
[formatter_standard]
format=%(asctime)s %(levelname)8s %(name)s: %(message)s
[handler_console]
class=StreamHandler
formatter=standard
[logger_root]
level=INFO
handlers=console
A more verbose example logging with a rotating file handler: log-verbose.conf
The same example above (also available at log.yml) can be achieved in YAML with:
version: 1
formatters:
standard:
format: "%(asctime)s %(levelname)8s %(name)s: %(message)s"
handlers:
console:
class: logging.StreamHandler
formatter: standard
root:
handlers: ['console']
level: DEBUG
Based on the existing Dockerfile an example Kubernetes manifest has been created and is located at examples/kubernetes.
ConfigMap
for modbus-proxy config:examples/kubernetes/10_modbus-proxy.configmap.yaml
After the adjustment of the configmap, the manifest could get applied to your Kubernetes Cluster:
# adjust config
vim examples/kubernetes/10_modbus-proxy.configmap.yaml
# apply to cluster
kubectl apply -f examples/kubernetes
- Tiago Coutinho coutinhotiago@gmail.com
None yet. Why not be the first?