The big picture


One of the RaspberryPi's is assigned as the "leader", and this hosts most of the services: web server, MQTT broker, database, etc. It also sends commands to any "workers". Together, the leader and all the workers are called a "cluster". See more information about clusters here


The RaspberryPis communicate through the local network (in more advanced cases, this network is hosted on one of the leader). Each Rpi broadcasts it's hostname through mDNS, and the leader also broadcasts an alias (default: `pioreactor.local`). The use of mDNS is fine if devices support it: currently Apple, Linux and Windows 10 products do. 

The PioreactorUI is hosted on the leader and exposes port 80, hence users on the local network can view the website at `pioreactor.local`. 

Jobs and threads

A job (labelled activity on the PioreactorUI) is a Python process that controls some element of the Pioreactor. For example, the background job controls whether the stirring should be turned on, how fast it should spin, and reports its state to the cluster leader. Communication between the job and the leader / user is through MQTT. Specifically, the job spawns a thread that subscribes to the topic pioreactor/<unit>/<experiment>/<job name>/<attribute>/set. For example, to change the stirring duty cycle to 50 on unit1, we send a message to pioreactor/unit1/my experiment/stirring/set/duty_cycle with payload 50. The job knows how to update it self. Similarly, to retrieve the current value of a job attribute (like $state), one can listen to the topic: pioreactor/<unit>/<experiment>/<job name>/<attribute>.

Another thread on the job is used to publish messages from the job back to the MQTT broker (on the leader).

Other threads may be used for timing, sensors, etc.



← Back to all documentation