Skip to main content

Networking tools

info

Previously, Raspberry Pis used a file called wpa_supplicant.conf to handle wifi network connections. This is not longer the case. Modern Raspberry Pi software, including the Pioreactor, doesn't use wpa_supplicant.conf at all. Instead, the tool nmcli replaces it. Don't follow tutorials that use wpa_supplicant.conf.

General networking tools​

The main "entry point" for networking on the Pioreactor (and Raspberry Pi's in general) is the tool nmcli. This controls discovering networks, connecting to them, and editing connections. For example, running nmcli con will diplay a list of possible networking connections, and connected ones in green.

Connecting to a wifi network using nmcli​

You can first discover all the visible networks with:

sudo nmcli d wifi list

If you see your network on the list, then run:

sudo nmcli device wifi connect <ssid name> password <ssid password> ifname wlan0

(If you get a "Can't find" error, try running the above list command again.)

Connecting to multiple networks simultaneously​

First, some terminology and understanding for us: a computer, like a Pi, has networking interfaces. For example, the onboard wifi on RPi’s is one such interface. The larger, model Bs, have an ethernet connector, which is another interface. Each interface can connect to 0 or 1 networks.

In order to connect to multiple networks, you need a Raspberry Pi with multiple interfaces. You can add interfaces with devices like a USB wifi device.

To see what your current interfaces are doing, use:

nmcli device

The right hand side shows your connected interfaces. For example, if have a ethernet connection to a router, you'll see an eth0 connection active. If you have an additional wifi hardware device attached to your Pi, you'll see wlan1.

tip

To connect to another wifi network using an external wlan1 interface, use:

sudo nmcli device wifi connect <ssid name> password <ssid password> ifname wlan1

If your leader is connected to multiple networks A and B, and you access the UI over network A, but your workers access over network B, the leader is in the position called a "gateway". You'll need to make some configuration changes to your cluster. Since the leader is attached to networks A and B, it has two IPs (use hostname -I to see all the ips), let's call them ipA and ipB respectively.

  1. Access the configuration in the UI.
  2. In the shared config.ini, edit the [mqtt] broker field to be equal to ipA. Also edit the leader_address to be ipB.
  3. In each of the workers (including the leader's config.ini, if it's a worker), add the following: 
    [mqtt]
    broker=ipB

You may need to restart your cluster for these to take full effect.

Starting the Pioreactor local access point using nmcli​

Note that you can only have one network connection for each network interface. So if you have only one wifi device (by default, RPi's only have one wifi device on them, which implies only one wifi interface), the following will replace any existing wifi connection.

sudo nmcli con up PioreactorAP

If have a second wifi device with interface wlan1, you can change the interface of the local access point using:

sudo nmcli con modify PioreactorAP ifname wlan1
sudo nmcli con up PioreactorAP

Changing leader_address and MQTT broker in your config.ini​

You may need to change the leader_address in your config.ini: this is the address where workers will communicate with your leader.

Likewise, you may need to change the mqtt broker in your config.ini for the same reason. It may be required to change it in the unit specific config.inis.

config_worker01.ini
[mqtt]
broker=leader.local

Changing web UI port from 80 to something else​

To change the web UI port from the default of 80, following these instructions:

note

This will change the web server port for all leaders and workers. There is currently not a way to have different ports on different Pioreactors. See this issue.

  1. SSH into each Pioreactor in your cluster, leader and workers.

  2. We'll edit the lighttpd configuration first:

    sudo nano /etc/lighttpd/lighttpd.conf

    and find the line starting with server.port. Change this to something else (preferably not a value below 1024). Save and exit.

  3. Restart lighttpd with:

    sudo systemctl restart lighttpd.service

    You should be able to access the web UI only on http://leader name:new port, for example: http://leader.local:8080

  4. In your configuration, under the [ui] section, change the port option from 80 to your new port value. Save. You may need to restart your cluster for this new port to propagate to all Pioreactors correctly.

Connecting to eduroam​

This is a work-in-progress, but here's one example.

Common questions​

My Pioreactor activities start very slowly from the UI​

It is possible that your mDNS is being blocked or restricted. If possible, log into your router and enable settings that allow "multicast", or "mDNS", or "IGMP proxing". If you are using a the builtin local access point, this solution is not the correct one.

Also, if you are able to provide a permanent IPv4 address to your leader, you can get a significant performance boost by setting the configuration some config parameters to the IPv4 address.

[cluster.topology]
# below is an example, your IPv4 may differ:
leader_hostname=... # your leader's hostname
leader_address=192.168.0.3

[mqtt]
broker_address=192.168.0.3

I am using static IPv4s, how do I set this up?​

You can put worker ipv4s in the shared config.ini file under cluster_addresses section. For example, for worker01:

[cluster.addresses]
worker01=192.1.1.1