-
Notifications
You must be signed in to change notification settings - Fork 241
Distributed operation
Vasily Evseenko edited this page Sep 22, 2024
·
1 revision
WFB-NG supports distributed operation with multiple RX or (RX+TX) antennas on different hosts connected by IP network. You can use multiple hotspots connected via ethernet and get "pseudo-cellular" network. For example you can gather video stream from moving transmitter in the forest without any wifi roaming overhead.
- On the central node, the cluster topology should be set in
/etc/wifibroadcast.cfg
. Seemaster.cfg
for reference. You should specify node addresses, wifi card names and ssh keys (optional). - Only
wfb_rx
andwfb_tx
binaries should be installed on the cluster nodes (except for the central master node) (for openwrt, wfb-ng and bash packages should be installed). Encryption keys and /etc/wifibroadcast.cfg should not be installed on them. - The cluster can be initialized in two modes: automatic (ssh) and manual
a. For automatic mode, you only need to set the ssh key and user (usually root) under which ssh will work on cluster nodes without entering a password. If you use
ssh_agent
, then you do not need to add the key, but make sure that the agent runs under root sincewfb-server
needs to be launched under it (this is required to create an IP tunnel) b. For manual mode, you will need to generate wfb-ng initialization scripts for the cluster nodes. This is done viawfb-server --profiles gs --gen-init X.X.X.X
where X.X.X.X is the IP address of the node. The resulting script needs to be copied to the node and run (for openwrt, save it as/usr/sbin/wfb-ng.sh
, set executable permissions and do/etc/init.d/wfb-ng enable
and/etc/init.d/wfb-ng start
). - Run wfb-ng in cluster mode on the central node:
wfb-server --profiles gs --cluster ssh
orwfb-server --profiles gs --cluster manual
. There is no cluster support in systemd scripts for wfb-ng yet, but you can add it yourself (this will only work for manual or ssh mode with keys) - Run
wfb-cli gs
on the central node and check that everything works. There will be very few differences from the single mode: instead of short antenna names likeX:Y
there should beADDR:X:Y
, whereADDR
is the node IP address in hex format.
Differences between ssh and manual modes:
- In ssh mode, if an error occurs on one of the nodes (for example, something did not start or broke), the entire cluster will stop. Also, all services on the nodes will automatically stop if the central node is turned off.
- In manual mode, services on the cluster nodes work independently of the central node.
- use only IP addresses (not FQDN names) for nodes! There is no separate check for this yet and this will lead to undefined behavior in the cluster logic!
- In manual mode you need to regenerate init scripts for each node in case of any changes in wifibroadcast.cfg on the master node, because internal allocation of UDP ports will be different ant this will lead to undefined behavior in the cluster logic!