The system that starts the build is the master, while the workers are the clients.
If you want to use the master system to compile as well, apply the client configuration to it.
You can set each host that will start a build as master.
Keep in mind that only compiles are distributed; preprocessing and linking still takes place on the master system.
On any system install distcc:
pacman -S distccEnable distcc in /etc/makepkg.conf by removing the prepended !.
Change DISTCC_HOSTS to a space-delimited list of the clients on your network.
If you want to use the master system, be sure to put it's IP address in there too.
Referred guides suggest to avoid the usage of "localhost" or "127.0.0.1", I don't know why.
sudo sed -i 's/!distcc/distcc/' /etc/makepkg.conf
sudo sed -i 's,#DISTCC_HOSTS="",DISTCC_HOSTS="10.30.5.9/20",' /etc/makepkg.confThe setup above uses just the Ryzen server, limiting to 20 cores (max. 32).
Distribution status can be displayed with distccmon-text 2 (the number is refresh time, in seconds).
NOTE: master nodes cannot have -march=native in CFLAGS or CXXFLAGS!
Specific commands can be distributed with slaves with the pump command, e.g.:
export DISTCC_HOSTS="10.30.5.9/30,cpp,lzo localhost" # <- can be placed in .zshrc
pump make -j40 CC=distcc CXX=distccSpecify accepted connections in /etc/conf.d/distccd:
DISTCC_ARGS="--allow-private --log-level info --log-stderr --log-file /tmp/distccd.log"
Specific IPs can be specified, e.g. with --allow 127.0.0.1 --allow 10.30.5.0/24.
Then enable and start distccd.service.
Each slave can be monitored by the HTTP statistics server at port 3633 by adding --stats to DISTCC_ARGS.