Raspberry Pi clusters have been a favourite undertaking of homelabbers and distributed computing lovers because the platform first launched over a decade in the past, and for good cause. For a low value this {hardware} makes it potential to experiment with parallel computing — one thing that in any other case isn’t simply accessible with out a number of time, cash, and {hardware}. That is much more true with the compute modules, as their dimension and price makes some staggering builds potential like this cluster sporting 112 GB of RAM.
The undertaking relies on the NanoCluster, a board that may maintain seven compute modules in a type issue which, as [Christian] describes it, is concerning the dimension of a espresso mug. Meaning not solely does it have a reasonably staggering quantity of RAM but in addition 28 processor cores to work with. Placing the {hardware} collectively is the simple half, although; [Christian] wished to seek out absolutely the simplest way of managing a system like this and selected gitops, which is a technique of sustaining a server the place the specified system state is saved in Git, and automation constantly ensures the working surroundings on the {hardware} matches what’s within the repository.
For this cluster, it implies that the nodes themselves will be swapped out and in, with new nodes routinely receiving directions after which configuring themselves routinely. Updates and adjustments made on Git are pushed to the nodes routinely as nicely and there’s not a lot that must be performed manually in any respect. In a lot the identical method that immutable Linux distributions transfer all the trouble of administering a system to one thing like a config file, instruments like gitops do the identical for servers and clusters like this, and it’s price trying out [Christian]’s undertaking to get an thought of simply how simple it may be now.
