Home · Getting Started · Troubleshooting · About
Follow these steps in order, from an empty desk to 311 trained models. Each page starts with its goal and where to run it, and ends with a check that the step worked before you move on.
| Step | Page | At the end you have | Run on |
|---|---|---|---|
| 1 | Hardware List | Everything bought and assembled, with enough power and cooling | – |
| 2 | Install OS | Five cards flashed with Raspberry Pi OS Lite, each node reachable over SSH | Your computer |
| 3 | The Pi Network | Internet access, fixed addresses, nodes reachable by name, master able to log in everywhere |
Your computer, master |
| 4 | Cluster Commands | One command to run or copy anything on every node | master |
| 5 | Cluster Security | Key-only logins, fail2ban and a firewall | master |
| 6 | Single Node Setup | Java on every node; HDFS working on master alone |
master |
| 7 | Cluster Setup | HDFS across four workers, with two copies of every block | master |
| 8 | Spark | Spark with one 4-core executor per worker, reading from HDFS | master |
| 9 | Workload | 311 models trained, one per worker at a time, and a benchmark | master |
Stuck? See Troubleshooting.
The steps link to these when you need them. They’re also handy once the cluster is running.
| Page | Covers |
|---|---|
| Software | Raspberry Pi OS, which edition to use, and the Java, Hadoop and Spark versions the guide pins |
| Maintenance | Updating packages and firmware on every node; backing up, restoring and cloning cards |