>Tutorial: Troubleshooting

causeparallelix-node verify reports 'ollama not running', and start refuses to launch. The node serves inference through Ollama, so it cannot start without it.

fixInstall Ollama (https://ollama.com/download) and start it. On Linux the one-line installer does this for you; otherwise run `ollama serve`. Then pull a model and re-verify.

causestart prints that node <id> is 'no longer staked on-chain' and exits. The nodeId you passed is not backed by a stake (never registered, or the stake was withdrawn).

fixRegister the node in the Console, then start with the nodeId it assigns. Confirm the machine's nodeKeyHash matches the one you registered with parallelix-node status.

causestart prints that node <id> is 'in unstake cooldown' and exits. You called requestUnstake on this node, so nothing is dispatched to it.

fixTo keep running it, cancelUnstake(nodeId) in the Console Staking page, then start again. Otherwise wait out the 7-day cooldown and withdraw.

causestart stops after three rejections with a hint about --node-id and nodeKeyHash. The coordinator does not recognise this machine for that node.

fixCheck the --node-id is correct and that this machine's nodeKeyHash (from init) is exactly the one you registered on-chain. verify confirms the key and coordinator reachability.

causeMetaMask rejects the approve or registerNode transaction for insufficient gas. Registration is two on-chain transactions (approve $PRLX, then registerNode) on Ethereum mainnet.

fixHold a little ETH for gas in the staking wallet. Remember the 4% buy tax when acquiring $PRLX: buy a little more than 50,000 to clear the floor after tax.

causeThe installer or CLI does not run on native Windows. Native Windows is not supported.

fixInstall WSL2 (https://learn.microsoft.com/windows/wsl/install), open Ubuntu, and run the one-line installer and every command inside it.

causeprobe reports tier 3 (GPU · high) on a high-memory Mac. Apple Silicon shares unified memory between CPU and GPU, so the CLI caps the Metal GPU at tier 3 rather than datacenter tier 4.

fixThis is expected. Declare tier 3 when you register. Run probe to read the exact tier the CLI assigns this machine.

causeAfter running init --force, the running daemon's heartbeats stop being accepted. --force generates a new node key and a new nodeKeyHash, so the old on-chain registration no longer matches this machine.

fixEither restore the backed-up previous key, or register a fresh node with the new nodeKeyHash. Use --force deliberately. status prints the current nodeKeyHash so you can compare it against the Console.

causeThe Console shows the node in Idle and earnings stop accruing. A node only accrues while it is online and heartbeating; if the machine is off, asleep, or paused, it earns nothing for that time.

fixRun it as a service so it survives reboot and auto-restarts, rather than a foreground start that dies with the terminal.