Troubleshooting

Start with the exact startup error and the last node log lines. Most failures fall into one of the following boundaries.

CPU unsupported

CPU UNSUPPORTED
This CPU is too old or does not expose the required instructions.
REQUIRED x86-64 SSE4.1 + PCLMULQDQ

On ARM64 the required pair is NEON and PMULL. Production does not fall back to the scalar reference implementation.

Some virtual-machine plans hide PCLMULQDQ even when the physical host supports it. Check the guest-visible CPU flags or choose a virtual CPU profile that exposes the required instructions.

Most Intel and AMD desktop or server processors manufactured since 2012 support SSE4.1 and PCLMULQDQ, but the hypervisor must expose them.

Run the non-mutating check:

parano1d --check-hardware

No peers

Check:

parano1d-cli peers

Confirm outbound DNS and TCP access, local time, and the P2P listener setting. If accepting inbound connections, allow TCP 9400 in both host and upstream firewalls.

Do not open RPC 9401 as a substitute for P2P.

Mining button or process is unavailable

Mining needs:

synchronization with the canonical tip;
two authenticated peers;
a ready embedded HistoryStep runtime;
a valid payout owner;
supported CPU instructions.

Use:

parano1d-cli status
parano1d-cli peers
parano1d-cli mining

An ordinary node started with --mode node never mines even if an old TOML file contains mining.enabled = true.

Synchronization appears stuck

Inspect whether the node is validating headers, verifying a terminal, transferring segments or applying the recent suffix. These phases have different CPU, network and disk profiles.

Try more than one healthy peer before treating unchanged height as corruption. An interrupted snapshot staging directory is discarded automatically at restart.

Address balance is missing after import

Automatic discovery stops at the first empty derived address and checks at most 20 funded candidates. If an older wallet skipped an unused index, generate the later addresses explicitly.

Receipts are local artifacts and are not restored by importing the key.

Send was rejected

Common causes are:

transaction epoch changed;
an input was spent by a new block;
active address changed;
output-slot hint became occupied;
fee quote is below current policy;
insufficient balance after fee;
too many selected inputs.

Refresh the wallet and construct a new intent. Do not edit serialized intent bytes; the authorization binds the complete logical transaction.

Clean shutdown takes time

Mining and proof tasks stop at safe cancellation boundaries. The GUI waits for the private node to close networking and flush MDBX. A short delay is expected under all-core work.

If a service does not stop after its configured timeout, preserve the logs before forcing it. Repeated hard kills can turn an ordinary delay into database recovery work.

Port already in use

Find which process owns 9400 or 9401. Only one node may bind a listener and one process may own a data directory.

Use distinct listener ports and distinct data directories for independent instances. A GUI-owned private node can also conflict with a manually started Core node on the defaults.

Report a reproducible fault

Include:

release version and operating system;
CPU model and --check-hardware output;
exact command or GUI action;
relevant log lines;
whether the node was synchronized and mining;
expected and observed result.

Never include wallet.key, the exported master secret, private photo material or bearer tokens.