Runtime 0.8.12
Keep the node running
On Linux with a usable systemd user session, the installer leaves your node running as an enabled service. It reports login or boot startup.
What a new install already has
The current one-line installer first starts a detached node, then hands it to rivethub.service. You do not need to initialize or start a second service after a successful install. The installer says whether the service starts when you log in or at boot without login. When lingering is off, it prints sudo loginctl enable-linger <user> for you to run; it never runs that command itself.
$ rh service status$ rh service logsA user service starts at login unless lingering is enabled. To choose boot without login yourself:
$ sudo loginctl enable-linger "$USER"The installer keeps --no-cloud in the node’s config when handing it to the service. --no-service skips the handover; --no-start installs without starting anything. If service setup fails, the installer explains the failure, disables the service and falls back to a detached node that will not restart by itself.
Older installs and manual setup
If you installed before the service handover was added, run the one-line installer again. On Linux with a usable systemd user session and rh 0.8.4, it moves an existing detached node under the service and can recover a unit looping beside that node.
$ curl -fsSL https://get.rivethub.io/local.sh | bashFor a by-hand install, use the same order as the product download guide. Run as your ordinary user for a user unit. If a unit is already looping, stop it first with rh service stop, then follow this sequence:
$ rh local --ensure$ rh service init$ rh stopwhile rh status 2>/dev/null | grep -q '^RivetHub is running'; do sleep 1; done$ rh service startIn 0.8.4, rh service start hands an existing detached node to the service and waits for it to exit. Running init and start beside that node no longer causes the owner-lock restart loop. The explicit stop-and-wait sequence above also works.
If you are still on 0.8.3, stop a looping unit with rh service stop, then use the stop-and-wait sequence above before starting it; update with rh update.
Init writes and enables the unit: standalone SQLite runs rh local; a full node config runs the supervisor. Root init creates a system unit instead; use a deliberate service account and config for a datahub. Use rh service stop, rh service restart and rh service status to manage it.
Environment settings
In 0.8.4, the generated unit’s EnvironmentFile directive is accepted by systemd and loads ~/.rivethub/.env. Protect that file with mode 0600. rh service restart and rh update rewrite an older generated unit, keep rivethub.service.bak and reload systemd. A unit edited by hand is left in place and the difference is printed; review it yourself.
$ rh service restartIf you are still on 0.8.3, its quoted directive can leave ~/.rivethub/.env unloaded; update with rh update.
macOS, no systemd and opting out
On macOS, WSL without systemd, a container without a usable systemd user session, or with --no-service, the installer leaves a detached node and says it will not restart by itself. Start it after a restart with rh local --ensure. With --no-start, nothing is started; for a new offline config, use rh local --no-cloud instead. The installer refuses --no-cloud combined with --no-start.
The service CLI manages systemd only. macOS operators must configure launchd themselves; this guide does not supply a tested plist. WSL needs a working systemd user session for the Linux steps. Starting on demand does not register a boot service. Desktop is preview software with no release.
Need a hand?
Check the troubleshooting guide or ask with a redacted report.