Choose the process form
The bare launcher starts a background server under Aion home and reports it. An explicit server command runs in the foreground. Use the form that fits your process supervisor.
aion server --config aion.tomlRead configuration for file discovery and the starter config. Use a local-only HTTP and gRPC listener for an unauthenticated evaluation.
Verify it
Status checks the recorded process identity and the live build identity, then reports serving and readiness separately.
Report which server is actually running: the pid-file record, the live build identity from its own listener, and readiness — each layer's absence reported as its own distinct state
aion server status [OPTIONS]-
--config <CONFIG> - Path to the TOML server configuration file. Used to resolve the store path and, when no pid file exists, the address to probe
The bare launcher’s log is under Aion home. Foreground operation emits JSON logs to stdout. Read the error and the configuration path it names.
An HTTP status alone is insufficient: unknown URLs can return the console HTML. Check the body and content type of identity and health readings.
Writer lock
The durable store has one writer lock. A new process waits for the current holder and reports the wait. Do not delete the lock file, change storage permissions broadly or kill an unrelated process to force a boot.
A restart with new configuration must use the same intended home and data directory. Point recovery at the existing data directory. An empty directory creates a new store.
Shutdown
The local stop command verifies the recorded process identity before signalling. The server closes admission, drains work and stops its managed processes. It reports the drain outcome and budget.
Stop the server running under this machine's Aion home.
aion stop [OPTIONS]-
--config <CONFIG> - Path to the TOML server configuration file. Used only to resolve the wait patience when `--patience` is absent and no running process's record answers: the stop budget the config describes — its drain window (`drain.timeout_seconds`), plus twice the managed workers' stop grace, plus the fixed harness termination grace
-
--patience <PATIENCE> - Seconds to wait for the server to exit before reporting it still draining. Absent, the running server's own recorded stop budget governs (drain window + twice the managed-worker stop grace + the fixed harness termination grace), then the stop budget its config describes.
A second termination signal requests forced shutdown. A successful workflow cancellation is a different act from stopping a server process.
Continue
Use upgrading for binary handover and migrations. Use backups before planned storage changes.
Several nodes
Follow cluster setup to assign membership, shards and peer listeners before starting a new cluster.