Struct Supervisor 

pub struct Supervisor { /* private fields */ }

Manages running node processes. Holds child process handles and runtime state.

Implementations

impl Supervisor

pub fn new(event_tx: Sender<NodeEvent>) -> Self

pub async fn start_node( &mut self, config: &NodeConfig, supervisor_ref: Arc<RwLock<Supervisor>>, registry_ref: Arc<RwLock<NodeRegistry>>, ) -> Result<NodeStarted>

Start a node by spawning the actual process.

Returns NodeStarted on success. Spawns a background monitoring task that watches the child process and handles restart logic.

pub async fn stop_node(&mut self, node_id: u32) -> Result<()>

Stop a node by gracefully terminating its process.

Sends SIGTERM (Unix) or kills (Windows), waits up to 10 seconds for exit, then sends SIGKILL if needed. The monitor task detects the Stopping status and exits cleanly without attempting a restart.

pub async fn stop_all_nodes( &mut self, configs: &[(u32, String)], ) -> StopNodeResult

Stop all running nodes, returning an aggregate result.

pub fn node_status(&self, node_id: u32) -> Result<NodeStatus>

Get the status of a node.

pub fn node_pid(&self, node_id: u32) -> Option<u32>

Get the PID of a running node.

pub fn node_uptime_secs(&self, node_id: u32) -> Option<u64>

Get the uptime of a running node in seconds.

pub fn node_pending_version(&self, node_id: u32) -> Option<String>

The target version when the node is in UpgradeScheduled state, otherwise None.

pub fn is_running(&self, node_id: u32) -> bool

Check whether a node is running.

pub fn node_counts(&self) -> (u32, u32, u32)

Get counts of nodes in each state: (running, stopped, errored).

pub fn adopt_from_registry(&mut self, registry: &NodeRegistry) -> Vec<u32>

Restore running-node state from a previous daemon instance.

For each registered node, determines the PID to adopt via resolve_adopted_pid: try <data_dir>/node.pid first, and if it’s missing or stale, fall back to a process-table scan matching the node’s binary path and --root-dir argument. Live matches are inserted into node_states as Running.

The scan is what covers the upgrade path: nodes spawned by a pre-adoption daemon never had a pid file written, so without the fallback the first restart after installing this fix would still leave every previously-running node classified as Stopped.

Must be called before the HTTP server starts accepting requests — the window between Supervisor::new and adoption is where the API would otherwise report live nodes as Stopped. Adopted nodes have no associated monitor_node task (the tokio::process::Child handle belonged to the previous daemon, and tokio::process::Child::wait only works for the process’s actual parent). Their exits are detected instead by the spawn_liveness_monitor polling task.

Returns the list of node IDs that were adopted.