feat(api): poll power state for PowerShelf in Ready

Cargo.toml

@@ -27,7 +27,7 @@ authors = ["NVIDIA Carbide Engineering <carbide-dev@exchange.nvidia.com>"]
 [workspace.dependencies]
 clap = { version = "4", features = ["derive", "env"] }
 libredfish = { git = "https://github.com/NVIDIA/libredfish.git", tag = "v0.44.4" }
-librms = { git = "https://github.com/NVIDIA/nv-rms-client.git", tag = "v0.0.12-rc3" }
+librms = { git = "https://github.com/NVIDIA/nv-rms-client.git", tag = "v0.0.12-rc4" }
 ansi-to-html = "0.2.2"
 
 tokio = { version = "1", features = ["full", "tracing"] }

crates/api-test-helper/src/mock_rms.rs

@@ -67,6 +67,10 @@ pub struct MockRmsApi {
         Mutex<VecDeque<Result<rms::GetPowerStateResponse, RackManagerError>>>,
     get_power_state_calls: Mutex<Vec<rms::GetPowerStateRequest>>,
 
+    get_power_state_by_device_list_responses:
+        Mutex<VecDeque<Result<rms::GetPowerStateByDeviceListResponse, RackManagerError>>>,
+    get_power_state_by_device_list_calls: Mutex<Vec<rms::GetPowerStateByDeviceListRequest>>,
+
     sequence_rack_power_responses:
         Mutex<VecDeque<Result<rms::SequenceRackPowerResponse, RackManagerError>>>,
     sequence_rack_power_calls: Mutex<Vec<rms::SequenceRackPowerRequest>>,
@@ -244,6 +248,8 @@ impl MockRmsApi {
             set_power_state_calls: Default::default(),
             set_power_state_by_device_list_responses: Default::default(),
             set_power_state_by_device_list_calls: Default::default(),
+            get_power_state_by_device_list_responses: Default::default(),
+            get_power_state_by_device_list_calls: Default::default(),
             get_power_state_responses: Default::default(),
             get_power_state_calls: Default::default(),
             sequence_rack_power_responses: Default::default(),
@@ -809,6 +815,16 @@ fn pop_or_err<T>(
 
 #[async_trait::async_trait]
 impl RmsApi for MockRmsApi {
+    async fn get_power_state_by_device_list(
+        &self,
+        cmd: rms::GetPowerStateByDeviceListRequest,
+    ) -> Result<rms::GetPowerStateByDeviceListResponse, RackManagerError> {
+        self.get_power_state_by_device_list_calls
+            .lock()
+            .await
+            .push(cmd);
+        pop_or_err(&mut self.get_power_state_by_device_list_responses.lock().await)
+    }
     async fn set_power_state(
         &self,
         cmd: rms::SetPowerStateRequest,

crates/power-shelf-controller/src/maintenance.rs

@@ -245,11 +245,12 @@ async fn invoke_rms_power_operation(
 }
 
 /// Build the `rms::NewNodeInfo` describing this power shelf for inclusion
-/// in a `SetPowerStateByDeviceList` request. Resolves the BMC IP from the
-/// database and BMC credentials via the credential manager, since the
-/// caller-supplied variant of the RPC requires the BMC connection details
-/// inline rather than relying on RMS's inventory.
-async fn build_power_shelf_node_info(
+/// in any caller-supplied `NodeSet` request (`SetPowerStateByDeviceList`
+/// from `Maintenance`, `GetDeviceInfoByDeviceList` from `Ready`). Resolves
+/// the BMC IP from the database and BMC credentials via the credential
+/// manager, since these RPCs require the BMC connection details inline
+/// rather than relying on RMS's inventory.
+pub(super) async fn build_power_shelf_node_info(
     power_shelf_id: &PowerShelfId,
     state: &PowerShelf,
     rack_id: String,

crates/power-shelf-controller/src/ready.rs

@@ -17,27 +17,33 @@
 
 //! Handler for PowerShelfControllerState::Ready.
 
+use carbide_rack::rack_manager_error;
 use carbide_uuid::power_shelf::PowerShelfId;
-use model::power_shelf::{PowerShelf, PowerShelfControllerState};
+use db::power_shelf as db_power_shelf;
+use librms::protos::rack_manager as rms;
+use model::power_shelf::{PowerShelf, PowerShelfControllerState, PowerShelfStatus};
+use sqlx::PgTransaction;
 use state_controller::state_handler::{
     StateHandlerContext, StateHandlerError, StateHandlerOutcome,
 };
 
 use crate::context::PowerShelfStateHandlerContextObjects;
+use crate::maintenance::build_power_shelf_node_info;
 
 /// Handles the Ready state for a power shelf.
 ///
 /// If the power shelf is marked for deletion, transitions to `Deleting`.
 /// If a maintenance request has been posted via
 /// `power_shelf_maintenance_requested`, transitions to `Maintenance` with the
-/// requested operation (PowerOn / PowerOff). Otherwise idles.
+/// requested operation (PowerOn / PowerOff). Otherwise polls RMS for the
+/// current power state (best-effort observation) and idles.
 ///
 /// TODO: Implement PowerShelf monitoring (health checks, status updates,
 /// power consumption / efficiency tracking).
 pub async fn handle_ready(
     power_shelf_id: &PowerShelfId,
     state: &mut PowerShelf,
-    _ctx: &mut StateHandlerContext<'_, PowerShelfStateHandlerContextObjects>,
+    ctx: &mut StateHandlerContext<'_, PowerShelfStateHandlerContextObjects>,
 ) -> Result<StateHandlerOutcome<PowerShelfControllerState>, StateHandlerError> {
     if state.is_marked_as_deleted() {
         return Ok(StateHandlerOutcome::transition(
@@ -58,9 +64,193 @@ pub async fn handle_ready(
         ));
     }
 
-    tracing::info!("PowerShelf {} is ready", power_shelf_id,);
-    Ok(StateHandlerOutcome::wait(format!(
-        "PowerShelf {} is ready",
-        power_shelf_id
-    )))
+    let txn = poll_rms_power_state(power_shelf_id, state, ctx).await;
+
+    Ok(StateHandlerOutcome::do_nothing().with_txn_opt(txn))
+}
+///
+/// On a successful response, the observed `pstate` for this power shelf is
+/// persisted to the `power_shelves.status` column and the in-memory `state`
+/// is updated to match. The returned `PgTransaction` (if any) carries that
+/// status write so the caller can attach it to the `Ready` outcome and have
+/// the state-controller framework commit it alongside the usual outcome
+/// bookkeeping.
+async fn poll_rms_power_state(
+    power_shelf_id: &PowerShelfId,
+    state: &mut PowerShelf,
+    ctx: &mut StateHandlerContext<'_, PowerShelfStateHandlerContextObjects>,
+) -> Option<PgTransaction<'static>> {
+    let Some(rms_client) = ctx.services.rms_client.as_ref() else {
+        tracing::debug!(
+            power_shelf_id = %power_shelf_id,
+            "PowerShelf Ready: skipping RMS GetPowerStateByDeviceList; RMS client not configured",
+        );
+        return None;
+    };
+
+    let Some(rack_id) = state.rack_id.as_ref() else {
+        tracing::debug!(
+            power_shelf_id = %power_shelf_id,
+            "PowerShelf Ready: skipping RMS GetPowerStateByDeviceList; power shelf has no rack association",
+        );
+        return None;
+    };
+
+    let device = match build_power_shelf_node_info(
+        power_shelf_id,
+        state,
+        rack_id.to_string(),
+        &ctx.services.db_pool,
+        ctx.services.credential_manager.as_ref(),
+    )
+    .await
+    {
+        Ok(device) => device,
+        Err(cause) => {
+            tracing::debug!(
+                power_shelf_id = %power_shelf_id,
+                rack_id = %rack_id,
+                cause = %cause,
+                "PowerShelf Ready: skipping RMS GetPowerStateByDeviceList; unable to build NodeSet",
+            );
+            return None;
+        }
+    };
+
+    let request = rms::GetPowerStateByDeviceListRequest {
+        nodes: Some(rms::NodeSet {
+            devices: vec![device],
+        }),
+        ..Default::default()
+    };
+
+    let rack_id_str = rack_id.to_string();
+    let response = match rms_client.get_power_state_by_device_list(request).await {
+        Ok(response) => response,
+        Err(error) => {
+            let error = rack_manager_error("get_power_state_by_device_list", error);
+            tracing::warn!(
+                power_shelf_id = %power_shelf_id,
+                rack_id = %rack_id_str,
+                error = %error,
+                "RMS GetPowerStateByDeviceList transport error",
+            );
+            return None;
+        }
+    };
+
+    let batch = response.response.clone().unwrap_or_default();
+    if !(batch.status == rms::ReturnCode::Success as i32 && batch.failed_nodes == 0) {
+        tracing::warn!(
+            power_shelf_id = %power_shelf_id,
+            rack_id = %rack_id_str,
+            batch_status = batch.status,
+            successful_nodes = batch.successful_nodes,
+            failed_nodes = batch.failed_nodes,
+            message = %batch.message,
+            "RMS GetPowerStateByDeviceList returned non-Success result",
+        );
+        return None;
+    }
+
+    tracing::info!(
+        power_shelf_id = %power_shelf_id,
+        rack_id = %rack_id_str,
+        successful_nodes = batch.successful_nodes,
+        pstates = ?response
+            .node_power_states
+            .iter()
+            .map(|node| (node.node_id.as_str(), node.pstate.as_str()))
+            .collect::<Vec<_>>(),
+        "RMS GetPowerStateByDeviceList succeeded",
+    );
+
+    persist_observed_power_state(power_shelf_id, state, ctx, &response.node_power_states).await
+}
+
+/// Look up the `NodePowerState` for this power shelf in the RMS response,
+/// stamp the value into `state.status`, and persist it via
+/// `db_power_shelf::update`. Returns the open `PgTransaction` so the caller
+/// can attach it to the `Ready` outcome.
+///
+/// Status persistence is best-effort: if RMS did not echo a result for this
+/// node, or if the DB write fails, the in-memory state is left untouched
+/// and `None` is returned — `Ready` must stay in `Ready` regardless.
+async fn persist_observed_power_state(
+    power_shelf_id: &PowerShelfId,
+    state: &mut PowerShelf,
+    ctx: &mut StateHandlerContext<'_, PowerShelfStateHandlerContextObjects>,
+    node_power_states: &[rms::NodePowerState],
+) -> Option<PgTransaction<'static>> {
+    let node_id = power_shelf_id.to_string();
+    let Some(observed) = node_power_states
+        .iter()
+        .find(|node| node.node_id == node_id)
+    else {
+        tracing::debug!(
+            power_shelf_id = %power_shelf_id,
+            "RMS GetPowerStateByDeviceList: no NodePowerState echoed for this power shelf; skipping status update",
+        );
+        return None;
+    };
+
+    let new_power_state = observed.pstate.to_lowercase();
+    let new_status = match state.status.as_ref() {
+        Some(existing) => PowerShelfStatus {
+            shelf_name: existing.shelf_name.clone(),
+            power_state: new_power_state.clone(),
+            health_status: existing.health_status.clone(),
+        },
+        None => PowerShelfStatus {
+            shelf_name: state.config.name.clone(),
+            power_state: new_power_state.clone(),
+            health_status: String::new(),
+        },
+    };
+
+    if state
+        .status
+        .as_ref()
+        .is_some_and(|s| s.power_state == new_status.power_state)
+    {
+        tracing::debug!(
+            power_shelf_id = %power_shelf_id,
+            power_state = %new_status.power_state,
+            "PowerShelf status power_state unchanged; skipping DB write",
+        );
+        return None;
+    }
+
+    let previous_status = state.status.replace(new_status);
+
+    let mut txn = match ctx.services.db_pool.begin().await {
+        Ok(txn) => txn,
+        Err(error) => {
+            state.status = previous_status;
+            tracing::warn!(
+                power_shelf_id = %power_shelf_id,
+                error = %error,
+                "PowerShelf Ready: failed to begin txn while persisting observed power state",
+            );
+            return None;
+        }
+    };
+
+    if let Err(error) = db_power_shelf::update(state, &mut txn).await {
+        state.status = previous_status;
+        tracing::warn!(
+            power_shelf_id = %power_shelf_id,
+            error = %error,
+            "PowerShelf Ready: failed to persist observed power state to DB",
+        );
+        return None;
+    }
+
+    tracing::info!(
+        power_shelf_id = %power_shelf_id,
+        power_state = %new_power_state,
+        "PowerShelf Ready: persisted observed power state from RMS",
+    );
+
+    Some(txn)
 }

crates/rack/src/rms_client.rs

@@ -158,6 +158,10 @@ pub mod test_support {
 
         fn build_mock_client(&self) -> MockRmsClient {
             MockRmsClient {
+                submitted_get_power_state_by_device_list_requests: Arc::new(Mutex::new(Vec::new())),
+                queued_get_power_state_by_device_list_responses: Arc::new(Mutex::new(
+                    VecDeque::new(),
+                )),
                 fail_add_node: self.fail_add_node.clone(),
                 fail_inventory_get: self.fail_inventory_get.clone(),
                 registered_nodes: self.registered_nodes.clone(),
@@ -452,6 +456,10 @@ pub mod test_support {
         switch_system_image_job_statuses:
             Arc<Mutex<HashMap<String, rms::GetSwitchSystemImageJobStatusResponse>>>,
         switch_system_image_job_errors: Arc<Mutex<HashMap<String, String>>>,
+        submitted_get_power_state_by_device_list_requests:
+            Arc<Mutex<Vec<rms::GetPowerStateByDeviceListRequest>>>,
+        queued_get_power_state_by_device_list_responses:
+            Arc<Mutex<VecDeque<Result<rms::GetPowerStateByDeviceListResponse, RackManagerError>>>>,
         submitted_get_device_info_by_device_list_requests:
             Arc<Mutex<Vec<rms::GetDeviceInfoByDeviceListRequest>>>,
         queued_get_device_info_by_device_list_responses:
@@ -473,6 +481,21 @@ pub mod test_support {
 
     #[async_trait::async_trait]
     impl RmsApi for MockRmsClient {
+        async fn get_power_state_by_device_list(
+            &self,
+            cmd: rms::GetPowerStateByDeviceListRequest,
+        ) -> Result<rms::GetPowerStateByDeviceListResponse, RackManagerError> {
+            self.submitted_get_power_state_by_device_list_requests
+                .lock()
+                .await
+                .push(cmd);
+            self.queued_get_power_state_by_device_list_responses
+                .lock()
+                .await
+                .pop_front()
+                .unwrap_or(Ok(rms::GetPowerStateByDeviceListResponse::default()))
+        }
+
         async fn get_device_info_by_device_list(
             &self,
             cmd: rms::GetDeviceInfoByDeviceListRequest,