Don't constrain baseboard when adding sled

nexus/src/app/rack.rs

@@ -52,7 +52,6 @@ use omicron_common::api::external::ListResultVec;
 use omicron_common::api::external::LookupResult;
 use omicron_common::api::external::Name;
 use omicron_common::api::external::NameOrId;
-use omicron_common::api::external::ResourceType;
 use omicron_common::api::internal::shared::ExternalPortDiscovery;
 use sled_agent_client::types::AddSledRequest;
 use sled_agent_client::types::EarlyNetworkConfigBody;
@@ -871,36 +870,15 @@ impl super::Nexus {
             )
             .await?;
 
-        // Grab the SPs from the last collection
-        let collection =
-            self.db_datastore.inventory_get_latest_collection(opctx).await?;
-
-        // If there isn't a collection, we don't know about the sled
-        let Some(collection) = collection else {
-            return Err(Error::unavail("no inventory data available"));
-        };
-
-        // Find the revision
-        let Some(sp) = collection.sps.get(&baseboard_id) else {
-            return Err(Error::ObjectNotFound {
-                type_name: ResourceType::Sled,
-                lookup_type:
-                    omicron_common::api::external::LookupType::ByCompositeId(
-                        format!("{sled:?}"),
-                    ),
-            });
-        };
-
-        // Convert the baseboard as necessary
-        let baseboard = sled_agent_client::types::Baseboard::Gimlet {
-            identifier: sled.serial.clone(),
-            model: sled.part.clone(),
-            revision: sp.baseboard_revision.into(),
+        // Convert `UninitializedSledId` to the sled-agent type
+        let baseboard_id = sled_agent_client::types::BaseboardId {
+            serial_number: sled.serial.clone(),
+            part_number: sled.part.clone(),
         };
 
         // Make the call to sled-agent
         let req = AddSledRequest {
-            sled_id: baseboard,
+            sled_id: baseboard_id,
             start_request: StartSledAgentRequest {
                 generation: 0,
                 schema_version: 1,

openapi/sled-agent.json

@@ -1240,7 +1240,7 @@
         "type": "object",
         "properties": {
           "sled_id": {
-            "$ref": "#/components/schemas/Baseboard"
+            "$ref": "#/components/schemas/BaseboardId"
           },
           "start_request": {
             "$ref": "#/components/schemas/StartSledAgentRequest"
@@ -1319,6 +1319,24 @@
           }
         ]
       },
+      "BaseboardId": {
+        "description": "A representation of a Baseboard ID as used in the inventory subsystem This type is essentially the same as a `Baseboard` except it doesn't have a revision or HW type (Gimlet, PC, Unknown).",
+        "type": "object",
+        "properties": {
+          "part_number": {
+            "description": "Oxide Part Number",
+            "type": "string"
+          },
+          "serial_number": {
+            "description": "Serial number (unique for a given part number)",
+            "type": "string"
+          }
+        },
+        "required": [
+          "part_number",
+          "serial_number"
+        ]
+      },
       "BgpConfig": {
         "type": "object",
         "properties": {

sled-agent/src/bootstrap/params.rs

@@ -174,10 +174,21 @@ impl TryFrom<UnvalidatedRackInitializeRequest> for RackInitializeRequest {
 pub type Certificate = nexus_client::types::Certificate;
 pub type RecoverySiloConfig = nexus_client::types::RecoverySiloConfig;
 
+/// A representation of a Baseboard ID as used in the inventory subsystem
+/// This type is essentially the same as a `Baseboard` except it doesn't have a
+/// revision or HW type (Gimlet, PC, Unknown).
+#[derive(Clone, Debug, Serialize, Deserialize, PartialEq, JsonSchema)]
+pub struct BaseboardId {
+    /// Oxide Part Number
+    pub part_number: String,
+    /// Serial number (unique for a given part number)
+    pub serial_number: String,
+}
+
 /// A request to Add a given sled after rack initialization has occurred
 #[derive(Clone, Debug, Serialize, Deserialize, PartialEq, JsonSchema)]
 pub struct AddSledRequest {
-    pub sled_id: Baseboard,
+    pub sled_id: BaseboardId,
     pub start_request: StartSledAgentRequest,
 }
 
@@ -255,9 +266,6 @@ pub struct StartSledAgentRequestBody {
     /// true.
     pub is_lrtq_learner: bool,
 
-    // Note: The order of these fields is load bearing, because we serialize
-    // `SledAgentRequest`s as toml. `subnet` serializes as a TOML table, so it
-    // must come after non-table fields.
     /// Portion of the IP space to be managed by the Sled Agent.
     pub subnet: Ipv6Subnet<SLED_PREFIX>,
 }

sled-agent/src/sled_agent.rs

@@ -9,7 +9,7 @@ use crate::bootstrap::config::BOOTSTRAP_AGENT_RACK_INIT_PORT;
 use crate::bootstrap::early_networking::{
     EarlyNetworkConfig, EarlyNetworkSetupError,
 };
-use crate::bootstrap::params::StartSledAgentRequest;
+use crate::bootstrap::params::{BaseboardId, StartSledAgentRequest};
 use crate::config::Config;
 use crate::instance_manager::{InstanceManager, ReservoirMode};
 use crate::long_running_tasks::LongRunningTaskHandles;
@@ -1187,8 +1187,8 @@ pub enum AddSledError {
     },
     #[error("Failed to connect to DDM")]
     DdmAdminClient(#[source] ddm_admin_client::DdmError),
-    #[error("Failed to learn bootstrap ip for {0}")]
-    NotFound(Baseboard),
+    #[error("Failed to learn bootstrap ip for {0:?}")]
+    NotFound(BaseboardId),
     #[error("Failed to initialize {sled_id}: {err}")]
     BootstrapTcpClient {
         sled_id: Baseboard,
@@ -1199,7 +1199,7 @@ pub enum AddSledError {
 /// Add a sled to an initialized rack.
 pub async fn sled_add(
     log: Logger,
-    sled_id: Baseboard,
+    sled_id: BaseboardId,
     request: StartSledAgentRequest,
 ) -> Result<(), AddSledError> {
     // Get all known bootstrap addresses via DDM
@@ -1227,16 +1227,20 @@ pub async fn sled_add(
         })
         .collect::<FuturesUnordered<_>>();
 
-    // Execute the futures until we find our matching sled or done searching
+    // Execute the futures until we find our matching sled or are done searching
     let mut target_ip = None;
+    let mut found_baseboard = None;
     while let Some((ip, result)) = addrs_to_sleds.next().await {
         match result {
             Ok(baseboard) => {
                 // Convert from progenitor type back to `sled-hardware`
                 // type.
-                let found = baseboard.into_inner().into();
-                if sled_id == found {
+                let found: Baseboard = baseboard.into_inner().into();
+                if sled_id.serial_number == found.identifier()
+                    && sled_id.part_number == found.model()
+                {
                     target_ip = Some(ip);
+                    found_baseboard = Some(found);
                     break;
                 }
             }
@@ -1259,10 +1263,14 @@ pub async fn sled_add(
         log.new(o!("BootstrapAgentClient" => bootstrap_addr.to_string())),
     );
 
+    // Safe to unwrap, because we would have bailed when checking target_ip
+    // above otherwise. baseboard and target_ip are set together.
+    let baseboard = found_baseboard.unwrap();
+
     client.start_sled_agent(&request).await.map_err(|err| {
-        AddSledError::BootstrapTcpClient { sled_id: sled_id.clone(), err }
+        AddSledError::BootstrapTcpClient { sled_id: baseboard.clone(), err }
     })?;
 
-    info!(log, "Peer agent initialized"; "peer_bootstrap_addr" => %bootstrap_addr, "peer_id" => %sled_id);
+    info!(log, "Peer agent initialized"; "peer_bootstrap_addr" => %bootstrap_addr, "peer_id" => %baseboard);
     Ok(())
 }