Distinguish between Scrimlets and Gimlets (#1510)

* Distinguish between Scrimlets and Gimlets

* Update scrimlet status on upsert

Author: smklein

common/src/sql/dbinit.sql

@@ -78,6 +78,9 @@ CREATE TABLE omicron.public.sled (
     /* FK into the Rack table */
     rack_id UUID NOT NULL,
 
+    /* Idenfities if this Sled is a Scrimlet */
+    is_scrimlet BOOL NOT NULL,
+
     /* The IP address and bound port of the sled agent server. */
     ip INET NOT NULL,
     port INT4 CHECK (port BETWEEN 0 AND 65535) NOT NULL,

nexus/db-model/src/schema.rs

@@ -346,6 +346,7 @@ table! {
         rcgen -> Int8,
 
         rack_id -> Uuid,
+        is_scrimlet -> Bool,
         ip -> Inet,
         port -> Int4,
         last_used_address -> Inet,

nexus/db-model/src/sled.rs

@@ -24,6 +24,8 @@ pub struct Sled {
 
     pub rack_id: Uuid,
 
+    is_scrimlet: bool,
+
     // ServiceAddress (Sled Agent).
     pub ip: ipv6::Ipv6Addr,
     pub port: SqlU16,
@@ -33,7 +35,12 @@ pub struct Sled {
 }
 
 impl Sled {
-    pub fn new(id: Uuid, addr: SocketAddrV6, rack_id: Uuid) -> Self {
+    pub fn new(
+        id: Uuid,
+        addr: SocketAddrV6,
+        is_scrimlet: bool,
+        rack_id: Uuid,
+    ) -> Self {
         let last_used_address = {
             let mut segments = addr.ip().segments();
             segments[7] += omicron_common::address::RSS_RESERVED_ADDRESSES;
@@ -44,12 +51,17 @@ impl Sled {
             time_deleted: None,
             rcgen: Generation::new(),
             rack_id,
+            is_scrimlet,
             ip: ipv6::Ipv6Addr::from(addr.ip()),
             port: addr.port().into(),
             last_used_address,
         }
     }
 
+    pub fn is_scrimlet(&self) -> bool {
+        self.is_scrimlet
+    }
+
     pub fn ip(&self) -> Ipv6Addr {
         self.ip.into()
     }

nexus/src/app/sled.rs

@@ -10,13 +10,15 @@ use crate::db::identity::Asset;
 use crate::db::lookup::LookupPath;
 use crate::db::model::DatasetKind;
 use crate::db::model::ServiceKind;
-use crate::internal_api::params::ZpoolPutRequest;
+use crate::internal_api::params::{
+    SledAgentStartupInfo, SledRole, ZpoolPutRequest,
+};
 use omicron_common::api::external::DataPageParams;
 use omicron_common::api::external::Error;
 use omicron_common::api::external::ListResultVec;
 use omicron_common::api::external::LookupResult;
 use sled_agent_client::Client as SledAgentClient;
-use std::net::{Ipv6Addr, SocketAddr, SocketAddrV6};
+use std::net::{Ipv6Addr, SocketAddr};
 use std::sync::Arc;
 use uuid::Uuid;
 
@@ -28,10 +30,21 @@ impl super::Nexus {
     pub async fn upsert_sled(
         &self,
         id: Uuid,
-        address: SocketAddrV6,
+        info: SledAgentStartupInfo,
     ) -> Result<(), Error> {
         info!(self.log, "registered sled agent"; "sled_uuid" => id.to_string());
-        let sled = db::model::Sled::new(id, address, self.rack_id);
+
+        let is_scrimlet = match info.role {
+            SledRole::Gimlet => false,
+            SledRole::Scrimlet => true,
+        };
+
+        let sled = db::model::Sled::new(
+            id,
+            info.sa_address,
+            is_scrimlet,
+            self.rack_id,
+        );
         self.db_datastore.sled_upsert(sled).await?;
         Ok(())
     }

nexus/src/db/datastore/mod.rs

@@ -431,7 +431,8 @@ mod test {
         );
         let rack_id = Uuid::new_v4();
         let sled_id = Uuid::new_v4();
-        let sled = Sled::new(sled_id, bogus_addr.clone(), rack_id);
+        let is_scrimlet = false;
+        let sled = Sled::new(sled_id, bogus_addr.clone(), is_scrimlet, rack_id);
         datastore.sled_upsert(sled).await.unwrap();
         sled_id
     }
@@ -790,12 +791,13 @@ mod test {
         let rack_id = Uuid::new_v4();
         let addr1 = "[fd00:1de::1]:12345".parse().unwrap();
         let sled1_id = "0de4b299-e0b4-46f0-d528-85de81a7095f".parse().unwrap();
-        let sled1 = db::model::Sled::new(sled1_id, addr1, rack_id);
+        let is_scrimlet = false;
+        let sled1 = db::model::Sled::new(sled1_id, addr1, is_scrimlet, rack_id);
         datastore.sled_upsert(sled1).await.unwrap();
 
         let addr2 = "[fd00:1df::1]:12345".parse().unwrap();
         let sled2_id = "66285c18-0c79-43e0-e54f-95271f271314".parse().unwrap();
-        let sled2 = db::model::Sled::new(sled2_id, addr2, rack_id);
+        let sled2 = db::model::Sled::new(sled2_id, addr2, is_scrimlet, rack_id);
         datastore.sled_upsert(sled2).await.unwrap();
 
         let ip = datastore.next_ipv6_address(&opctx, sled1_id).await.unwrap();

nexus/src/db/datastore/sled.rs

@@ -34,6 +34,8 @@ impl DataStore {
                 dsl::time_modified.eq(Utc::now()),
                 dsl::ip.eq(sled.ip),
                 dsl::port.eq(sled.port),
+                dsl::rack_id.eq(sled.rack_id),
+                dsl::is_scrimlet.eq(sled.is_scrimlet()),
             ))
             .returning(Sled::as_returning())
             .get_result_async(self.pool())

nexus/src/internal_api/http_entrypoints.rs

@@ -36,7 +36,7 @@ type NexusApiDescription = ApiDescription<Arc<ServerContext>>;
 /// Returns a description of the internal nexus API
 pub fn internal_api() -> NexusApiDescription {
     fn register_endpoints(api: &mut NexusApiDescription) -> Result<(), String> {
-        api.register(cpapi_sled_agents_post)?;
+        api.register(sled_agent_put)?;
         api.register(rack_initialization_complete)?;
         api.register(zpool_put)?;
         api.register(dataset_put)?;
@@ -63,26 +63,22 @@ struct SledAgentPathParam {
 }
 
 /// Report that the sled agent for the specified sled has come online.
-// TODO: Should probably be "PUT", since:
-// 1. We're upserting the value
-// 2. The client supplies the UUID
-// 3. This call is idempotent (mod "time_modified").
 #[endpoint {
      method = POST,
      path = "/sled-agents/{sled_id}",
  }]
-async fn cpapi_sled_agents_post(
+async fn sled_agent_put(
     rqctx: Arc<RequestContext<Arc<ServerContext>>>,
     path_params: Path<SledAgentPathParam>,
     sled_info: TypedBody<SledAgentStartupInfo>,
 ) -> Result<HttpResponseUpdatedNoContent, HttpError> {
     let apictx = rqctx.context();
     let nexus = &apictx.nexus;
     let path = path_params.into_inner();
-    let si = sled_info.into_inner();
+    let info = sled_info.into_inner();
     let sled_id = &path.sled_id;
     let handler = async {
-        nexus.upsert_sled(*sled_id, si.sa_address).await?;
+        nexus.upsert_sled(*sled_id, info).await?;
         Ok(HttpResponseUpdatedNoContent())
     };
     apictx.internal_latencies.instrument_dropshot_handler(&rqctx, handler).await

nexus/types/src/internal_api/params.rs

@@ -13,11 +13,28 @@ use std::net::SocketAddrV6;
 use std::str::FromStr;
 use uuid::Uuid;
 
+/// Describes the role of the sled within the rack.
+///
+/// Note that this may change if the sled is physically moved
+/// within the rack.
+#[derive(Serialize, Deserialize, JsonSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum SledRole {
+    /// The sled is a general compute sled.
+    Gimlet,
+    /// The sled is attached to the network switch, and has additional
+    /// responsibilities.
+    Scrimlet,
+}
+
 /// Sent by a sled agent on startup to Nexus to request further instruction
 #[derive(Serialize, Deserialize, JsonSchema)]
 pub struct SledAgentStartupInfo {
-    /// the address of the sled agent's API endpoint
+    /// The address of the sled agent's API endpoint
     pub sa_address: SocketAddrV6,
+
+    /// Describes the responsibilities of the sled
+    pub role: SledRole,
 }
 
 /// Sent by a sled agent on startup to Nexus to request further instruction

openapi/nexus-internal.json

@@ -281,7 +281,7 @@
     "/sled-agents/{sled_id}": {
       "post": {
         "summary": "Report that the sled agent for the specified sled has come online.",
-        "operationId": "cpapi_sled_agents_post",
+        "operationId": "sled_agent_put",
         "parameters": [
           {
             "in": "path",
@@ -1788,15 +1788,32 @@
         "description": "Sent by a sled agent on startup to Nexus to request further instruction",
         "type": "object",
         "properties": {
+          "role": {
+            "description": "Describes the responsibilities of the sled",
+            "allOf": [
+              {
+                "$ref": "#/components/schemas/SledRole"
+              }
+            ]
+          },
           "sa_address": {
-            "description": "the address of the sled agent's API endpoint",
+            "description": "The address of the sled agent's API endpoint",
             "type": "string"
           }
         },
         "required": [
+          "role",
           "sa_address"
         ]
       },
+      "SledRole": {
+        "description": "Describes the role of the sled within the rack.\n\nNote that this may change if the sled is physically moved within the rack.",
+        "type": "string",
+        "enum": [
+          "gimlet",
+          "scrimlet"
+        ]
+      },
       "ZpoolPutRequest": {
         "description": "Sent by a sled agent on startup to Nexus to request further instruction",
         "type": "object",

sled-agent/src/bootstrap/agent.rs

@@ -263,11 +263,20 @@ impl Agent {
 
             return Ok(SledAgentResponse { id: server.id() });
         }
+
+        // TODO(https://github.com/oxidecomputer/omicron/issues/823):
+        // Currently, the prescence or abscence of RSS is our signal
+        // for "is this a scrimlet or not".
+        // Longer-term, we should make this call based on the underlying
+        // hardware.
+        let is_scrimlet = self.rss.lock().await.is_some();
+
         // Server does not exist, initialize it.
         let server = SledServer::start(
             &self.sled_config,
             self.parent_log.clone(),
             sled_address,
+            is_scrimlet,
             request.rack_id,
         )
         .await