Add desired host phase 2 contents to OmicronSledConfig (PR 1/4)

dev-tools/omdb/src/bin/omdb/db.rs

@@ -131,6 +131,7 @@ use nexus_sled_agent_shared::inventory::BootPartitionDetails;
 use nexus_sled_agent_shared::inventory::ConfigReconcilerInventory;
 use nexus_sled_agent_shared::inventory::ConfigReconcilerInventoryResult;
 use nexus_sled_agent_shared::inventory::ConfigReconcilerInventoryStatus;
+use nexus_sled_agent_shared::inventory::HostPhase2DesiredContents;
 use nexus_sled_agent_shared::inventory::OmicronSledConfig;
 use nexus_sled_agent_shared::inventory::OmicronZoneImageSource;
 use nexus_sled_agent_shared::inventory::OrphanedDataset;
@@ -7567,12 +7568,30 @@ fn inv_collection_print_sled_config(label: &str, config: &OmicronSledConfig) {
         datasets,
         zones,
         remove_mupdate_override,
+        host_phase_2,
     } = config;
 
     println!("\n{label} SLED CONFIG");
     println!("    generation: {}", generation);
     println!("    remove_mupdate_override: {remove_mupdate_override:?}");
 
+    let display_host_phase_2_desired = |desired| match desired {
+        HostPhase2DesiredContents::CurrentContents => {
+            Cow::Borrowed("keep existing current contents")
+        }
+        HostPhase2DesiredContents::Artifact { hash } => {
+            Cow::Owned(format!("artifact {hash}"))
+        }
+    };
+    println!(
+        "    desired host phase 2 slot a: {}",
+        display_host_phase_2_desired(host_phase_2.slot_a)
+    );
+    println!(
+        "    desired host phase 2 slot b: {}",
+        display_host_phase_2_desired(host_phase_2.slot_b)
+    );
+
     if disks.is_empty() {
         println!("    disk config empty");
     } else {

nexus-sled-agent-shared/src/inventory.rs

@@ -576,6 +576,47 @@ pub enum SledRole {
     Scrimlet,
 }
 
+/// Describes the desired contents of a host phase 2 slot (i.e., the boot
+/// partition on one of the internal M.2 drives).
+#[derive(
+    Clone, Copy, Debug, Deserialize, Serialize, JsonSchema, PartialEq, Eq,
+)]
+#[serde(tag = "type", rename_all = "snake_case")]
+pub enum HostPhase2DesiredContents {
+    /// Do not change the current contents.
+    ///
+    /// We use this value when we've detected a sled has been mupdated (and we
+    /// don't want to overwrite phase 2 images until we understand how to
+    /// recover from that mupdate) and as the default value when reading an
+    /// [`OmicronSledConfig`] that was ledgered before this concept existed.
+    CurrentContents,
+
+    /// Set the phase 2 slot to the given artifact.
+    ///
+    /// The artifact will come from an unpacked and distributed TUF repo.
+    Artifact { hash: ArtifactHash },
+}
+
+/// Describes the desired contents for both host phase 2 slots.
+#[derive(Clone, Debug, Deserialize, Serialize, JsonSchema, PartialEq, Eq)]
+#[serde(rename_all = "snake_case")]
+pub struct HostPhase2DesiredSlots {
+    pub slot_a: HostPhase2DesiredContents,
+    pub slot_b: HostPhase2DesiredContents,
+}
+
+impl HostPhase2DesiredSlots {
+    /// Return a `HostPhase2DesiredSlots` with both slots set to
+    /// [`HostPhase2DesiredContents::CurrentContents`]; i.e., "make no changes
+    /// to the current contents of either slot".
+    pub const fn current_contents() -> Self {
+        Self {
+            slot_a: HostPhase2DesiredContents::CurrentContents,
+            slot_b: HostPhase2DesiredContents::CurrentContents,
+        }
+    }
+}
+
 /// Describes the set of Reconfigurator-managed configuration elements of a sled
 #[derive(Clone, Debug, Deserialize, Serialize, JsonSchema, PartialEq, Eq)]
 pub struct OmicronSledConfig {
@@ -584,6 +625,8 @@ pub struct OmicronSledConfig {
     pub datasets: IdMap<DatasetConfig>,
     pub zones: IdMap<OmicronZoneConfig>,
     pub remove_mupdate_override: Option<MupdateOverrideUuid>,
+    #[serde(default = "HostPhase2DesiredSlots::current_contents")]
+    pub host_phase_2: HostPhase2DesiredSlots,
 }
 
 impl Default for OmicronSledConfig {
@@ -594,6 +637,7 @@ impl Default for OmicronSledConfig {
             datasets: IdMap::default(),
             zones: IdMap::default(),
             remove_mupdate_override: None,
+            host_phase_2: HostPhase2DesiredSlots::current_contents(),
         }
     }
 }

nexus/db-model/src/inventory.rs

@@ -44,6 +44,8 @@ use nexus_db_schema::schema::{
 use nexus_sled_agent_shared::inventory::BootImageHeader;
 use nexus_sled_agent_shared::inventory::BootPartitionDetails;
 use nexus_sled_agent_shared::inventory::ConfigReconcilerInventoryStatus;
+use nexus_sled_agent_shared::inventory::HostPhase2DesiredContents;
+use nexus_sled_agent_shared::inventory::HostPhase2DesiredSlots;
 use nexus_sled_agent_shared::inventory::MupdateOverrideBootInventory;
 use nexus_sled_agent_shared::inventory::MupdateOverrideInventory;
 use nexus_sled_agent_shared::inventory::MupdateOverrideNonBootInventory;
@@ -1944,6 +1946,9 @@ pub struct InvOmicronSledConfig {
     pub id: DbTypedUuid<OmicronSledConfigKind>,
     pub generation: Generation,
     pub remove_mupdate_override: Option<DbTypedUuid<MupdateOverrideKind>>,
+
+    #[diesel(embed)]
+    pub host_phase_2: DbHostPhase2DesiredSlots,
 }
 
 impl InvOmicronSledConfig {
@@ -1952,12 +1957,51 @@ impl InvOmicronSledConfig {
         id: OmicronSledConfigUuid,
         generation: external::Generation,
         remove_mupdate_override: Option<MupdateOverrideUuid>,
+        host_phase_2: HostPhase2DesiredSlots,
     ) -> Self {
         Self {
             inv_collection_id: inv_collection_id.into(),
             id: id.into(),
             generation: Generation(generation),
             remove_mupdate_override: remove_mupdate_override.map(From::from),
+            host_phase_2: host_phase_2.into(),
+        }
+    }
+}
+
+#[derive(Queryable, Clone, Debug, Selectable, Insertable)]
+#[diesel(table_name = inv_omicron_sled_config)]
+pub struct DbHostPhase2DesiredSlots {
+    pub host_phase_2_desired_slot_a: Option<ArtifactHash>,
+    pub host_phase_2_desired_slot_b: Option<ArtifactHash>,
+}
+
+impl From<HostPhase2DesiredSlots> for DbHostPhase2DesiredSlots {
+    fn from(value: HostPhase2DesiredSlots) -> Self {
+        let remap = |desired| match desired {
+            HostPhase2DesiredContents::CurrentContents => None,
+            HostPhase2DesiredContents::Artifact { hash } => {
+                Some(ArtifactHash(hash))
+            }
+        };
+        Self {
+            host_phase_2_desired_slot_a: remap(value.slot_a),
+            host_phase_2_desired_slot_b: remap(value.slot_b),
+        }
+    }
+}
+
+impl From<DbHostPhase2DesiredSlots> for HostPhase2DesiredSlots {
+    fn from(value: DbHostPhase2DesiredSlots) -> Self {
+        let remap = |maybe_artifact| match maybe_artifact {
+            None => HostPhase2DesiredContents::CurrentContents,
+            Some(ArtifactHash(hash)) => {
+                HostPhase2DesiredContents::Artifact { hash }
+            }
+        };
+        Self {
+            slot_a: remap(value.host_phase_2_desired_slot_a),
+            slot_b: remap(value.host_phase_2_desired_slot_b),
         }
     }
 }

nexus/db-model/src/schema_versions.rs

@@ -16,7 +16,7 @@ use std::{collections::BTreeMap, sync::LazyLock};
 ///
 /// This must be updated when you change the database schema.  Refer to
 /// schema/crdb/README.adoc in the root of this repository for details.
-pub const SCHEMA_VERSION: Version = Version::new(157, 0, 0);
+pub const SCHEMA_VERSION: Version = Version::new(158, 0, 0);
 
 /// List of all past database schema versions, in *reverse* order
 ///
@@ -28,6 +28,7 @@ static KNOWN_VERSIONS: LazyLock<Vec<KnownVersion>> = LazyLock::new(|| {
         // |  leaving the first copy as an example for the next person.
         // v
         // KnownVersion::new(next_int, "unique-dirname-with-the-sql-files"),
+        KnownVersion::new(158, "sled-config-desired-host-phase-2"),
         KnownVersion::new(157, "user-data-export"),
         KnownVersion::new(156, "boot-partitions-inventory"),
         KnownVersion::new(155, "vpc-firewall-icmp"),

nexus/db-queries/src/db/datastore/inventory.rs

@@ -2774,6 +2774,7 @@ impl DataStore {
                             disks: IdMap::default(),
                             datasets: IdMap::default(),
                             zones: IdMap::default(),
+                            host_phase_2: sled_config.host_phase_2.into(),
                         },
                     });
                 }
@@ -3871,6 +3872,7 @@ impl ConfigReconcilerRows {
             sled_config_id,
             config.generation,
             config.remove_mupdate_override,
+            config.host_phase_2.clone(),
         ));
         self.disks.extend(config.disks.iter().map(|disk| {
             InvOmicronSledConfigDisk::new(

nexus/db-schema/src/schema.rs

@@ -1788,6 +1788,8 @@ table! {
 
         generation -> Int8,
         remove_mupdate_override -> Nullable<Uuid>,
+        host_phase_2_desired_slot_a -> Nullable<Text>,
+        host_phase_2_desired_slot_b -> Nullable<Text>,
     }
 }
 

nexus/inventory/src/collector.rs

@@ -410,6 +410,7 @@ mod test {
     use gateway_messages::SpPort;
     use id_map::IdMap;
     use nexus_sled_agent_shared::inventory::ConfigReconcilerInventoryStatus;
+    use nexus_sled_agent_shared::inventory::HostPhase2DesiredSlots;
     use nexus_sled_agent_shared::inventory::OmicronSledConfig;
     use nexus_sled_agent_shared::inventory::OmicronZoneConfig;
     use nexus_sled_agent_shared::inventory::OmicronZoneImageSource;
@@ -434,6 +435,7 @@ mod test {
             datasets,
             zones,
             remove_mupdate_override,
+            host_phase_2,
         } = config;
 
         writeln!(s, "        generation: {generation}").unwrap();
@@ -442,6 +444,11 @@ mod test {
             "        remove_mupdate_override: {remove_mupdate_override:?}"
         )
         .unwrap();
+        {
+            let HostPhase2DesiredSlots { slot_a, slot_b } = host_phase_2;
+            writeln!(s, "        host_phase_2.slot_a: {slot_a:?}").unwrap();
+            writeln!(s, "        host_phase_2.slot_b: {slot_b:?}").unwrap();
+        }
         for disk in disks {
             writeln!(
                 s,
@@ -687,6 +694,7 @@ mod test {
                 .into_iter()
                 .collect(),
                 remove_mupdate_override: None,
+                host_phase_2: HostPhase2DesiredSlots::current_contents(),
             })
             .await
             .expect("failed to write initial zone version to fake sled agent");

nexus/inventory/src/examples.rs

@@ -21,6 +21,7 @@ use nexus_sled_agent_shared::inventory::BootImageHeader;
 use nexus_sled_agent_shared::inventory::BootPartitionDetails;
 use nexus_sled_agent_shared::inventory::ConfigReconcilerInventory;
 use nexus_sled_agent_shared::inventory::ConfigReconcilerInventoryStatus;
+use nexus_sled_agent_shared::inventory::HostPhase2DesiredSlots;
 use nexus_sled_agent_shared::inventory::Inventory;
 use nexus_sled_agent_shared::inventory::InventoryDataset;
 use nexus_sled_agent_shared::inventory::InventoryDisk;
@@ -345,20 +346,23 @@ pub fn representative() -> Representative {
         datasets: Default::default(),
         zones: sled14.zones.into_iter().collect(),
         remove_mupdate_override: None,
+        host_phase_2: HostPhase2DesiredSlots::current_contents(),
     };
     let sled16 = OmicronSledConfig {
         generation: sled16.generation,
         disks: Default::default(),
         datasets: Default::default(),
         zones: sled16.zones.into_iter().collect(),
         remove_mupdate_override: None,
+        host_phase_2: HostPhase2DesiredSlots::current_contents(),
     };
     let sled17 = OmicronSledConfig {
         generation: sled17.generation,
         disks: Default::default(),
         datasets: Default::default(),
         zones: sled17.zones.into_iter().collect(),
         remove_mupdate_override: None,
+        host_phase_2: HostPhase2DesiredSlots::current_contents(),
     };
 
     // Create iterator producing fixed IDs.

nexus/test-utils/src/lib.rs

@@ -34,6 +34,7 @@ use nexus_config::MgdConfig;
 use nexus_config::NUM_INITIAL_RESERVED_IP_ADDRESSES;
 use nexus_config::NexusConfig;
 use nexus_db_queries::db::pub_test_utils::crdb;
+use nexus_sled_agent_shared::inventory::HostPhase2DesiredSlots;
 use nexus_sled_agent_shared::inventory::OmicronSledConfig;
 use nexus_sled_agent_shared::inventory::OmicronZoneDataset;
 use nexus_sled_agent_shared::recovery_silo::RecoverySiloConfig;
@@ -1147,6 +1148,7 @@ impl<'a, N: NexusServer> ControlPlaneTestContextBuilder<'a, N> {
                     .map(From::from)
                     .collect(),
                 remove_mupdate_override: None,
+                host_phase_2: HostPhase2DesiredSlots::current_contents(),
             })
             .await
             .expect("Failed to configure sled agent with our zones");
@@ -1186,6 +1188,7 @@ impl<'a, N: NexusServer> ControlPlaneTestContextBuilder<'a, N> {
                     datasets: IdMap::default(),
                     zones: IdMap::default(),
                     remove_mupdate_override: None,
+                    host_phase_2: HostPhase2DesiredSlots::current_contents(),
                 })
                 .await
                 .expect("Failed to configure sled agent with our zones");

nexus/types/src/deployment.rs

@@ -20,6 +20,7 @@ pub use crate::inventory::ZpoolName;
 use blueprint_diff::ClickhouseClusterConfigDiffTablesForSingleBlueprint;
 use blueprint_display::BpDatasetsTableSchema;
 use daft::Diffable;
+use nexus_sled_agent_shared::inventory::HostPhase2DesiredSlots;
 use nexus_sled_agent_shared::inventory::OmicronSledConfig;
 use nexus_sled_agent_shared::inventory::OmicronZoneConfig;
 use nexus_sled_agent_shared::inventory::OmicronZoneImageSource;
@@ -743,6 +744,9 @@ impl BlueprintSledConfig {
                 })
                 .collect(),
             remove_mupdate_override: self.remove_mupdate_override,
+            // TODO BlueprintSledConfig should have a corresponding field.
+            // https://github.com/oxidecomputer/omicron/issues/8542
+            host_phase_2: HostPhase2DesiredSlots::current_contents(),
         }
     }