blueprint_planner background task

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

@@ -46,6 +46,7 @@ use nexus_types::deployment::ClickhousePolicy;
 use nexus_types::deployment::OximeterReadMode;
 use nexus_types::deployment::OximeterReadPolicy;
 use nexus_types::internal_api::background::AbandonedVmmReaperStatus;
+use nexus_types::internal_api::background::BlueprintPlannerStatus;
 use nexus_types::internal_api::background::BlueprintRendezvousStatus;
 use nexus_types::internal_api::background::InstanceReincarnationStatus;
 use nexus_types::internal_api::background::InstanceUpdaterStatus;
@@ -1056,6 +1057,9 @@ fn print_task_details(bgtask: &BackgroundTask, details: &serde_json::Value) {
         "abandoned_vmm_reaper" => {
             print_task_abandoned_vmm_reaper(details);
         }
+        "blueprint_planner" => {
+            print_task_blueprint_planner(details);
+        }
         "blueprint_executor" => {
             print_task_blueprint_executor(details);
         }
@@ -1204,6 +1208,44 @@ fn print_task_abandoned_vmm_reaper(details: &serde_json::Value) {
     };
 }
 
+fn print_task_blueprint_planner(details: &serde_json::Value) {
+    let status =
+        match serde_json::from_value::<BlueprintPlannerStatus>(details.clone())
+        {
+            Ok(status) => status,
+            Err(error) => {
+                eprintln!(
+                    "warning: failed to interpret task details: {:?}: {:?}",
+                    error, details
+                );
+                return;
+            }
+        };
+    match status {
+        BlueprintPlannerStatus::Disabled => {
+            println!("    blueprint planning explicitly disabled by config!");
+        }
+        BlueprintPlannerStatus::Error(error) => {
+            println!("    task did not complete successfully: {error}");
+        }
+        BlueprintPlannerStatus::Unchanged { parent_blueprint_id } => {
+            println!("    plan unchanged from parent {parent_blueprint_id}");
+        }
+        BlueprintPlannerStatus::Planned { parent_blueprint_id, error } => {
+            println!(
+                "    planned new blueprint from parent {parent_blueprint_id}, \
+                     but could not make it the target: {error}"
+            );
+        }
+        BlueprintPlannerStatus::Targeted { blueprint_id, .. } => {
+            println!(
+                "    planned new blueprint {blueprint_id}, \
+                     and made it the current target"
+            );
+        }
+    }
+}
+
 fn print_task_blueprint_executor(details: &serde_json::Value) {
     let mut value = details.clone();
     // Extract and remove the event report. (If we don't do this, the

dev-tools/omdb/tests/env.out

@@ -47,6 +47,10 @@ task: "blueprint_loader"
     Loads the current target blueprint from the DB
 
 
+task: "blueprint_planner"
+    Updates the target blueprint
+
+
 task: "blueprint_rendezvous"
     reconciles blueprints and inventory collection, updating Reconfigurator-
     owned rendezvous tables that other subsystems consume
@@ -243,6 +247,10 @@ task: "blueprint_loader"
     Loads the current target blueprint from the DB
 
 
+task: "blueprint_planner"
+    Updates the target blueprint
+
+
 task: "blueprint_rendezvous"
     reconciles blueprints and inventory collection, updating Reconfigurator-
     owned rendezvous tables that other subsystems consume
@@ -426,6 +434,10 @@ task: "blueprint_loader"
     Loads the current target blueprint from the DB
 
 
+task: "blueprint_planner"
+    Updates the target blueprint
+
+
 task: "blueprint_rendezvous"
     reconciles blueprints and inventory collection, updating Reconfigurator-
     owned rendezvous tables that other subsystems consume

dev-tools/omdb/tests/successes.out

@@ -259,6 +259,10 @@ task: "blueprint_loader"
     Loads the current target blueprint from the DB
 
 
+task: "blueprint_planner"
+    Updates the target blueprint
+
+
 task: "blueprint_rendezvous"
     reconciles blueprints and inventory collection, updating Reconfigurator-
     owned rendezvous tables that other subsystems consume
@@ -521,6 +525,13 @@ task: "bfd_manager"
     started at <REDACTED_TIMESTAMP> (<REDACTED DURATION>s ago) and ran for <REDACTED DURATION>ms
     last completion reported error: failed to resolve addresses for Dendrite services: no record found for Query { name: Name("_dendrite._tcp.control-plane.oxide.internal."), query_type: SRV, query_class: IN }
 
+task: "blueprint_planner"
+  configured period: every <REDACTED_DURATION>m
+  currently executing: no
+  last completed activation: <REDACTED ITERATIONS>, triggered by a dependent task completing
+    started at <REDACTED_TIMESTAMP> (<REDACTED DURATION>s ago) and ran for <REDACTED DURATION>ms
+    blueprint planning explicitly disabled by config!
+
 task: "blueprint_rendezvous"
   configured period: every <REDACTED_DURATION>m
   currently executing: no
@@ -1040,6 +1051,13 @@ task: "bfd_manager"
     started at <REDACTED_TIMESTAMP> (<REDACTED DURATION>s ago) and ran for <REDACTED DURATION>ms
     last completion reported error: failed to resolve addresses for Dendrite services: no record found for Query { name: Name("_dendrite._tcp.control-plane.oxide.internal."), query_type: SRV, query_class: IN }
 
+task: "blueprint_planner"
+  configured period: every <REDACTED_DURATION>m
+  currently executing: no
+  last completed activation: <REDACTED ITERATIONS>, triggered by a dependent task completing
+    started at <REDACTED_TIMESTAMP> (<REDACTED DURATION>s ago) and ran for <REDACTED DURATION>ms
+    blueprint planning explicitly disabled by config!
+
 task: "blueprint_rendezvous"
   configured period: every <REDACTED_DURATION>m
   currently executing: no
@@ -1450,12 +1468,14 @@ parent:    <none>
     oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_external_dns_..........<REDACTED_UUID>...........      ..........<REDACTED_UUID>...........   in service    none    none          off        
     oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_internal_dns_..........<REDACTED_UUID>...........      ..........<REDACTED_UUID>...........   in service    none    none          off        
     oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_nexus_..........<REDACTED_UUID>...........             ..........<REDACTED_UUID>...........   in service    none    none          off        
+    oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_ntp_..........<REDACTED_UUID>...........               ..........<REDACTED_UUID>...........   in service    none    none          off        
 
 
     omicron zones:
     ---------------------------------------------------------------------------------------------------------
     zone type         zone id                                image source      disposition   underlay IP     
     ---------------------------------------------------------------------------------------------------------
+    boundary_ntp      ..........<REDACTED_UUID>...........   install dataset   in service    ::1             
     clickhouse        ..........<REDACTED_UUID>...........   install dataset   in service    ::1             
     cockroach_db      ..........<REDACTED_UUID>...........   install dataset   in service    ::1             
     crucible_pantry   ..........<REDACTED_UUID>...........   install dataset   in service    ::1             
@@ -1552,12 +1572,14 @@ parent:    <none>
     oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_external_dns_..........<REDACTED_UUID>...........      ..........<REDACTED_UUID>...........   in service    none    none          off        
     oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_internal_dns_..........<REDACTED_UUID>...........      ..........<REDACTED_UUID>...........   in service    none    none          off        
     oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_nexus_..........<REDACTED_UUID>...........             ..........<REDACTED_UUID>...........   in service    none    none          off        
+    oxp_..........<REDACTED_UUID>.........../crypt/zone/oxz_ntp_..........<REDACTED_UUID>...........               ..........<REDACTED_UUID>...........   in service    none    none          off        
 
 
     omicron zones:
     ---------------------------------------------------------------------------------------------------------
     zone type         zone id                                image source      disposition   underlay IP     
     ---------------------------------------------------------------------------------------------------------
+    boundary_ntp      ..........<REDACTED_UUID>...........   install dataset   in service    ::1             
     clickhouse        ..........<REDACTED_UUID>...........   install dataset   in service    ::1             
     cockroach_db      ..........<REDACTED_UUID>...........   install dataset   in service    ::1             
     crucible_pantry   ..........<REDACTED_UUID>...........   install dataset   in service    ::1             

docs/reconfigurator.adoc

@@ -124,18 +124,16 @@ The Planner
                +----------+  |  +----------/
                           |  |  |
                           v  v  v
-
-                         "planner"
-              (eventually a background task)
-                             |
-                             v                      no
-                    is a new blueprint necessary? ------> done
+                    planner background task
                              |
-                             | yes
                              v
                     generate a new blueprint
                              |
                              |
+                             v                                                no
+                    is the new blueprint different from the current target? ------> done
+                             |
+                             | yes
                              v
                     commit blueprint to database
                              |

nexus-config/src/nexus_config.rs

@@ -592,19 +592,27 @@ pub struct PhantomDiskConfig {
 #[serde_as]
 #[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
 pub struct BlueprintTasksConfig {
+    /// background planner chicken switch
+    pub disable_planner: bool,
+
     /// period (in seconds) for periodic activations of the background task that
     /// reads the latest target blueprint from the database
     #[serde_as(as = "DurationSeconds<u64>")]
     pub period_secs_load: Duration,
 
+    /// period (in seconds) for periodic activations of the background task that
+    /// plans and updates the target blueprint
+    #[serde_as(as = "DurationSeconds<u64>")]
+    pub period_secs_plan: Duration,
+
     /// period (in seconds) for periodic activations of the background task that
     /// executes the latest target blueprint
     #[serde_as(as = "DurationSeconds<u64>")]
     pub period_secs_execute: Duration,
 
     /// period (in seconds) for periodic activations of the background task that
     /// reconciles the latest blueprint and latest inventory collection into
-    /// Rencofigurator rendezvous tables
+    /// Reconfigurator rendezvous tables
     #[serde_as(as = "DurationSeconds<u64>")]
     pub period_secs_rendezvous: Duration,
 
@@ -1055,7 +1063,9 @@ mod test {
             physical_disk_adoption.period_secs = 30
             decommissioned_disk_cleaner.period_secs = 30
             phantom_disks.period_secs = 30
+            blueprints.disable_planner = true
             blueprints.period_secs_load = 10
+            blueprints.period_secs_plan = 60
             blueprints.period_secs_execute = 60
             blueprints.period_secs_rendezvous = 300
             blueprints.period_secs_collect_crdb_node_ids = 180
@@ -1220,7 +1230,9 @@ mod test {
                             period_secs: Duration::from_secs(30),
                         },
                         blueprints: BlueprintTasksConfig {
+                            disable_planner: true,
                             period_secs_load: Duration::from_secs(10),
+                            period_secs_plan: Duration::from_secs(60),
                             period_secs_execute: Duration::from_secs(60),
                             period_secs_collect_crdb_node_ids:
                                 Duration::from_secs(180),
@@ -1364,7 +1376,9 @@ mod test {
             physical_disk_adoption.period_secs = 30
             decommissioned_disk_cleaner.period_secs = 30
             phantom_disks.period_secs = 30
+            blueprints.disable_planner = true
             blueprints.period_secs_load = 10
+            blueprints.period_secs_plan = 60
             blueprints.period_secs_execute = 60
             blueprints.period_secs_rendezvous = 300
             blueprints.period_secs_collect_crdb_node_ids = 180

nexus/background-task-interface/src/init.rs

@@ -22,6 +22,7 @@ pub struct BackgroundTasks {
     pub task_decommissioned_disk_cleaner: Activator,
     pub task_phantom_disks: Activator,
     pub task_blueprint_loader: Activator,
+    pub task_blueprint_planner: Activator,
     pub task_blueprint_executor: Activator,
     pub task_blueprint_rendezvous: Activator,
     pub task_crdb_node_id_collector: Activator,

nexus/examples/config-second.toml

@@ -118,7 +118,9 @@ phantom_disks.period_secs = 30
 physical_disk_adoption.period_secs = 30
 support_bundle_collector.period_secs = 30
 decommissioned_disk_cleaner.period_secs = 60
+blueprints.disable_planner = true
 blueprints.period_secs_load = 10
+blueprints.period_secs_plan = 60
 blueprints.period_secs_execute = 60
 blueprints.period_secs_rendezvous = 300
 blueprints.period_secs_collect_crdb_node_ids = 180

nexus/examples/config.toml

@@ -104,7 +104,9 @@ phantom_disks.period_secs = 30
 physical_disk_adoption.period_secs = 30
 support_bundle_collector.period_secs = 30
 decommissioned_disk_cleaner.period_secs = 60
+blueprints.disable_planner = true
 blueprints.period_secs_load = 10
+blueprints.period_secs_plan = 60
 blueprints.period_secs_execute = 60
 blueprints.period_secs_rendezvous = 300
 blueprints.period_secs_collect_crdb_node_ids = 180