PlanningReport: Use JSON-safe map keys (#9080)

(This also includes #9077 to avoid failing CI.)

---------

Co-authored-by: iliana etaoin <iliana@oxide.computer>

Author: jgallagher

dev-tools/reconfigurator-cli/tests/output/cmds-target-release-stdout

@@ -7944,9 +7944,9 @@ planning report:
 * skipping noop zone image source check on sled d81c6a84-79b8-4958-ae41-ea46c9b19763: all 9 zones are already from artifacts
 * 3 remaining out-of-date zones
 * 3 zones waiting to be expunged:
-  * zone 0c71b3b2-6ceb-4e8f-b020-b08675e83038 (nexus): image out-of-date, but zone's nexus_generation 1 is still active
-  * zone 3eeb8d49-eb1a-43f8-bb64-c2338421c2c6 (nexus): image out-of-date, but zone's nexus_generation 1 is still active
-  * zone 466a9f29-62bf-4e63-924a-b9efdb86afec (nexus): image out-of-date, but zone's nexus_generation 1 is still active
+  * zone 0c71b3b2-6ceb-4e8f-b020-b08675e83038: nexus image out-of-date, but nexus_generation 1 is still active
+  * zone 3eeb8d49-eb1a-43f8-bb64-c2338421c2c6: nexus image out-of-date, but nexus_generation 1 is still active
+  * zone 466a9f29-62bf-4e63-924a-b9efdb86afec: nexus image out-of-date, but nexus_generation 1 is still active
 * updating top-level nexus_generation to: 2
 
 
@@ -8035,9 +8035,9 @@ planning report:
 * skipping noop zone image source check on sled d81c6a84-79b8-4958-ae41-ea46c9b19763: all 9 zones are already from artifacts
 * 3 remaining out-of-date zones
 * 3 zones waiting to be expunged:
-  * zone 0c71b3b2-6ceb-4e8f-b020-b08675e83038 (nexus): image out-of-date, but zone's nexus_generation 1 is still active
-  * zone 3eeb8d49-eb1a-43f8-bb64-c2338421c2c6 (nexus): image out-of-date, but zone's nexus_generation 1 is still active
-  * zone 466a9f29-62bf-4e63-924a-b9efdb86afec (nexus): image out-of-date, but zone's nexus_generation 1 is still active
+  * zone 0c71b3b2-6ceb-4e8f-b020-b08675e83038: nexus image out-of-date, but nexus_generation 1 is still active
+  * zone 3eeb8d49-eb1a-43f8-bb64-c2338421c2c6: nexus image out-of-date, but nexus_generation 1 is still active
+  * zone 466a9f29-62bf-4e63-924a-b9efdb86afec: nexus image out-of-date, but nexus_generation 1 is still active
 
 
 

nexus/src/app/background/tasks/blueprint_planner.rs

@@ -22,6 +22,7 @@ use omicron_common::api::external::LookupType;
 use omicron_uuid_kinds::CollectionUuid;
 use omicron_uuid_kinds::GenericUuid as _;
 use serde_json::json;
+use slog_error_chain::InlineErrorChain;
 use std::sync::Arc;
 use tokio::sync::watch::{self, Receiver, Sender};
 
@@ -292,7 +293,16 @@ impl BackgroundTask for BlueprintPlanner {
         &'a mut self,
         opctx: &'a OpContext,
     ) -> BoxFuture<'a, serde_json::Value> {
-        Box::pin(async move { json!(self.plan(opctx).await) })
+        Box::pin(async move {
+            let status = self.plan(opctx).await;
+            match serde_json::to_value(status) {
+                Ok(val) => val,
+                Err(err) => json!({
+                    "error": format!("could not serialize task status: {}",
+                                     InlineErrorChain::new(&err)),
+                }),
+            }
+        })
     }
 }
 

nexus/types/src/deployment/planning_report.rs

@@ -848,8 +848,8 @@ pub struct PlanningZoneUpdatesStepReport {
     pub out_of_date_zones: BTreeMap<SledUuid, Vec<PlanningOutOfDateZone>>,
     pub expunged_zones: BTreeMap<SledUuid, Vec<BlueprintZoneConfig>>,
     pub updated_zones: BTreeMap<SledUuid, Vec<BlueprintZoneConfig>>,
-    pub unsafe_zones: BTreeMap<BlueprintZoneConfig, ZoneUnsafeToShutdown>,
-    pub waiting_zones: BTreeMap<BlueprintZoneConfig, ZoneWaitingToExpunge>,
+    pub unsafe_zones: BTreeMap<OmicronZoneUuid, ZoneUnsafeToShutdown>,
+    pub waiting_zones: BTreeMap<OmicronZoneUuid, ZoneWaitingToExpunge>,
 }
 
 impl PlanningZoneUpdatesStepReport {
@@ -934,15 +934,15 @@ impl PlanningZoneUpdatesStepReport {
         zone: &BlueprintZoneConfig,
         reason: ZoneUnsafeToShutdown,
     ) {
-        self.unsafe_zones.insert(zone.clone(), reason);
+        self.unsafe_zones.insert(zone.id, reason);
     }
 
     pub fn waiting_zone(
         &mut self,
         zone: &BlueprintZoneConfig,
         reason: ZoneWaitingToExpunge,
     ) {
-        self.waiting_zones.insert(zone.clone(), reason);
+        self.waiting_zones.insert(zone.id, reason);
     }
 }
 
@@ -1001,28 +1001,16 @@ impl fmt::Display for PlanningZoneUpdatesStepReport {
         if !unsafe_zones.is_empty() {
             let (n, s) = plural_map(unsafe_zones);
             writeln!(f, "* {n} zone{s} not ready to shut down safely:")?;
-            for (zone, reason) in unsafe_zones.iter() {
-                writeln!(
-                    f,
-                    "  * zone {} ({}): {}",
-                    zone.id,
-                    zone.zone_type.kind().report_str(),
-                    reason,
-                )?;
+            for (zone_id, reason) in unsafe_zones.iter() {
+                writeln!(f, "  * zone {zone_id}: {reason}")?;
             }
         }
 
         if !waiting_zones.is_empty() {
             let (n, s) = plural_map(waiting_zones);
             writeln!(f, "* {n} zone{s} waiting to be expunged:")?;
-            for (zone, reason) in waiting_zones.iter() {
-                writeln!(
-                    f,
-                    "  * zone {} ({}): {}",
-                    zone.id,
-                    zone.zone_type.kind().report_str(),
-                    reason,
-                )?;
+            for (zone_id, reason) in waiting_zones.iter() {
+                writeln!(f, "  * zone {zone_id}: {reason}")?;
             }
         }
 
@@ -1076,7 +1064,9 @@ pub enum ZoneUnsafeToShutdown {
 impl fmt::Display for ZoneUnsafeToShutdown {
     fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
         match self {
-            Self::Cockroachdb { reason } => write!(f, "{reason}"),
+            Self::Cockroachdb { reason } => {
+                write!(f, "cockroach unsafe to shut down: {reason}")
+            }
             Self::BoundaryNtp {
                 total_boundary_ntp_zones: t,
                 synchronized_count: s,
@@ -1107,7 +1097,7 @@ impl fmt::Display for ZoneWaitingToExpunge {
             Self::Nexus { zone_generation } => {
                 write!(
                     f,
-                    "image out-of-date, but zone's nexus_generation \
+                    "nexus image out-of-date, but nexus_generation \
                      {zone_generation} is still active"
                 )
             }