[sled-agent-config-reconciler] Add separate OmicronDatasets type (#8218)

This is more consistent with how the reconciler remembers zones and
disks. This is almost all just moving code around. The only nontrivial
changes are:

1. In the spot where we ought to delete datasets, we at least "forget"
them in-memory and log an error about leaking the ZFS dataset
2. The dataset serialization task remembers fewer details about datasets
it ensured (just the names - that's all we need for the error checking
it does when dealing with nested datasets)

Staged on top of #8064. PR 1 of 3 working towards #8173.

Author: jgallagher

Cargo.lock

@@ -7351,30 +7351,28 @@ fn inv_collection_print_sleds(collection: &Collection) {
                     "LAST RECONCILED CONFIG",
                     &last_reconciliation.last_reconciled_config,
                 );
-                let disk_errs = collect_config_reconciler_errors(
-                    &last_reconciliation.external_disks,
-                );
-                let dataset_errs = collect_config_reconciler_errors(
-                    &last_reconciliation.datasets,
-                );
-                let zone_errs = collect_config_reconciler_errors(
-                    &last_reconciliation.zones,
-                );
-                for (label, errs) in [
-                    ("disk", disk_errs),
-                    ("dataset", dataset_errs),
-                    ("zone", zone_errs),
-                ] {
-                    if errs.is_empty() {
-                        println!("    all {label}s reconciled successfully");
-                    } else {
-                        println!(
-                            "    {} {label} reconciliation errors:",
-                            errs.len()
-                        );
-                        for err in errs {
-                            println!("      {err}");
-                        }
+            }
+            let disk_errs = collect_config_reconciler_errors(
+                &last_reconciliation.external_disks,
+            );
+            let dataset_errs =
+                collect_config_reconciler_errors(&last_reconciliation.datasets);
+            let zone_errs =
+                collect_config_reconciler_errors(&last_reconciliation.zones);
+            for (label, errs) in [
+                ("disk", disk_errs),
+                ("dataset", dataset_errs),
+                ("zone", zone_errs),
+            ] {
+                if errs.is_empty() {
+                    println!("        all {label}s reconciled successfully");
+                } else {
+                    println!(
+                        "        {} {label} reconciliation errors:",
+                        errs.len()
+                    );
+                    for err in errs {
+                        println!("          {err}");
                     }
                 }
             }

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

@@ -12,6 +12,7 @@ use chrono::{DateTime, Utc};
 use daft::Diffable;
 use id_map::IdMap;
 use id_map::IdMappable;
+use omicron_common::disk::{DatasetKind, DatasetName};
 use omicron_common::ledger::Ledgerable;
 use omicron_common::{
     api::{
@@ -327,6 +328,10 @@ impl OmicronZoneConfig {
             Some(self.id),
         )
     }
+
+    pub fn dataset_name(&self) -> Option<DatasetName> {
+        self.zone_type.dataset_name()
+    }
 }
 
 /// Describes a persistent ZFS dataset associated with an Omicron zone
@@ -613,6 +618,41 @@ impl OmicronZoneType {
             | OmicronZoneType::Oximeter { .. } => None,
         }
     }
+
+    /// If this kind of zone has an associated dataset, return the dataset's
+    /// name. Otherwise, return `None`.
+    pub fn dataset_name(&self) -> Option<DatasetName> {
+        let (dataset, dataset_kind) = match self {
+            OmicronZoneType::BoundaryNtp { .. }
+            | OmicronZoneType::InternalNtp { .. }
+            | OmicronZoneType::Nexus { .. }
+            | OmicronZoneType::Oximeter { .. }
+            | OmicronZoneType::CruciblePantry { .. } => None,
+            OmicronZoneType::Clickhouse { dataset, .. } => {
+                Some((dataset, DatasetKind::Clickhouse))
+            }
+            OmicronZoneType::ClickhouseKeeper { dataset, .. } => {
+                Some((dataset, DatasetKind::ClickhouseKeeper))
+            }
+            OmicronZoneType::ClickhouseServer { dataset, .. } => {
+                Some((dataset, DatasetKind::ClickhouseServer))
+            }
+            OmicronZoneType::CockroachDb { dataset, .. } => {
+                Some((dataset, DatasetKind::Cockroach))
+            }
+            OmicronZoneType::Crucible { dataset, .. } => {
+                Some((dataset, DatasetKind::Crucible))
+            }
+            OmicronZoneType::ExternalDns { dataset, .. } => {
+                Some((dataset, DatasetKind::ExternalDns))
+            }
+            OmicronZoneType::InternalDns { dataset, .. } => {
+                Some((dataset, DatasetKind::InternalDns))
+            }
+        }?;
+
+        Some(DatasetName::new(dataset.pool_name, dataset_kind))
+    }
 }
 
 /// Like [`OmicronZoneType`], but without any associated data.

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

@@ -1,7 +1,7 @@
 {
   "$schema": "http://json-schema.org/draft-07/schema#",
   "title": "OmicronZonesConfigLocal",
-  "description": "Combines the Nexus-provided `OmicronZonesConfig` (which describes what Nexus wants for all of its zones) with the locally-determined configuration for these zones.",
+  "description": "Legacy type of the ledgered zone config.",
   "type": "object",
   "required": [
     "ledger_generation",
@@ -10,20 +10,10 @@
   ],
   "properties": {
     "ledger_generation": {
-      "description": "ledger-managed generation number\n\nThis generation is managed by the ledger facility itself.  It's bumped whenever we write a new ledger.  In practice, we don't currently have any reason to bump this _for a given Omicron generation_ so it's somewhat redundant.  In principle, if we needed to modify the ledgered configuration due to some event that doesn't change the Omicron config (e.g., if we wanted to move the root filesystem to a different path), we could do that by bumping this generation.",
-      "allOf": [
-        {
-          "$ref": "#/definitions/Generation"
-        }
-      ]
+      "$ref": "#/definitions/Generation"
     },
     "omicron_generation": {
-      "description": "generation of the Omicron-provided part of the configuration\n\nThis generation number is outside of Sled Agent's control.  We store exactly what we were given and use this number to decide when to fail requests to establish an outdated configuration.\n\nYou can think of this as a major version number, with `ledger_generation` being a minor version number.  See `is_newer_than()`.",
-      "allOf": [
-        {
-          "$ref": "#/definitions/Generation"
-        }
-      ]
+      "$ref": "#/definitions/Generation"
     },
     "zones": {
       "type": "array",
@@ -269,7 +259,6 @@
       }
     },
     "OmicronZoneConfigLocal": {
-      "description": "Combines the Nexus-provided `OmicronZoneConfig` (which describes what Nexus wants for this zone) with any locally-determined configuration (like the path to the root filesystem)",
       "type": "object",
       "required": [
         "root",

schema/all-zones-requests.json

@@ -45,6 +45,7 @@ expectorate.workspace = true
 illumos-utils = { workspace = true, features = ["testing"] }
 omicron-test-utils.workspace = true
 proptest.workspace = true
+schemars.workspace = true
 scopeguard.workspace = true
 serde_json.workspace = true
 sled-storage = { workspace = true, features = ["testing"] }