From ed22cf3f0e719790e9131f3a6d2186cd07686563 Mon Sep 17 00:00:00 2001
From: lif <>
Date: Tue, 6 Feb 2024 10:14:44 +0000
Subject: [PATCH] wip: start adding test that will eventually validate behavior
 of responding to TimedOut answers from nexus

---
 openapi/nexus-internal.json |   2 +-
 sled-agent/src/instance.rs  | 127 ++++++++++++++++++++++++++++++++----
 2 files changed, 114 insertions(+), 15 deletions(-)

diff --git a/openapi/nexus-internal.json b/openapi/nexus-internal.json
index b84bc9708a5..dbef2b281d8 100644
--- a/openapi/nexus-internal.json
+++ b/openapi/nexus-internal.json
@@ -514,7 +514,7 @@
     "/instances/{instance_id}/creation-success": {
       "put": {
         "summary": "Asynchronously report the successful result of certain instance_put calls",
-        "description": "(such as the potentially long-running one made during instance creation)",
+        "description": "(such as the potentially long-running one made during instance creation). If Nexus has already marked the instance as failed, this returns `TimedOut` so the sled-agent making the request knows to terminate the instance.",
         "operationId": "cpapi_handle_instance_put_success",
         "parameters": [
           {
diff --git a/sled-agent/src/instance.rs b/sled-agent/src/instance.rs
index 45df6bea2c0..0907ad1da7d 100644
--- a/sled-agent/src/instance.rs
+++ b/sled-agent/src/instance.rs
@@ -1470,7 +1470,33 @@ mod tests {
     ) -> Instance {
         let id = Uuid::new_v4();
         let propolis_id = Uuid::new_v4();
+
         let ticket = InstanceTicket::new_without_manager_for_test(id);
+
+        let initial_state =
+            fake_instance_initial_state(propolis_id, propolis_addr);
+
+        let services = fake_instance_manager_services(
+            logctx,
+            storage_handle,
+            nexus_client_with_resolver,
+        );
+
+        Instance::new(
+            logctx.log.new(o!("component" => "Instance")),
+            id,
+            propolis_id,
+            ticket,
+            initial_state,
+            services,
+        )
+        .unwrap()
+    }
+
+    fn fake_instance_initial_state(
+        propolis_id: Uuid,
+        propolis_addr: SocketAddr,
+    ) -> InstanceInitialState {
         let hardware = InstanceHardware {
             properties: InstanceProperties {
                 ncpus: InstanceCpuCount(1),
@@ -1495,7 +1521,7 @@ mod tests {
             cloud_init_bytes: None,
         };
 
-        let initial_state = InstanceInitialState {
+        InstanceInitialState {
             hardware,
             instance_runtime: InstanceRuntimeState {
                 propolis_id: Some(propolis_id),
@@ -1510,8 +1536,14 @@ mod tests {
                 time_updated: Default::default(),
             },
             propolis_addr,
-        };
+        }
+    }
 
+    fn fake_instance_manager_services(
+        logctx: &LogContext,
+        storage_handle: StorageHandle,
+        nexus_client_with_resolver: NexusClientWithResolver,
+    ) -> InstanceManagerServices {
         let vnic_allocator =
             VnicAllocator::new("Foo", Etherstub("mystub".to_string()));
         let port_manager = PortManager::new(
@@ -1526,24 +1558,14 @@ mod tests {
             cleanup_context,
         );
 
-        let services = InstanceManagerServices {
+        InstanceManagerServices {
             nexus_client: nexus_client_with_resolver,
             vnic_allocator,
             port_manager,
             storage: storage_handle,
             zone_bundler,
             zone_builder_factory: ZoneBuilderFactory::fake(),
-        };
-
-        Instance::new(
-            logctx.log.new(o!("component" => "Instance")),
-            id,
-            propolis_id,
-            ticket,
-            initial_state,
-            services,
-        )
-        .unwrap()
+        }
     }
 
     #[tokio::test]
@@ -1724,4 +1746,81 @@ mod tests {
 
         logctx.cleanup_successful();
     }
+
+    #[tokio::test]
+    async fn test_instance_manager_creation_nexus_timeout() {
+        let logctx = omicron_test_utils::dev::test_setup_log(
+            "test_instance_manager_creation_nexus_timeout",
+        );
+
+        let storage_handle = fake_storage_manager_with_u2().await;
+
+        let (nexus_client, nexus_server, state_rx) = fake_nexus_server(&logctx);
+
+        let (_dns_server, resolver, _dns_config_dir) =
+            timeout(TIMEOUT_DURATION, dns_server(&logctx, &nexus_server))
+                .await
+                .expect("timed out making DNS server and Resolver");
+
+        let nexus_client_with_resolver =
+            NexusClientWithResolver::new_with_client(nexus_client, resolver);
+
+        let InstanceManagerServices {
+            nexus_client,
+            vnic_allocator: _,
+            port_manager,
+            storage,
+            zone_bundler,
+            zone_builder_factory,
+        } = fake_instance_manager_services(
+            &logctx,
+            storage_handle,
+            nexus_client_with_resolver,
+        );
+
+        let etherstub = Etherstub("mystub".to_string());
+
+        let mgr = crate::instance_manager::InstanceManager::new(
+            logctx.log.new(o!("component" => "InstanceManager")),
+            nexus_client,
+            etherstub,
+            port_manager,
+            storage,
+            zone_bundler,
+            zone_builder_factory,
+        )
+        .unwrap();
+
+        let (propolis_server, _propolis_client) =
+            propolis_mock_server(&logctx.log);
+        let propolis_addr = propolis_server.local_addr();
+
+        // automock'd things used during this test
+        let _mock_vnic_contexts = mock_vnic_contexts();
+        let _mock_zone_contexts = mock_zone_contexts();
+
+        let instance_id = Uuid::new_v4();
+        let propolis_id = Uuid::new_v4();
+        let InstanceInitialState {
+            hardware,
+            instance_runtime,
+            vmm_runtime,
+            propolis_addr,
+        } = fake_instance_initial_state(propolis_id.clone(), propolis_addr);
+
+        mgr.ensure_registered(
+            instance_id,
+            propolis_id,
+            hardware,
+            instance_runtime,
+            vmm_runtime,
+            propolis_addr,
+        )
+        .await
+        .unwrap();
+
+        todo!();
+
+        logctx.cleanup_successful();
+    }
 }