From 475fb621d43b47ab3f4e9bd69fb21a695d147d73 Mon Sep 17 00:00:00 2001
From: John Gallagher <john@oxidecomputer.com>
Date: Fri, 16 Feb 2024 09:06:59 -0500
Subject: [PATCH] Fix bad service NIC slots (#5080)

This is the second half of the fix for #5056. #5065 (already merged)
fixed _how_ we were getting service NICs with nonzero slot values, and
this PR adds a schema migration to apply a one-time fix to any existing
service NICs with nonzero slot values. This matters to the
Reconfigurator, because currently the NICs sled-agent thinks it has
don't match the NICs recorded in CRDB (differing only by slot number).

Closes #5056.
---
 nexus/db-model/src/schema.rs               | 2 +-
 nexus/db-queries/src/db/mod.rs             | 5 +++++
 nexus/db-queries/src/db/pool_connection.rs | 2 +-
 nexus/tests/integration_tests/schema.rs    | 6 ++++++
 schema/crdb/35.0.0/up.sql                  | 6 ++++++
 schema/crdb/dbinit.sql                     | 2 +-
 6 files changed, 20 insertions(+), 3 deletions(-)
 create mode 100644 schema/crdb/35.0.0/up.sql

diff --git a/nexus/db-model/src/schema.rs b/nexus/db-model/src/schema.rs
index 9ebf7516cf..3d9050b1af 100644
--- a/nexus/db-model/src/schema.rs
+++ b/nexus/db-model/src/schema.rs
@@ -13,7 +13,7 @@ use omicron_common::api::external::SemverVersion;
 ///
 /// This should be updated whenever the schema is changed. For more details,
 /// refer to: schema/crdb/README.adoc
-pub const SCHEMA_VERSION: SemverVersion = SemverVersion::new(34, 0, 0);
+pub const SCHEMA_VERSION: SemverVersion = SemverVersion::new(35, 0, 0);
 
 table! {
     disk (id) {
diff --git a/nexus/db-queries/src/db/mod.rs b/nexus/db-queries/src/db/mod.rs
index 184ebe5f8a..d5262166ee 100644
--- a/nexus/db-queries/src/db/mod.rs
+++ b/nexus/db-queries/src/db/mod.rs
@@ -35,6 +35,11 @@ pub mod subquery;
 pub(crate) mod true_or_cast_error;
 mod update_and_check;
 
+/// Batch statement to disable full table scans.
+// This is `pub` so tests that don't go through our connection pool can disable
+// full table scans the same way pooled connections do.
+pub use pool_connection::DISALLOW_FULL_TABLE_SCAN_SQL;
+
 #[cfg(test)]
 mod test_utils;
 
diff --git a/nexus/db-queries/src/db/pool_connection.rs b/nexus/db-queries/src/db/pool_connection.rs
index d9c50ff26c..66fb125a7c 100644
--- a/nexus/db-queries/src/db/pool_connection.rs
+++ b/nexus/db-queries/src/db/pool_connection.rs
@@ -78,7 +78,7 @@ static CUSTOM_TYPE_KEYS: &'static [&'static str] = &[
 ];
 const CUSTOM_TYPE_SCHEMA: &'static str = "public";
 
-const DISALLOW_FULL_TABLE_SCAN_SQL: &str =
+pub const DISALLOW_FULL_TABLE_SCAN_SQL: &str =
     "set disallow_full_table_scans = on; set large_full_scan_rows = 0;";
 
 #[derive(Debug)]
diff --git a/nexus/tests/integration_tests/schema.rs b/nexus/tests/integration_tests/schema.rs
index 2d496fcd8e..1c23f1b842 100644
--- a/nexus/tests/integration_tests/schema.rs
+++ b/nexus/tests/integration_tests/schema.rs
@@ -10,6 +10,7 @@ use nexus_db_model::schema::SCHEMA_VERSION as LATEST_SCHEMA_VERSION;
 use nexus_db_queries::db::datastore::{
     all_sql_for_version_migration, EARLIEST_SUPPORTED_VERSION,
 };
+use nexus_db_queries::db::DISALLOW_FULL_TABLE_SCAN_SQL;
 use nexus_test_utils::{db, load_test_config, ControlPlaneTestContextBuilder};
 use omicron_common::api::external::SemverVersion;
 use omicron_common::api::internal::shared::SwitchLocation;
@@ -118,6 +119,11 @@ async fn apply_update(
 
     let client = crdb.connect().await.expect("failed to connect");
 
+    client
+        .batch_execute(DISALLOW_FULL_TABLE_SCAN_SQL)
+        .await
+        .expect("failed to disallow full table scans");
+
     // We skip this for the earliest supported version because these tables
     // might not exist yet.
     if version != EARLIEST_SUPPORTED_VERSION {
diff --git a/schema/crdb/35.0.0/up.sql b/schema/crdb/35.0.0/up.sql
new file mode 100644
index 0000000000..f50e8dc881
--- /dev/null
+++ b/schema/crdb/35.0.0/up.sql
@@ -0,0 +1,6 @@
+-- This isn't realy a schema migration, but is instead a one-off fix for
+-- incorrect data (https://github.com/oxidecomputer/omicron/issues/5056) after
+-- the root cause for the incorrect data has been addressed.
+UPDATE omicron.public.network_interface
+    SET slot = 0
+    WHERE kind = 'service' AND time_deleted IS NULL;
diff --git a/schema/crdb/dbinit.sql b/schema/crdb/dbinit.sql
index 254080e92d..6c82b63e6e 100644
--- a/schema/crdb/dbinit.sql
+++ b/schema/crdb/dbinit.sql
@@ -3515,7 +3515,7 @@ INSERT INTO omicron.public.db_metadata (
     version,
     target_version
 ) VALUES
-    ( TRUE, NOW(), NOW(), '34.0.0', NULL)
+    ( TRUE, NOW(), NOW(), '35.0.0', NULL)
 ON CONFLICT DO NOTHING;
 
 COMMIT;