fix: make deduplicate_discovery_checks migration production-safe (#178)

- Add @disable_ddl_transaction and @disable_migration_lock for concurrent index
- Increase statement_timeout to 10min for DELETE operation on large tables
- Use concurrently: true on index creation to avoid table locks
- Refactor detect_airfiber_counter_overrides to fix credo nesting warning
- Prevents timeout errors in production deployment

Reviewed-on: graham/towerops-web#178

.pre-commit-config.yaml

@@ -1 +1 @@
-/nix/store/4xlkmdd947h5qlhj2rmbyavq08grkz27-pre-commit-config.json
+/nix/store/zx636v118rzzqzwafgrw2aybh7l0szrl-pre-commit-config.json

lib/towerops/workers/device_poller_worker.ex

@@ -1059,17 +1059,13 @@ defmodule Towerops.Workers.DevicePollerWorker do
       # needs proprietary counters. Use v1 since these devices may not support v2c.
       v1_opts = Keyword.put(client_opts, :version, "1")
 
-      # Check for LTU first (more specific)
-      case Client.get(v1_opts, "1.3.6.1.4.1.41112.1.10.1.2.2.0") do
-        {:ok, val} when val != nil ->
-          :airfiber_ltu
-
-        _ ->
-          # Check for regular AirFiber stats table (index field)
-          case Client.get(v1_opts, "1.3.6.1.4.1.41112.1.3.3.1.1.1") do
-            {:ok, val} when val != nil -> :airfiber
-            _ -> nil
-          end
+      # Check for LTU first (more specific), fall back to regular AirFiber
+      with {:ltu, {:error, _}} <- {:ltu, Client.get(v1_opts, "1.3.6.1.4.1.41112.1.10.1.2.2.0")},
+           {:ok, val} when val != nil <- Client.get(v1_opts, "1.3.6.1.4.1.41112.1.3.3.1.1.1") do
+        :airfiber
+      else
+        {:ltu, {:ok, val}} when val != nil -> :airfiber_ltu
+        _ -> nil
       end
     end
   end

priv/repo/migrations/20260325215632_deduplicate_discovery_checks.exs

@@ -1,7 +1,14 @@
 defmodule Towerops.Repo.Migrations.DeduplicateDiscoveryChecks do
   use Ecto.Migration
 
+  # Required for CREATE INDEX CONCURRENTLY
+  @disable_ddl_transaction true
+  @disable_migration_lock true
+
   def up do
+    # Increase timeout for DELETE operation on large table
+    execute("SET statement_timeout = '10min'")
+
     # Delete duplicate auto-discovered checks, keeping the oldest one per
     # (device_id, check_type, source_id) combination.
     execute("""
@@ -21,11 +28,16 @@ defmodule Towerops.Repo.Migrations.DeduplicateDiscoveryChecks do
     )
     """)
 
+    # Reset timeout
+    execute("RESET statement_timeout")
+
     # Prevent future duplicates for auto-discovered checks.
+    # Use concurrently: true to avoid locking table during index creation.
     create(
       unique_index(:checks, [:device_id, :check_type, :source_id],
         where: "source_type = 'auto_discovery' AND source_id IS NOT NULL",
-        name: :checks_device_type_source_unique_index
+        name: :checks_device_type_source_unique_index,
+        concurrently: true
       )
     )
   end
@@ -33,7 +45,8 @@ defmodule Towerops.Repo.Migrations.DeduplicateDiscoveryChecks do
   def down do
     drop_if_exists(
       index(:checks, [:device_id, :check_type, :source_id],
-        name: :checks_device_type_source_unique_index
+        name: :checks_device_type_source_unique_index,
+        concurrently: true
       )
     )
   end