From 3fad3190537e0049c256068f3806a295c63b9104 Mon Sep 17 00:00:00 2001
From: nexxo <nexxo@clupilot.local>
Date: Sun, 26 Jul 2026 00:02:18 +0200
Subject: [PATCH] fix(hosts): re-read the host before deciding there is no DNS
 record

A registration finishing while the purge waited on the run locks wrote
dns_record_id after the purge had already loaded the host, so it skipped the
deletion and then deleted the row holding the only id.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 app/Provisioning/Jobs/PurgeHost.php          |  6 ++++++
 tests/Feature/Provisioning/HostStepsTest.php | 21 ++++++++++++++++++++
 2 files changed, 27 insertions(+)

diff --git a/app/Provisioning/Jobs/PurgeHost.php b/app/Provisioning/Jobs/PurgeHost.php
index 6a13a38..3d395fb 100644
--- a/app/Provisioning/Jobs/PurgeHost.php
+++ b/app/Provisioning/Jobs/PurgeHost.php
@@ -54,6 +54,12 @@ class PurgeHost implements ShouldQueue
             Cache::lock('run:'.$run->uuid, 2200)->block(2100, fn () => $run->delete());
         }
 
+        // Re-read after waiting on the run locks above: a registration that was
+        // still running may have written dns_record_id in the meantime, and the
+        // stale copy loaded before the wait would skip the deletion and then
+        // delete the row that held the id.
+        $host->refresh();
+
         // The record identifier lives on the row about to be deleted, so this
         // has to happen first — otherwise the host's management address stays
         // published with nothing left to clean it up.
diff --git a/tests/Feature/Provisioning/HostStepsTest.php b/tests/Feature/Provisioning/HostStepsTest.php
index 7ff769c..d5d1a6a 100644
--- a/tests/Feature/Provisioning/HostStepsTest.php
+++ b/tests/Feature/Provisioning/HostStepsTest.php
@@ -388,3 +388,24 @@ it('does not hand the same name to two datacenter codes that look alike', functi
     // failed the unique constraint mid-onboarding.
     expect($names)->toBe(['eu-west-01', 'eu-west-02']);
 });
+
+it('sees a DNS record registered while the purge was waiting', function () {
+    $s = fakeServices();
+    app()->instance(App\Services\Dns\HetznerDnsClient::class, $s['dns']);
+
+    $host = App\Models\Host::factory()->active()->create([
+        'datacenter' => 'fsn', 'wg_ip' => '10.66.0.90', 'dns_name' => null,
+    ]);
+    $purge = new App\Provisioning\Jobs\PurgeHost($host->uuid);
+
+    // Registration finishes while the purge is blocked on the run lock: the
+    // copy it loaded first still says there is no record to delete.
+    $run = App\Models\ProvisioningRun::factory()->forHost($host)->create(['pipeline' => 'host']);
+    (new App\Provisioning\Steps\Host\RegisterHostDns($s['dns']))->execute($run);
+    $fqdn = $host->fresh()->dns_name.'.node.clupilot.com';
+
+    $purge->handle();
+
+    expect($s['dns']->records)->not->toHaveKey($fqdn)
+        ->and(App\Models\Host::query()->whereKey($host->id)->exists())->toBeFalse();
+});