fix(hosts): a failed DNS deletion must not orphan the record

Swallowing it and carrying on deleted the row that held the only reference to
that record, publishing the machine's management address for good. The purge now
fails instead — and is retried, since it re-reads the host and its other steps
are idempotent — so a broken DNS provider becomes a visible failed job rather
than a silent leak.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

app/Provisioning/Jobs/PurgeHost.php

@@ -4,7 +4,6 @@ namespace App\Provisioning\Jobs;
 
 use App\Models\Host;
 use App\Services\Dns\HetznerDnsClient;
-use Illuminate\Support\Facades\Log;
 use App\Models\VpnPeer;
 use Illuminate\Bus\Queueable;
 use Illuminate\Contracts\Queue\ShouldQueue;
@@ -22,7 +21,16 @@ class PurgeHost implements ShouldQueue
 {
     use Dispatchable, InteractsWithQueue, Queueable, SerializesModels;
 
-    public $tries = 1;
+    // Retried, because the DNS deletion below can fail for a minute and the job
+    // is re-entrant: it re-reads the host, already-deleted runs are simply gone,
+    // and the peer removal is idempotent.
+    public $tries = 3;
+
+    /** @return list<int> seconds between attempts */
+    public function backoff(): array
+    {
+        return [30, 120];
+    }
 
     public $timeout = 2200;
 
@@ -50,13 +58,13 @@ class PurgeHost implements ShouldQueue
         // has to happen first — otherwise the host's management address stays
         // published with nothing left to clean it up.
         if (filled($host->dns_record_id)) {
-            try {
+            // Deliberately NOT swallowed: the row about to be deleted holds the
-                app(HetznerDnsClient::class)->deleteRecord($host->dns_record_id);
+            // only reference to that record, so continuing past a failure would
-            } catch (\Throwable $e) {
+            // leave the machine's management address published for good with no
-                Log::warning('could not delete host DNS record', [
+            // way left to find it. Failing here lets the queue retry, and a
-                    'host' => $host->uuid, 'record' => $host->dns_record_id, 'error' => $e->getMessage(),
+            // permanently broken DNS provider becomes a visible failed job
-                ]);
+            // rather than a silent leak.
-            }
+            app(HetznerDnsClient::class)->deleteRecord($host->dns_record_id);
         }
 
         if (filled($host->wg_pubkey)) {

app/Services/Dns/FakeHetznerDnsClient.php

@@ -16,6 +16,8 @@ class FakeHetznerDnsClient implements HetznerDnsClient
 
     public bool $failUpsert = false;
 
+    public bool $failDelete = false;
+
     private int $counter = 1;
 
     public function upsertRecord(string $fqdn, string $type, string $value): string
@@ -31,6 +33,10 @@ class FakeHetznerDnsClient implements HetznerDnsClient
 
     public function deleteRecord(string $recordId): void
     {
+        if ($this->failDelete) {
+            throw new RuntimeException('hetzner dns 5xx');
+        }
+
         $this->records = array_filter($this->records, fn ($id) => $id !== $recordId);
     }
 }

tests/Feature/Provisioning/HostStepsTest.php

@@ -317,3 +317,30 @@ it('takes the host DNS record with it when the host is purged', function () {
     // the identifier needed to remove it deleted along with the row.
     expect($s['dns']->records)->not->toHaveKey($fqdn);
 });
+
+it('keeps the host until its DNS record is really gone', function () {
+    $s = fakeServices();
+    app()->instance(App\Services\Dns\HetznerDnsClient::class, $s['dns']);
+
+    $host = App\Models\Host::factory()->active()->create([
+        'datacenter' => 'hel', 'wg_ip' => '10.66.0.60', 'dns_name' => null,
+    ]);
+    $run = App\Models\ProvisioningRun::factory()->forHost($host)->create(['pipeline' => 'host']);
+    (new App\Provisioning\Steps\Host\RegisterHostDns($s['dns']))->execute($run);
+
+    $s['dns']->failDelete = true;
+
+    // The row carries the only reference to that record — deleting it anyway
+    // would publish the machine's management address for good.
+    expect(fn () => (new App\Provisioning\Jobs\PurgeHost($host->uuid))->handle())
+        ->toThrow(RuntimeException::class);
+
+    expect(App\Models\Host::query()->whereKey($host->id)->exists())->toBeTrue();
+
+    // Once DNS is back, the purge completes.
+    $s['dns']->failDelete = false;
+    (new App\Provisioning\Jobs\PurgeHost($host->uuid))->handle();
+
+    expect(App\Models\Host::query()->whereKey($host->id)->exists())->toBeFalse()
+        ->and($s['dns']->records)->toBe([]);
+});