From 82916167cfa5fd990a1c0aee76795988983656b0 Mon Sep 17 00:00:00 2001
From: nexxo <nexxo@clupilot.local>
Date: Sat, 25 Jul 2026 20:08:07 +0200
Subject: [PATCH] fix(monitoring): MONITORING_ATTEMPTS is now exact
 (off-by-one); accept int or dict monitor id
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- attempt is 0-based, so compare one-based: ATTEMPTS=2 really means two tries
- bridge accepts add_monitor returning {monitorID}|{id}|int (v1.2.1 returns a
  dict — re-verified: first create on a fresh Kuma is HTTP 200 with the id)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 .../Steps/Customer/RegisterMonitoring.php            |  6 ++++--
 docker/kuma-bridge/app.py                            | 12 ++++++++++--
 docs/monitoring-uptime-kuma.md                       |  2 +-
 tests/Feature/Provisioning/CustomerStepsTest.php     |  9 +++++----
 4 files changed, 20 insertions(+), 9 deletions(-)

diff --git a/app/Provisioning/Steps/Customer/RegisterMonitoring.php b/app/Provisioning/Steps/Customer/RegisterMonitoring.php
index b4b0a1a..7ee6959 100644
--- a/app/Provisioning/Steps/Customer/RegisterMonitoring.php
+++ b/app/Provisioning/Steps/Customer/RegisterMonitoring.php
@@ -41,8 +41,10 @@ class RegisterMonitoring extends CustomerStep
                 throw $e; // operator wants monitoring to gate delivery
             }
 
-            $allowed = (int) config('provisioning.monitoring.attempts', 2);
-            if ($run->attempt < $allowed) {
+            // `attempt` is 0-based (this call IS attempt number attempt+1), so
+            // compare one-based: MONITORING_ATTEMPTS=2 really means two tries.
+            $allowed = max(1, (int) config('provisioning.monitoring.attempts', 2));
+            if ($run->attempt + 1 < $allowed) {
                 return StepResult::retry(30, 'monitoring unavailable: '.$e->getMessage());
             }
 
diff --git a/docker/kuma-bridge/app.py b/docker/kuma-bridge/app.py
index 72a699b..fb498b7 100644
--- a/docker/kuma-bridge/app.py
+++ b/docker/kuma-bridge/app.py
@@ -163,11 +163,19 @@ def create_monitor(payload: MonitorIn) -> dict:
             maxretries=payload.retries or 1,   # Kuma's own field name
         )
     )
-    monitor_id = created.get("monitorID") if isinstance(created, dict) else None
+    # uptime-kuma-api 1.2.x returns {"msg": ..., "monitorID": N} (verified against
+    # Kuma 1.x). Accept a bare id too, so a future library change can't break us.
+    if isinstance(created, dict):
+        monitor_id = created.get("monitorID", created.get("id"))
+    elif isinstance(created, int):
+        monitor_id = created
+    else:
+        monitor_id = None
+
     if monitor_id is None:
         raise HTTPException(status_code=502, detail=f"Kuma returned no monitor id: {created}")
 
-    return {"monitor": {"id": monitor_id}}
+    return {"monitor": {"id": int(monitor_id)}}
 
 
 @app.get("/monitors/{monitor_id}", dependencies=[Depends(require_token)])
diff --git a/docs/monitoring-uptime-kuma.md b/docs/monitoring-uptime-kuma.md
index df67748..f6d9c08 100644
--- a/docs/monitoring-uptime-kuma.md
+++ b/docs/monitoring-uptime-kuma.md
@@ -64,7 +64,7 @@ Alle Monitor-Routen erfordern `Authorization: Bearer $BRIDGE_TOKEN`.
 Monitoring ist Beobachtung, nicht das Produkt. Fällt Kuma oder die Bridge aus,
 darf die Bereitstellung eines **bezahlten** Kunden nicht scheitern:
 
-- Der Schritt `register_monitoring` versucht es `MONITORING_ATTEMPTS` mal (Standard 2),
+- Der Schritt `register_monitoring` versucht es genau `MONITORING_ATTEMPTS` mal (Standard 2 = 1 Versuch + 1 Wiederholung),
 - danach läuft die Bereitstellung **degradiert** weiter und schreibt ein
   sichtbares `info`-Ereignis in die Admin-Konsole,
 - mit `MONITORING_REQUIRED=true` wird Monitoring stattdessen erzwungen (der Lauf
diff --git a/tests/Feature/Provisioning/CustomerStepsTest.php b/tests/Feature/Provisioning/CustomerStepsTest.php
index 563e403..d8323db 100644
--- a/tests/Feature/Provisioning/CustomerStepsTest.php
+++ b/tests/Feature/Provisioning/CustomerStepsTest.php
@@ -371,13 +371,14 @@ it('keeps provisioning when monitoring is unreachable (retry, then degrade)', fu
     $s['monitoring']->failWith = 'connection refused';
     ['run' => $run, 'instance' => $instance] = reservedRun();
 
-    // Early attempts retry — the outage may be transient.
+    // MONITORING_ATTEMPTS=2 means exactly two tries: attempt 0 retries…
+    config()->set('provisioning.monitoring.attempts', 2);
     $first = app(RegisterMonitoring::class)->execute($run->fresh());
     expect($first->type)->toBe('retry')->and($first->reason)->toContain('connection refused');
 
-    // Once the allowance is used up, continue DEGRADED with a visible event
-    // instead of failing a paid customer's provisioning.
-    $run->update(['attempt' => (int) config('provisioning.monitoring.attempts', 2)]);
+    // …and the SECOND try (attempt 1) is the last: continue DEGRADED with a
+    // visible event instead of failing a paid customer's provisioning.
+    $run->update(['attempt' => 1]);
     expect(app(RegisterMonitoring::class)->execute($run->fresh())->type)->toBe('advance');
 
     expect($instance->fresh()->monitoringTargets()->count())->toBe(0)