From 976d8b4dca00d7edb145f5cc3fc146aaada40577 Mon Sep 17 00:00:00 2001
From: Konstantin Renner <konstantin@renner.uno>
Date: Thu, 14 May 2026 20:36:22 +0200
Subject: [PATCH] immich: HPA for machine-learning

Bursty CPU profile (library scans, facial recognition, smart search
queries) that idles between. Cache PVC is RWX, so multi-replica is
safe. Cap at 2 to limit aggregate model RAM. Slow scale rates and
long scaleDown stabilization (10m) avoid churning the model cache,
which takes ~30s of cold-start to rebuild.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---
 kubernetes/apps/immich/machine-learning.yaml | 33 ++++++++++++++++++++
 1 file changed, 33 insertions(+)

diff --git a/kubernetes/apps/immich/machine-learning.yaml b/kubernetes/apps/immich/machine-learning.yaml
index bcfc9972..fcca0ae1 100644
--- a/kubernetes/apps/immich/machine-learning.yaml
+++ b/kubernetes/apps/immich/machine-learning.yaml
@@ -70,3 +70,36 @@ spec:
   ports:
     - port: 3003
       targetPort: http
+---
+apiVersion: autoscaling/v2
+kind: HorizontalPodAutoscaler
+metadata:
+  name: immich-machine-learning
+  namespace: immich
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: immich-machine-learning
+  minReplicas: 1
+  maxReplicas: 2
+  metrics:
+    - type: Resource
+      resource:
+        name: cpu
+        target:
+          type: Utilization
+          averageUtilization: 70
+  behavior:
+    scaleUp:
+      stabilizationWindowSeconds: 0
+      policies:
+        - type: Pods
+          value: 1
+          periodSeconds: 120
+    scaleDown:
+      stabilizationWindowSeconds: 600
+      policies:
+        - type: Pods
+          value: 1
+          periodSeconds: 300