2026-05-21 17:10:19 +00:00
1 changed files with 20 additions and 0 deletions
@@ -18,3 +18,23 @@ spec:
      target:
        type: Utilization
        averageUtilization: 60
+  behavior:
+    scaleUp:
+      # Wait 2min of sustained high CPU before scaling up. Schemas is
+      # event-driven and the per-request work is bursty even with the
+      # cache + warmup, so single spikes shouldn't pull replicas up.
+      stabilizationWindowSeconds: 120
+      policies:
+      - type: Pods
+        value: 1
+        periodSeconds: 60
+    scaleDown:
+      # Default 300s window kept pods pinned at maxReplicas long after
+      # the triggering spike had subsided. 120s is long enough to avoid
+      # flapping but lets the deployment return to minReplicas quickly
+      # once the workload calms.
+      stabilizationWindowSeconds: 120
+      policies:
+      - type: Pods
+        value: 1
+        periodSeconds: 60