fix(k8s): add scaleUp/scaleDown stabilization to schemas HPA (#844)

2026-05-21 17:10:16 +00:00
parent 4e50a051d0
commit 111c2e4b19
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