From de80e3afe448237c69a535517ca88c12d079ee6d Mon Sep 17 00:00:00 2001 From: Yassin Kortam Date: Wed, 2 Sep 2026 10:19:03 -0700 Subject: [PATCH] fix(helm): scale the classic chart's HPA out at the documented 60 percent CPU (#35975) * fix(helm): scale the classic chart's HPA out at the documented 60 percent CPU The litellm-helm chart shipped targetCPUUtilizationPercentage: 80, which is unexamined helm create scaffold rather than a chosen number. It arrived packaged with the stock minReplicas: 1, maxReplicas: 100, a commented-out targetMemoryUtilizationPercentage: 80, and the boilerplate "such as Minikube" comment, the same provenance as the 128Mi resource example this file just corrected. 60 is the documented recommendation. The mechanism behind it is scale-up lag: the chart's own startupProbe is failureThreshold: 30 times periodSeconds: 10, so a replica can take up to 300 seconds to become ready, and a pod added at 80 percent utilization arrives minutes after saturation. The memory target stays commented out on purpose. The prisma query engine's resident memory is a high-water mark that ratchets to the pod's worst-ever write and is never returned, so a memory-target HPA reads the largest write a pod ever did rather than what it is doing now, and replicas ratchet up without scaling back in. hpa_tests.yaml carried its second suite after a YAML document separator, and helm-unittest loads only the first document per file, so that suite never ran; an assertion planted in it still passed. Fold it into the one live suite and add coverage pinning the rendered CPU target, the absence of a memory metric by default, and that overrides still take effect. Bump the chart to 1.1.2, since rendered output changes for anyone running with autoscaling enabled. * fix(helm): bump litellm-helm to 1.1.3 after rebase onto 1.1.2 Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --------- Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- helm/litellm-helm/Chart.yaml | 2 +- helm/litellm-helm/tests/hpa_tests.yaml | 42 ++++++++++++++++++++++---- helm/litellm-helm/values.yaml | 11 ++++++- 3 files changed, 47 insertions(+), 8 deletions(-) diff --git a/helm/litellm-helm/Chart.yaml b/helm/litellm-helm/Chart.yaml index 3959d85edf3..a3cb388ffc6 100644 --- a/helm/litellm-helm/Chart.yaml +++ b/helm/litellm-helm/Chart.yaml @@ -18,7 +18,7 @@ type: application # This is the chart version. This version number should be incremented each time you make changes # to the chart and its templates, including the app version. # Versions are expected to follow Semantic Versioning (https://semver.org/) -version: 1.1.2 +version: 1.1.3 # This is the version number of the application being deployed. This version number should be # incremented each time you make changes to the application. Versions are not expected to diff --git a/helm/litellm-helm/tests/hpa_tests.yaml b/helm/litellm-helm/tests/hpa_tests.yaml index ec18c3591d3..cd062dd5971 100644 --- a/helm/litellm-helm/tests/hpa_tests.yaml +++ b/helm/litellm-helm/tests/hpa_tests.yaml @@ -1,4 +1,4 @@ -suite: "hpa with behavior" +suite: "hpa" templates: - hpa.yaml tests: @@ -23,14 +23,44 @@ tests: - equal: { path: spec.behavior.scaleUp.stabilizationWindowSeconds, value: 60 } - equal: { path: spec.behavior.scaleDown.stabilizationWindowSeconds, value: 90 } ---- -suite: "hpa without behavior" -templates: - - hpa.yaml -tests: - it: "does not render behavior when not set" set: autoscaling.enabled: true asserts: - isKind: { of: HorizontalPodAutoscaler } - isNull: { path: spec.behavior } + + - it: "scales on cpu at the documented 60 percent by default" + set: + autoscaling.enabled: true + asserts: + - isKind: { of: HorizontalPodAutoscaler } + - equal: { path: "spec.metrics[0].resource.name", value: cpu } + - equal: { path: "spec.metrics[0].resource.target.type", value: Utilization } + - equal: { path: "spec.metrics[0].resource.target.averageUtilization", value: 60 } + + - it: "does not scale on memory by default" + set: + autoscaling.enabled: true + asserts: + - lengthEqual: { path: spec.metrics, count: 1 } + + - it: "honours an explicit cpu target override" + set: + autoscaling.enabled: true + autoscaling.targetCPUUtilizationPercentage: 75 + asserts: + - equal: { path: "spec.metrics[0].resource.target.averageUtilization", value: 75 } + + - it: "renders a memory metric only when a memory target is set" + set: + autoscaling.enabled: true + autoscaling.targetMemoryUtilizationPercentage: 80 + asserts: + - lengthEqual: { path: spec.metrics, count: 2 } + - equal: { path: "spec.metrics[1].resource.name", value: memory } + - equal: { path: "spec.metrics[1].resource.target.averageUtilization", value: 80 } + + - it: "renders no hpa when autoscaling is disabled" + asserts: + - hasDocuments: { count: 0 } diff --git a/helm/litellm-helm/values.yaml b/helm/litellm-helm/values.yaml index f8df98de102..637be2322e3 100644 --- a/helm/litellm-helm/values.yaml +++ b/helm/litellm-helm/values.yaml @@ -200,7 +200,16 @@ autoscaling: enabled: false minReplicas: 1 maxReplicas: 100 - targetCPUUtilizationPercentage: 80 + # 60 is the documented recommendation. See "Recommended Machine Specifications" + # in https://docs.litellm.ai/docs/proxy/prod. A new replica clears the startupProbe + # above only after up to failureThreshold x periodSeconds = 300 seconds, so a target + # high enough to trip near saturation adds capacity minutes after it was needed. + targetCPUUtilizationPercentage: 60 + # Deliberately left unset rather than given a value. The prisma query engine's + # resident memory is a high-water mark that ratchets to the pod's worst-ever write + # and is never returned, so a memory target reads the largest write a pod ever did + # rather than what it is doing now, and replicas ratchet up without scaling back in. + # Memory is a floor to provision under 'resources', not a signal to scale on. # targetMemoryUtilizationPercentage: 80 # behavior: {}