diff --git a/infra/k8s/gitlab/runner-rl-values.yaml b/infra/k8s/gitlab/runner-rl-values.yaml index 5fafb3c8e..8a81ed7da 100644 --- a/infra/k8s/gitlab/runner-rl-values.yaml +++ b/infra/k8s/gitlab/runner-rl-values.yaml @@ -34,10 +34,11 @@ runners: image = "rust:1.89-slim" privileged = false node_selector_overwrite_allowed = ".*" - cpu_request_overwrite_max_allowed = "4000m" - cpu_limit_overwrite_max_allowed = "4000m" - memory_request_overwrite_max_allowed = "16Gi" - memory_limit_overwrite_max_allowed = "16Gi" + # Hyperopt RL jobs need 6000m CPU + 20Gi mem (PSO parallel trials on L4 8 vCPU) + cpu_request_overwrite_max_allowed = "8000m" + cpu_limit_overwrite_max_allowed = "8000m" + memory_request_overwrite_max_allowed = "24Gi" + memory_limit_overwrite_max_allowed = "24Gi" poll_timeout = 600 # Allow CI jobs to override runtimeClassName via KUBERNETES_RUNTIME_CLASS_NAME runtime_class_name_overwrite_allowed = ".*" diff --git a/infra/k8s/monitoring/tempo.yaml b/infra/k8s/monitoring/tempo.yaml index d4f2dc764..8f4948f61 100644 --- a/infra/k8s/monitoring/tempo.yaml +++ b/infra/k8s/monitoring/tempo.yaml @@ -77,10 +77,10 @@ spec: resources: requests: cpu: 100m - memory: 128Mi + memory: 256Mi limits: cpu: 500m - memory: 512Mi + memory: 1Gi readinessProbe: httpGet: path: /ready