refactor(observability): extract local-dev Helm values into separate files

The kube-prometheus-stack and Jaeger releases were configured via long
chains of `--set` flags, which obscure the configuration and make the
script hard to extend. Extract them into two checked-in values files
the setup script consumes via `--values`.

- tasks/scripts/observability-prometheus-values.yaml — slim chart config
  plus Grafana auto-provisioning of a Jaeger datasource (stable uid so
  dashboards can reference it).
- tasks/scripts/observability-jaeger-values.yaml — all-in-one Jaeger.
- PROMSTACK_VALUES and JAEGER_VALUES env vars allow pointing at custom
  files for local experimentation.

Author: TaylorMutch

tasks/scripts/observability-jaeger-values.yaml

@@ -0,0 +1,25 @@
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Helm values for the local-dev jaegertracing/jaeger install.
+# Consumed by tasks/scripts/observability-k8s-setup.sh.
+#
+# All-in-one mode: single pod with in-memory storage, OTLP/gRPC receiver on
+# :4317, OTLP/HTTP on :4318, UI on :16686. Enough for dashboards and trace
+# inspection during dev; nothing persists across pod restarts.
+
+allInOne:
+  enabled: true
+
+storage:
+  type: memory
+
+provisionDataStore:
+  cassandra: false
+
+agent:
+  enabled: false
+collector:
+  enabled: false
+query:
+  enabled: false

tasks/scripts/observability-k8s-setup.sh

@@ -20,12 +20,16 @@
 
 set -euo pipefail
 
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
 MONITORING_NAMESPACE="${MONITORING_NAMESPACE:-monitoring}"
 OBSERVABILITY_NAMESPACE="${OBSERVABILITY_NAMESPACE:-observability}"
 PROMSTACK_RELEASE="${PROMSTACK_RELEASE:-kube-prometheus-stack}"
 PROMSTACK_VERSION="${PROMSTACK_VERSION:-75.0.0}"
+PROMSTACK_VALUES="${PROMSTACK_VALUES:-${SCRIPT_DIR}/observability-prometheus-values.yaml}"
 JAEGER_RELEASE="${JAEGER_RELEASE:-jaeger}"
 JAEGER_VERSION="${JAEGER_VERSION:-3.4.0}"
+JAEGER_VALUES="${JAEGER_VALUES:-${SCRIPT_DIR}/observability-jaeger-values.yaml}"
 HEALTH_TIMEOUT="${HEALTH_TIMEOUT:-180}"
 
 # ---------------------------------------------------------------------------
@@ -51,11 +55,7 @@ helm upgrade --install "${PROMSTACK_RELEASE}" prometheus-community/kube-promethe
     --version "${PROMSTACK_VERSION}" \
     --namespace "${MONITORING_NAMESPACE}" \
     --create-namespace \
-    --set alertmanager.enabled=false \
-    --set nodeExporter.enabled=false \
-    --set kubeStateMetrics.enabled=false \
-    --set grafana.adminPassword=admin \
-    --set prometheus.prometheusSpec.serviceMonitorSelectorNilUsesHelmValues=false \
+    --values "${PROMSTACK_VALUES}" \
     --wait --timeout "${HEALTH_TIMEOUT}s"
 
 # ---------------------------------------------------------------------------
@@ -67,12 +67,7 @@ helm upgrade --install "${JAEGER_RELEASE}" jaegertracing/jaeger \
     --version "${JAEGER_VERSION}" \
     --namespace "${OBSERVABILITY_NAMESPACE}" \
     --create-namespace \
-    --set allInOne.enabled=true \
-    --set storage.type=memory \
-    --set provisionDataStore.cassandra=false \
-    --set agent.enabled=false \
-    --set collector.enabled=false \
-    --set query.enabled=false \
+    --values "${JAEGER_VALUES}" \
     --wait --timeout "${HEALTH_TIMEOUT}s"
 
 # ---------------------------------------------------------------------------

tasks/scripts/observability-prometheus-values.yaml

@@ -0,0 +1,35 @@
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Helm values for the local-dev kube-prometheus-stack install.
+# Consumed by tasks/scripts/observability-k8s-setup.sh.
+#
+# Slimmed down for k3d: drops Alertmanager + node/kube-state exporters.
+# Real-cluster operators install the bundle separately and follow the
+# docs/kubernetes/monitoring.mdx guide.
+
+alertmanager:
+  enabled: false
+nodeExporter:
+  enabled: false
+kubeStateMetrics:
+  enabled: false
+
+prometheus:
+  prometheusSpec:
+    # Don't restrict to ServiceMonitors carrying the Helm release label —
+    # the openshell chart sets its own selector via monitoring.serviceMonitor.labels.
+    serviceMonitorSelectorNilUsesHelmValues: false
+
+grafana:
+  adminPassword: admin
+  # Auto-provision Jaeger as a trace datasource. The Prometheus datasource is
+  # added by the chart automatically.
+  additionalDataSources:
+    - name: Jaeger
+      uid: jaeger
+      type: jaeger
+      access: proxy
+      # Service installed by tasks/scripts/observability-jaeger-values.yaml.
+      url: http://jaeger-query.observability.svc.cluster.local:16686
+      editable: true