perf(observability): Reduce CPU/memory requests to fix CI resource exhaustion

Root cause: GitHub Actions runners have limited CPU. High resource requests
prevented pods from scheduling (Kiali: "0/1 nodes available: 1 Insufficient cpu").

Solution: Reduce CPU requests by 65% to allow Kubernetes to overcommit resources.
We accept slower CI execution in exchange for successful completion.

Resource request changes:
- Prometheus: 50m CPU, 512Mi → 50m CPU, 512Mi (no change, already optimal)
- Grafana: 50m CPU, 256Mi → 50m CPU, 256Mi (no change, already optimal)
- Loki: 200m CPU, 512Mi → 50m CPU, 256Mi
- Tempo: 200m CPU, 512Mi → 50m CPU, 256Mi
- OTEL Collector: 200m CPU, 512Mi → 50m CPU, 256Mi (2 replicas = 100m total)
- Phoenix: 100m CPU, 256Mi → 50m CPU, 128Mi
- Kiali: 100m CPU, 256Mi → 50m CPU, 128Mi
- Alertmanager: 100m CPU, 128Mi → 25m CPU, 64Mi
- Korrel8r: 100m CPU, 128Mi → 25m CPU, 64Mi
- Kube-state-metrics: 10m CPU, 32Mi (no change, already optimal)

Total CPU requests: 1310m → 460m (65% reduction)

Additional optimization:
- Prometheus scrape_interval: 15s → 30s (50% reduction in scrape frequency)
- Prometheus evaluation_interval: 15s → 30s

Expected outcome:
- Pods can schedule in resource-constrained CI environment
- CI run will be slower but will complete successfully
- Local Kind deployments unaffected (have more resources)

Ref: Run 19666137964 failure analysis in /tmp/ci_failure_analysis_19666137964.md

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: Ladas

argocd/applications/helm/kiali.yaml

@@ -28,8 +28,8 @@ spec:
           replicas: 1
           resources:
             requests:
-              cpu: 100m
-              memory: 256Mi
+              cpu: 50m      # Reduced for CI (allows overcommit)
+              memory: 128Mi # Reduced for CI
             limits:
               cpu: 500m
               memory: 1Gi

components/02-observability/alertmanager/deployment.yaml

@@ -81,8 +81,8 @@ spec:
           # Resource limits
           resources:
             requests:
-              memory: "128Mi"
-              cpu: "100m"
+              memory: "64Mi"  # Reduced for CI (allows overcommit)
+              cpu: "25m"      # Reduced for CI
             limits:
               memory: "256Mi"
               cpu: "200m"

components/02-observability/korrel8r/deployment.yaml

@@ -43,8 +43,8 @@ spec:
           readOnly: true
         resources:
           requests:
-            cpu: 100m
-            memory: 128Mi
+            cpu: 25m  # Reduced for CI (allows overcommit)
+            memory: 64Mi # Reduced for CI
           limits:
             cpu: 500m
             memory: 512Mi

components/02-observability/loki/deployment.yaml

@@ -79,8 +79,8 @@ spec:
 
         resources:
           requests:
-            cpu: 200m
-            memory: 512Mi
+            cpu: 50m      # Reduced for CI (allows overcommit)
+            memory: 256Mi # Reduced for CI
           limits:
             cpu: 1000m
             memory: 2Gi

components/02-observability/otel-collector/deployment.yaml

@@ -78,8 +78,8 @@ spec:
 
         resources:
           requests:
-            cpu: 200m
-            memory: 512Mi
+            cpu: 50m      # Reduced for CI (allows overcommit, 2 replicas = 100m total)
+            memory: 256Mi # Reduced for CI
           limits:
             cpu: 1000m
             memory: 2Gi

components/02-observability/phoenix/deployment.yaml

@@ -38,8 +38,8 @@ spec:
               cpu: 500m
               memory: 1Gi
             requests:
-              cpu: 100m
-              memory: 256Mi
+              cpu: 50m      # Reduced for CI (allows overcommit)
+              memory: 128Mi # Reduced for CI
           livenessProbe:
             httpGet:
               path: /

components/02-observability/prometheus/configmap.yaml

@@ -8,8 +8,8 @@ metadata:
 data:
   prometheus.yml: |
     global:
-      scrape_interval: 15s
-      evaluation_interval: 15s
+      scrape_interval: 30s      # 30s instead of 15s to reduce CPU load in CI
+      evaluation_interval: 30s   # 30s instead of 15s
       external_labels:
         cluster: 'kagenti-demo'
         environment: 'kind-local'

components/02-observability/tempo/deployment.yaml

@@ -68,8 +68,8 @@ spec:
 
         resources:
           requests:
-            cpu: 200m
-            memory: 512Mi
+            cpu: 50m      # Reduced for CI (allows overcommit)
+            memory: 256Mi # Reduced for CI
           limits:
             cpu: 1000m
             memory: 2Gi