fix: Kubernetes 시스템 로그 수집 안정화

kubernetes/monitoring/grafana-dashboard-monitoring-stack.yaml

@@ -10,7 +10,7 @@ data:
       "uid": "codeplace-monitoring-stack",
       "title": "CodePlace Monitoring Stack",
       "schemaVersion": 39,
-      "version": 2,
+      "version": 3,
       "refresh": "30s",
       "tags": ["codeplace", "monitoring", "scrape-contract"],
       "links": [
@@ -571,6 +571,22 @@ data:
               "legendFormat": "{{namespace}} {{job}}"
             }
           ]
+        },
+        {
+          "type": "logs",
+          "title": "Recent kube-vip Logs",
+          "description": "Cluster control-plane kube-vip logs for VIP ownership, leader election, and ARP troubleshooting.",
+          "gridPos": {"x": 0, "y": 49, "w": 24, "h": 8},
+          "datasource": {"type": "loki", "uid": "Loki"},
+          "options": {
+            "showLabels": false,
+            "showTime": true,
+            "wrapLogMessage": true,
+            "dedupStrategy": "none"
+          },
+          "targets": [
+            {"refId": "A", "expr": "{namespace=\"kube-system\", app_kubernetes_io_name=\"kube-vip-ds\"}"}
+          ]
         }
       ]
     }

kubernetes/monitoring/kubernetes-event-exporter.yaml

@@ -44,6 +44,7 @@ data:
   config.yaml: |
     logLevel: warn
     logFormat: json
+    omitLookup: true
     maxEventAgeSeconds: 60
     metricsNamePrefix: event_exporter_
     route:

kubernetes/monitoring/logs/README.md

@@ -4,7 +4,7 @@ This directory contains Helm values for the Kubernetes logs stack.
 
 - Loki: SingleBinary mode, filesystem storage, Longhorn PVC.
 - Loki gateway: 2 replicas for the stateless write/query entrypoint.
-- Alloy: DaemonSet log collector for every pod in `code-place-dev`, `code-place-prod`, and `monitoring`, plus Traefik pods in `kube-system`. It mounts host `/var/log` so `/var/log/pods` is readable from the Alloy container. It tolerates the standard `control-plane` and `master` NoSchedule taints so tainted control-plane nodes are still covered.
+- Alloy: DaemonSet log collector for every pod in `code-place-dev`, `code-place-prod`, and `monitoring`, plus Traefik and kube-vip pods in `kube-system`. It mounts host `/var/log` so `/var/log/pods` is readable from the Alloy container. It tolerates the standard `control-plane` and `master` NoSchedule taints so tainted control-plane nodes are still covered.
 - Kubernetes Event Exporter: Warning events are written to stdout as JSON and collected by Alloy from the `monitoring` namespace.
 - Retention: `code-place-dev` 3 days, `code-place-prod` 7 days.
 - Monitoring: Loki and Alloy ServiceMonitors are enabled for kube-prometheus-stack.
@@ -47,6 +47,10 @@ helm upgrade --install alloy grafana/alloy \
 
 kubectl apply -k kubernetes/monitoring
 
+# Event exporter reads its ConfigMap only at startup.
+kubectl -n monitoring rollout restart deployment/kubernetes-event-exporter
+kubectl -n monitoring rollout status deployment/kubernetes-event-exporter --timeout=3m
+
 # One-time migration after the replacement Traefik PodMonitor is present.
 kubectl -n monitoring get podmonitor traefik
 kubectl -n monitoring delete servicemonitor traefik --ignore-not-found
@@ -79,6 +83,7 @@ Grafana should show a `Loki` datasource. Useful Explore queries:
 {namespace="code-place-prod", app="backend"} | json
 {namespace="code-place-dev", container=~"backend|celery-worker|judge-server"}
 {namespace="kube-system", app_kubernetes_io_name="traefik"} | json | DownstreamStatus >= 400
+{namespace="kube-system", app_kubernetes_io_name="kube-vip-ds"}
 {namespace="monitoring", app_kubernetes_io_name="kubernetes-event-exporter"}
 ```
 

kubernetes/monitoring/logs/alloy-values.yaml

@@ -1,5 +1,5 @@
 # Helm values for grafana/alloy.
-# Alloy runs as a DaemonSet and collects CodePlace, monitoring, and Traefik logs.
+# Alloy runs as a DaemonSet and collects CodePlace, monitoring, Traefik, and kube-vip logs.
 
 controller:
   type: daemonset
@@ -37,7 +37,7 @@ alloy:
         rule {
           source_labels = ["__meta_kubernetes_namespace", "__meta_kubernetes_pod_label_app", "__meta_kubernetes_pod_label_app_kubernetes_io_name"]
           separator     = "/"
-          regex         = "(code-place-dev|code-place-prod|monitoring)/.*|kube-system/.*/traefik"
+          regex         = "(code-place-dev|code-place-prod|monitoring)/.*|kube-system/.*/(traefik|kube-vip-ds)"
           action        = "keep"
         }