|
| 1 | +# Monitoring stack values tuned for Oracle Cloud Always Free tier |
| 2 | +# 2 ARM64 nodes, 4 OCPUs + 24GB RAM each, limited block volume quota |
| 3 | +# |
| 4 | +# Total monitoring resource budget (requests): |
| 5 | +# CPU: ~400m (of 8000m available across 2 nodes) |
| 6 | +# RAM: ~576Mi (of 48Gi available across 2 nodes) |
| 7 | + |
| 8 | +victoria-metrics-k8s-stack: |
| 9 | + |
| 10 | + # --- VMSingle (metrics storage) --- |
| 11 | + vmsingle: |
| 12 | + enabled: true |
| 13 | + spec: |
| 14 | + retentionPeriod: "7d" |
| 15 | + replicaCount: 1 |
| 16 | + extraArgs: {} |
| 17 | + # Use emptyDir instead of PVC to avoid consuming block volume quota |
| 18 | + # (OCI Always Free tier: 200Gi total, 150Gi already used by protocol nodes) |
| 19 | + storage: {} |
| 20 | + resources: |
| 21 | + requests: |
| 22 | + cpu: 100m |
| 23 | + memory: 256Mi |
| 24 | + limits: |
| 25 | + cpu: 500m |
| 26 | + memory: 1Gi |
| 27 | + securityContext: |
| 28 | + runAsNonRoot: true |
| 29 | + runAsUser: 1000 |
| 30 | + runAsGroup: 1000 |
| 31 | + containers: |
| 32 | + - name: vmsingle |
| 33 | + securityContext: |
| 34 | + allowPrivilegeEscalation: false |
| 35 | + capabilities: |
| 36 | + drop: |
| 37 | + - ALL |
| 38 | + readOnlyRootFilesystem: true |
| 39 | + seccompProfile: |
| 40 | + type: RuntimeDefault |
| 41 | + |
| 42 | + # --- VMAgent (scrape + remote write) --- |
| 43 | + vmagent: |
| 44 | + enabled: true |
| 45 | + spec: |
| 46 | + port: "8429" |
| 47 | + selectAllByDefault: true |
| 48 | + scrapeInterval: 30s |
| 49 | + extraArgs: |
| 50 | + promscrape.streamParse: "true" |
| 51 | + promscrape.dropOriginalLabels: "true" |
| 52 | + resources: |
| 53 | + requests: |
| 54 | + cpu: 50m |
| 55 | + memory: 128Mi |
| 56 | + limits: |
| 57 | + cpu: 200m |
| 58 | + memory: 256Mi |
| 59 | + securityContext: |
| 60 | + runAsNonRoot: true |
| 61 | + runAsUser: 1000 |
| 62 | + runAsGroup: 1000 |
| 63 | + containers: |
| 64 | + - name: vmagent |
| 65 | + securityContext: |
| 66 | + allowPrivilegeEscalation: false |
| 67 | + capabilities: |
| 68 | + drop: |
| 69 | + - ALL |
| 70 | + readOnlyRootFilesystem: true |
| 71 | + seccompProfile: |
| 72 | + type: RuntimeDefault |
| 73 | + |
| 74 | + # --- VMAlert (recording rules evaluation) --- |
| 75 | + # Kept enabled for recording rules but with minimal resources |
| 76 | + vmalert: |
| 77 | + enabled: true |
| 78 | + spec: |
| 79 | + port: "8080" |
| 80 | + selectAllByDefault: true |
| 81 | + evaluationInterval: 30s |
| 82 | + extraArgs: |
| 83 | + http.pathPrefix: "/" |
| 84 | + # Alertmanager is disabled; sink alert notifications to blackhole |
| 85 | + notifier.blackhole: "true" |
| 86 | + resources: |
| 87 | + requests: |
| 88 | + cpu: 50m |
| 89 | + memory: 64Mi |
| 90 | + limits: |
| 91 | + cpu: 200m |
| 92 | + memory: 256Mi |
| 93 | + |
| 94 | + # --- Alertmanager (DISABLED - not needed for initial deployment) --- |
| 95 | + alertmanager: |
| 96 | + enabled: false |
| 97 | + |
| 98 | + # --- Grafana --- |
| 99 | + grafana: |
| 100 | + enabled: true |
| 101 | + # Use emptyDir for SQLite database (no PVC needed) |
| 102 | + persistence: |
| 103 | + enabled: false |
| 104 | + resources: |
| 105 | + requests: |
| 106 | + cpu: 100m |
| 107 | + memory: 256Mi |
| 108 | + limits: |
| 109 | + cpu: 500m |
| 110 | + memory: 512Mi |
| 111 | + securityContext: |
| 112 | + runAsNonRoot: true |
| 113 | + runAsUser: 472 |
| 114 | + runAsGroup: 472 |
| 115 | + fsGroup: 472 |
| 116 | + containerSecurityContext: |
| 117 | + allowPrivilegeEscalation: false |
| 118 | + capabilities: |
| 119 | + drop: |
| 120 | + - ALL |
| 121 | + readOnlyRootFilesystem: false |
| 122 | + seccompProfile: |
| 123 | + type: RuntimeDefault |
| 124 | + sidecar: |
| 125 | + dashboards: |
| 126 | + enabled: true |
| 127 | + label: grafana_dashboard |
| 128 | + labelValue: "1" |
| 129 | + folder: /var/lib/grafana/dashboards |
| 130 | + defaultFolderName: default |
| 131 | + provider: |
| 132 | + name: default |
| 133 | + orgid: 1 |
| 134 | + datasources: |
| 135 | + enabled: true |
| 136 | + label: grafana_datasource |
| 137 | + labelValue: "1" |
| 138 | + |
| 139 | + # --- Node Exporter (DaemonSet - runs on each node) --- |
| 140 | + prometheus-node-exporter: |
| 141 | + enabled: true |
| 142 | + resources: |
| 143 | + requests: |
| 144 | + cpu: 50m |
| 145 | + memory: 64Mi |
| 146 | + limits: |
| 147 | + cpu: 100m |
| 148 | + memory: 128Mi |
| 149 | + service: |
| 150 | + labels: |
| 151 | + jobLabel: node-exporter |
| 152 | + extraArgs: |
| 153 | + - --collector.filesystem.ignored-mount-points=^/(dev|proc|sys|var/lib/docker/.+|var/lib/kubelet/.+)($|/) |
| 154 | + - --collector.filesystem.ignored-fs-types=^(autofs|binfmt_misc|bpf|cgroup2?|configfs|debugfs|devpts|devtmpfs|fusectl|hugetlbfs|iso9660|mqueue|nsfs|overlay|proc|procfs|pstore|rpc_pipefs|securityfs|selinuxfs|squashfs|erofs|sysfs|tracefs)$ |
| 155 | + |
| 156 | + # --- Kube State Metrics --- |
| 157 | + kube-state-metrics: |
| 158 | + enabled: true |
| 159 | + resources: |
| 160 | + requests: |
| 161 | + cpu: 100m |
| 162 | + memory: 128Mi |
| 163 | + limits: |
| 164 | + cpu: 200m |
| 165 | + memory: 256Mi |
| 166 | + |
| 167 | + # --- VMAuth (not needed for single-node setup) --- |
| 168 | + vmauth: |
| 169 | + enabled: false |
| 170 | + |
| 171 | + # --- VMCluster (not needed - using VMSingle) --- |
| 172 | + vmcluster: |
| 173 | + enabled: false |
| 174 | + |
| 175 | + # --- Operator --- |
| 176 | + victoria-metrics-operator: |
| 177 | + enabled: true |
| 178 | + crds: |
| 179 | + plain: true |
| 180 | + cleanup: |
| 181 | + enabled: true |
| 182 | + operator: |
| 183 | + disable_prometheus_converter: false |
| 184 | + |
| 185 | + # --- Default dashboards --- |
| 186 | + defaultDashboards: |
| 187 | + enabled: true |
| 188 | + defaultTimezone: utc |
| 189 | + |
| 190 | + # --- Default rules (recording + alerting) --- |
| 191 | + defaultRules: |
| 192 | + create: true |
| 193 | + |
| 194 | + # TODO: Add custom Grafana dashboard ConfigMaps for Disentangle-specific |
| 195 | + # metrics (P2P peers, DAG height, mining rate, RPC latency) once the |
| 196 | + # protocol node exposes a Prometheus /metrics endpoint. |
0 commit comments