diff --git a/docker/compose/validator-prod/alertmanager/alertmanager.yml b/docker/compose/validator-prod/alertmanager/alertmanager.yml
new file mode 100644
index 0000000000000000000000000000000000000000..f72951393486b0e986ac0b0032b6c669ee499745
--- /dev/null
+++ b/docker/compose/validator-prod/alertmanager/alertmanager.yml
@@ -0,0 +1,16 @@
+global:
+ resolve_timeout: 1m
+
+route:
+ receiver: 'mail-notifications'
+
+receivers:
+- name: 'mail-notifications'
+ email_configs:
+ - to: 'mydedicatednodealertaddress@domain.tld'
+ from: 'sender@domain.tld'
+ smarthost: 'smtp.server.tld:587'
+ auth_username: 'sender@domain.tld'
+ auth_identity: 'sender@domain.tld'
+ auth_password: 'yrymyemufalyjing'
+ send_resolved: true
diff --git a/docker/compose/validator-prod/docker-compose.yml b/docker/compose/validator-prod/docker-compose.yml
new file mode 100644
index 0000000000000000000000000000000000000000..a09c4fb363ae66e32ceeefde1eb020566f779c10
--- /dev/null
+++ b/docker/compose/validator-prod/docker-compose.yml
@@ -0,0 +1,93 @@
+version: "3.4"
+
+services:
+ duniter-validator:
+ image: duniter/duniter-v2s:sha-142d4763
+ restart: unless-stopped
+ ports:
+ - 127.0.0.1:9615:9615
+ - 127.0.0.1:9933:9933
+ - 127.0.0.1:9944:9944
+ - 30333:30333
+ volumes:
+ - ./node.key:/etc/duniter/node.key
+ - duniter-validator-data:/var/lib/duniter/
+ environment:
+ - DUNITER_CHAIN_NAME=gdev
+ command:
+ - "--execution"
+ - "Wasm"
+ - "--node-key-file"
+ - "/etc/duniter/node-key"
+ - "--public-addr"
+ # SERVER_DOMAIN should be replaced by a domain name that point on your server
+ #
+ # The PEER_ID should be replaced by the output of this command:
+ # docker run --rm -it --entrypoint duniter -v $PWD:/var/lib/duniter/ duniter/duniter-v2s:sha-142d4763 key generate-node-key --file /var/lib/duniter/node.key
+ - "/dns/SERVER_DOMAIN/tcp/30333/p2p/PEER_ID"
+ - "--rpc-methods=Unsafe"
+ - "--validator"
+ - "--rpc-cors"
+ - "all"
+ - "--prometheus-external"
+
+ prometheus:
+ image: prom/prometheus:v2.30.3
+ restart: unless-stopped
+ ports:
+ - 9000:9090
+ volumes:
+ - ./prometheus:/etc/prometheus
+ - prometheus-data:/prometheus
+ command: --web.enable-lifecycle --config.file=/etc/prometheus/prometheus.yml
+ extra_hosts:
+ - "host.docker.internal:host-gateway"
+
+ alertmanager:
+ image: prom/alertmanager:v0.24.0
+ restart: unless-stopped
+ ports:
+ - 127.0.0.1:9093:9093
+ volumes:
+ - ./alertmanager:/etc/alertmanager
+ - alertmanager-data:/alertmanager
+ command: --config.file=/etc/alertmanager/alertmanager.yml
+
+ process-exporter:
+ image: ncabatoff/process-exporter
+ restart: unless-stopped
+ privileged: True
+ ports:
+ - 127.0.0.1:9256:9256
+ volumes:
+ - /proc:/host/proc
+ - ./process-exporter:/config
+ command: --procfs /host/proc -config.path /config/config.yml
+
+ node_exporter:
+ image: quay.io/prometheus/node-exporter:latest
+ container_name: node_exporter
+ command:
+ - '--path.rootfs=/host'
+ network_mode: host
+ pid: host
+ restart: unless-stopped
+ volumes:
+ - '/:/host:ro,rslave'
+
+ grafana:
+ image: grafana/grafana:8.5.4
+ restart: unless-stopped
+ ports:
+ - 127.0.0.1:3000:3000
+ volumes:
+ - ./grafana/provisioning/datasources:/etc/grafana/provisioning/datasources
+ - grafana-data:/var/lib/grafana
+ environment:
+ - GF_INSTALL_PLUGINS=camptocamp-prometheus-alertmanager-datasource
+
+volumes:
+ alertmanager-data:
+ duniter-validator-data:
+ grafana-data:
+ prometheus-data:
diff --git a/docker/compose/validator-prod/grafana/provisioning/datasources/prometheus_ds.yml b/docker/compose/validator-prod/grafana/provisioning/datasources/prometheus_ds.yml
new file mode 100644
index 0000000000000000000000000000000000000000..2842d119c8e21df4998295e12eedf373d4c64fb9
--- /dev/null
+++ b/docker/compose/validator-prod/grafana/provisioning/datasources/prometheus_ds.yml
@@ -0,0 +1,10 @@
+datasources:
+ - name: Prometheus
+ access: proxy
+ type: prometheus
+ url: http://prometheus:9090
+ isDefault: true
+ - name: Prometheus AlertManager
+ access: proxy
+ type: camptocamp-prometheus-alertmanager-datasource
+ url: http://alertmanager:9093
diff --git a/docker/compose/validator-prod/grafana/templates/polkadot-essentials-2021.json b/docker/compose/validator-prod/grafana/templates/polkadot-essentials-2021.json
new file mode 100644
index 0000000000000000000000000000000000000000..921f468eb7df9f26bad64b99e96c4547c3b0c8a0
--- /dev/null
+++ b/docker/compose/validator-prod/grafana/templates/polkadot-essentials-2021.json
@@ -0,0 +1,3671 @@
+{
+ "annotations": {
+ "list": [
+ {
+ "$$hashKey": "object:15",
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "limit": 100,
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "Server and Substrate chain main metrics, suitable for a node monitoring in 2021 - \r\nTutorial: https://medium.com/bld-nodes/monitoring-substrate-node-polkadot-kusama-parachains-validator-guide-922734ea4cdb",
+ "editable": true,
+ "gnetId": 13840,
+ "graphTooltip": 0,
+ "id": 1,
+ "iteration": 1654016981360,
+ "links": [
+ {
+ "icon": "external link",
+ "tags": [],
+ "targetBlank": true,
+ "title": "Polkastats.io",
+ "tooltip": "",
+ "type": "link",
+ "url": "https://polkastats.io"
+ },
+ {
+ "icon": "external link",
+ "tags": [],
+ "title": "Telemetry",
+ "type": "link",
+ "url": "https://telemetry.polkadot.io"
+ },
+ {
+ "icon": "external link",
+ "tags": [],
+ "title": "Polkadot JS app",
+ "type": "link",
+ "url": "https://polkadot.js.org/apps/#/explorer"
+ }
+ ],
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": true,
+ "colorValue": false,
+ "colors": ["#299c46", "rgba(237, 129, 40, 0.89)", "#d44a3a"],
+ "datasource": "Prometheus AlertManager",
+ "decimals": null,
+ "description": "",
+ "fieldConfig": {
+ "defaults": {},
+ "overrides": []
+ },
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 0,
+ "y": 0
+ },
+ "id": 84,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "annotations": false,
+ "expr": "",
+ "labelSelector": "*",
+ "legendFormat": "",
+ "refId": "A",
+ "target": "Query",
+ "type": "single"
+ }
+ ],
+ "thresholds": "1,5",
+ "title": "All alerts",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "columns": [],
+ "datasource": "Prometheus AlertManager",
+ "fieldConfig": {
+ "defaults": {},
+ "overrides": []
+ },
+ "fontSize": "100%",
+ "gridPos": {
+ "h": 3,
+ "w": 21,
+ "x": 3,
+ "y": 0
+ },
+ "id": 86,
+ "links": [],
+ "pageSize": null,
+ "scroll": true,
+ "showHeader": true,
+ "sort": {
+ "col": 0,
+ "desc": true
+ },
+ "styles": [
+ {
+ "alias": "Time",
+ "align": "auto",
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "pattern": "Time",
+ "type": "date"
+ },
+ {
+ "alias": "",
+ "align": "auto",
+ "colorMode": "row",
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "mappingType": 1,
+ "pattern": "severity",
+ "thresholds": ["1", "2"],
+ "type": "string",
+ "unit": "short",
+ "valueMaps": [
+ {
+ "text": "Critical",
+ "value": "3"
+ },
+ {
+ "text": "Warning",
+ "value": "2"
+ },
+ {
+ "text": "Info",
+ "value": "1"
+ }
+ ]
+ },
+ {
+ "alias": "",
+ "align": "auto",
+ "colorMode": "row",
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "decimals": 2,
+ "pattern": "/.*/",
+ "thresholds": ["0"],
+ "type": "number",
+ "unit": "short"
+ }
+ ],
+ "targets": [
+ {
+ "annotations": true,
+ "expr": "",
+ "hide": false,
+ "labelSelector": "*",
+ "legendFormat": "{{ summary }}",
+ "refId": "A",
+ "target": "Query",
+ "type": "table"
+ },
+ {
+ "annotations": false,
+ "expr": "",
+ "labelSelector": "*",
+ "legendFormat": "{{ instance }}",
+ "refId": "B",
+ "target": "Query",
+ "type": "table"
+ }
+ ],
+ "title": "All Alerts Info",
+ "transform": "table",
+ "type": "table-old"
+ },
+ {
+ "collapsed": false,
+ "datasource": "Prometheus",
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 3
+ },
+ "id": 50,
+ "panels": [],
+ "title": "Server Health",
+ "type": "row"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "description": "Busy state of all CPU cores together",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "thresholds"
+ },
+ "mappings": [
+ {
+ "id": 0,
+ "op": "=",
+ "text": "N/A",
+ "type": 1,
+ "value": "null"
+ }
+ ],
+ "max": 100,
+ "min": 0,
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 85
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 95
+ }
+ ]
+ },
+ "unit": "percent"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 3,
+ "x": 0,
+ "y": 4
+ },
+ "id": 54,
+ "links": [],
+ "options": {
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true,
+ "text": {}
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "(((count(count(node_cpu_seconds_total{instance=\"$node\",job=\"$job\"}) by (cpu))) - avg(sum by (mode)(irate(node_cpu_seconds_total{mode='idle',instance=\"$node\",job=\"$job\"}[5m])))) * 100) / count(count(node_cpu_seconds_total{instance=\"$node\",job=\"$job\"}) by (cpu))",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "title": "CPU Busy",
+ "type": "gauge"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "description": "Busy state of all CPU cores together (5 min average)",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "thresholds"
+ },
+ "mappings": [
+ {
+ "id": 0,
+ "op": "=",
+ "text": "N/A",
+ "type": 1,
+ "value": "null"
+ }
+ ],
+ "max": 100,
+ "min": 0,
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 85
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 95
+ }
+ ]
+ },
+ "unit": "percent"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 3,
+ "x": 3,
+ "y": 4
+ },
+ "id": 56,
+ "links": [],
+ "options": {
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true,
+ "text": {}
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "avg(node_load5{instance=\"$node\",job=\"$job\"}) / count(count(node_cpu_seconds_total{instance=\"$node\",job=\"$job\"}) by (cpu)) * 100",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "title": "Sys Load (5m avg)",
+ "type": "gauge"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "description": "Busy state of all CPU cores together (15 min average)",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "thresholds"
+ },
+ "mappings": [
+ {
+ "id": 0,
+ "op": "=",
+ "text": "N/A",
+ "type": 1,
+ "value": "null"
+ }
+ ],
+ "max": 100,
+ "min": 0,
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 85
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 95
+ }
+ ]
+ },
+ "unit": "percent"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 3,
+ "x": 6,
+ "y": 4
+ },
+ "id": 58,
+ "links": [],
+ "options": {
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true,
+ "text": {}
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "avg(node_load15{instance=\"$node\",job=\"$job\"}) / count(count(node_cpu_seconds_total{instance=\"$node\",job=\"$job\"}) by (cpu)) * 100",
+ "hide": false,
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "title": "Sys Load (15m avg)",
+ "type": "gauge"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "description": "Non available RAM memory",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "thresholds"
+ },
+ "decimals": 0,
+ "mappings": [],
+ "max": 100,
+ "min": 0,
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 80
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 90
+ }
+ ]
+ },
+ "unit": "percent"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 3,
+ "x": 9,
+ "y": 4
+ },
+ "hideTimeOverride": false,
+ "id": 60,
+ "links": [],
+ "options": {
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true,
+ "text": {}
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "((node_memory_MemTotal_bytes{instance=\"$node\",job=\"$job\"} - node_memory_MemFree_bytes{instance=\"$node\",job=\"$job\"}) / (node_memory_MemTotal_bytes{instance=\"$node\",job=\"$job\"} )) * 100",
+ "format": "time_series",
+ "hide": true,
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ },
+ {
+ "expr": "100 - ((node_memory_MemAvailable_bytes{instance=\"$node\",job=\"$job\"} * 100) / node_memory_MemTotal_bytes{instance=\"$node\",job=\"$job\"})",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "refId": "B",
+ "step": 900
+ }
+ ],
+ "title": "RAM Used",
+ "type": "gauge"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "description": "Used Swap",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "thresholds"
+ },
+ "mappings": [
+ {
+ "id": 0,
+ "op": "=",
+ "text": "N/A",
+ "type": 1,
+ "value": "null"
+ }
+ ],
+ "max": 100,
+ "min": 0,
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 10
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 25
+ }
+ ]
+ },
+ "unit": "percent"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 3,
+ "x": 12,
+ "y": 4
+ },
+ "id": 62,
+ "links": [],
+ "options": {
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true,
+ "text": {}
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "((node_memory_SwapTotal_bytes{instance=\"$node\",job=\"$job\"} - node_memory_SwapFree_bytes{instance=\"$node\",job=\"$job\"}) / (node_memory_SwapTotal_bytes{instance=\"$node\",job=\"$job\"} )) * 100",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "title": "SWAP Used",
+ "type": "gauge"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "description": "Used Root FS",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "thresholds"
+ },
+ "decimals": 1,
+ "mappings": [
+ {
+ "id": 0,
+ "op": "=",
+ "text": "N/A",
+ "type": 1,
+ "value": "null"
+ }
+ ],
+ "max": 100,
+ "min": 0,
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 80
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 90
+ }
+ ]
+ },
+ "unit": "percent"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 3,
+ "x": 15,
+ "y": 4
+ },
+ "id": 64,
+ "links": [],
+ "options": {
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true,
+ "text": {}
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "100 - ((node_filesystem_avail_bytes{instance=\"$node\",job=\"$job\",mountpoint=\"/\",fstype!=\"rootfs\"} * 100) / node_filesystem_size_bytes{instance=\"$node\",job=\"$job\",mountpoint=\"/\",fstype!=\"rootfs\"})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "title": "Root FS Used",
+ "type": "gauge"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "Prometheus",
+ "description": "Total number of CPU cores",
+ "fieldConfig": {
+ "defaults": {},
+ "overrides": []
+ },
+ "format": "short",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 2,
+ "w": 2,
+ "x": 18,
+ "y": 4
+ },
+ "id": 66,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "count(count(node_cpu_seconds_total{instance=\"$node\",job=\"$job\"}) by (cpu))",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "thresholds": "",
+ "title": "CPU Cores",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 1,
+ "description": "System uptime",
+ "fieldConfig": {
+ "defaults": {},
+ "overrides": []
+ },
+ "format": "s",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 2,
+ "w": 4,
+ "x": 20,
+ "y": 4
+ },
+ "hideTimeOverride": true,
+ "id": 70,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "$$hashKey": "object:1094",
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "$$hashKey": "object:1095",
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "null",
+ "nullText": null,
+ "postfix": "s",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "node_time_seconds{instance=\"$node\",job=\"$job\"} - node_boot_time_seconds{instance=\"$node\",job=\"$job\"}",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 1800
+ }
+ ],
+ "thresholds": "",
+ "title": "Uptime",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "$$hashKey": "object:1097",
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 0,
+ "description": "Total RootFS",
+ "fieldConfig": {
+ "defaults": {},
+ "overrides": []
+ },
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 2,
+ "w": 2,
+ "x": 18,
+ "y": 6
+ },
+ "id": 68,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "node_filesystem_size_bytes{instance=\"$node\",job=\"$job\",mountpoint=\"/\",fstype!=\"rootfs\"}",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "thresholds": "70,90",
+ "title": "RootFS Total",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 0,
+ "description": "Total RAM",
+ "fieldConfig": {
+ "defaults": {},
+ "overrides": []
+ },
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 2,
+ "w": 2,
+ "x": 20,
+ "y": 6
+ },
+ "id": 82,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "70%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "node_memory_MemTotal_bytes{instance=\"$node\",job=\"$job\"}",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "thresholds": "",
+ "title": "RAM Total",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "Prometheus",
+ "decimals": 0,
+ "description": "Total SWAP",
+ "fieldConfig": {
+ "defaults": {},
+ "overrides": []
+ },
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 2,
+ "w": 2,
+ "x": 22,
+ "y": 6
+ },
+ "id": 74,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "70%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "node_memory_SwapTotal_bytes{instance=\"$node\",job=\"$job\"}",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 900
+ }
+ ],
+ "thresholds": "",
+ "title": "SWAP Total",
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {
+ "Busy": "#EAB839",
+ "Busy Iowait": "#890F02",
+ "Busy other": "#1F78C1",
+ "Idle": "#052B51",
+ "Idle - Waiting for something to happen": "#052B51",
+ "guest": "#9AC48A",
+ "idle": "#052B51",
+ "iowait": "#EAB839",
+ "irq": "#BF1B00",
+ "nice": "#C15C17",
+ "softirq": "#E24D42",
+ "steal": "#FCE2DE",
+ "system": "#508642",
+ "user": "#5195CE"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "description": "Basic CPU info",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 4,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 6,
+ "x": 0,
+ "y": 8
+ },
+ "hiddenSeries": false,
+ "id": 76,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": 250,
+ "sort": null,
+ "sortDesc": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": true,
+ "pluginVersion": "7.5.7",
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "$$hashKey": "object:3565",
+ "alias": "Busy Iowait",
+ "color": "#890F02"
+ },
+ {
+ "$$hashKey": "object:3566",
+ "alias": "Idle",
+ "color": "#7EB26D"
+ },
+ {
+ "$$hashKey": "object:3567",
+ "alias": "Busy System",
+ "color": "#EAB839"
+ },
+ {
+ "$$hashKey": "object:3568",
+ "alias": "Busy User",
+ "color": "#0A437C"
+ },
+ {
+ "$$hashKey": "object:3569",
+ "alias": "Busy Other",
+ "color": "#B877D9"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "exemplar": true,
+ "expr": "sum by (instance)(irate(node_cpu_seconds_total{mode=\"system\",instance=\"$node\",job=\"$job\"}[5m])) * 100",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "Busy System",
+ "refId": "A",
+ "step": 240
+ },
+ {
+ "expr": "sum by (instance)(irate(node_cpu_seconds_total{mode='user',instance=\"$node\",job=\"$job\"}[5m])) * 100",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "Busy User",
+ "refId": "B",
+ "step": 240
+ },
+ {
+ "expr": "sum by (instance)(irate(node_cpu_seconds_total{mode='iowait',instance=\"$node\",job=\"$job\"}[5m])) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Busy Iowait",
+ "refId": "C",
+ "step": 240
+ },
+ {
+ "expr": "sum by (instance)(irate(node_cpu_seconds_total{mode=~\".*irq\",instance=\"$node\",job=\"$job\"}[5m])) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Busy IRQs",
+ "refId": "D",
+ "step": 240
+ },
+ {
+ "expr": "sum (irate(node_cpu_seconds_total{mode!='idle',mode!='user',mode!='system',mode!='iowait',mode!='irq',mode!='softirq',instance=\"$node\",job=\"$job\"}[5m])) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Busy Other",
+ "refId": "E",
+ "step": 240
+ },
+ {
+ "expr": "sum by (mode)(irate(node_cpu_seconds_total{mode='idle',instance=\"$node\",job=\"$job\"}[5m])) * 100",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "Idle",
+ "refId": "F",
+ "step": 240
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "CPU Basic",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:3600",
+ "format": "short",
+ "label": "",
+ "logBase": 1,
+ "max": "100",
+ "min": "0",
+ "show": true
+ },
+ {
+ "$$hashKey": "object:3601",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "Apps": "#629E51",
+ "Buffers": "#614D93",
+ "Cache": "#6D1F62",
+ "Cached": "#511749",
+ "Committed": "#508642",
+ "Free": "#0A437C",
+ "Hardware Corrupted - Amount of RAM that the kernel identified as corrupted / not working": "#CFFAFF",
+ "Inactive": "#584477",
+ "PageTables": "#0A50A1",
+ "Page_Tables": "#0A50A1",
+ "RAM_Free": "#E0F9D7",
+ "SWAP Used": "#BF1B00",
+ "Slab": "#806EB7",
+ "Slab_Cache": "#E0752D",
+ "Swap": "#BF1B00",
+ "Swap Used": "#BF1B00",
+ "Swap_Cache": "#C15C17",
+ "Swap_Free": "#2F575E",
+ "Unused": "#EAB839"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 2,
+ "description": "Basic memory usage",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 4,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 6,
+ "x": 6,
+ "y": 8
+ },
+ "hiddenSeries": false,
+ "id": 78,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": 350,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "$$hashKey": "object:4023",
+ "alias": "RAM Total",
+ "color": "#E0F9D7",
+ "fill": 0,
+ "stack": false
+ },
+ {
+ "$$hashKey": "object:4024",
+ "alias": "RAM Cache + Buffer",
+ "color": "#052B51"
+ },
+ {
+ "$$hashKey": "object:4025",
+ "alias": "RAM Free",
+ "color": "#7EB26D"
+ },
+ {
+ "$$hashKey": "object:4026",
+ "alias": "Avaliable",
+ "color": "#DEDAF7",
+ "fill": 0,
+ "stack": false
+ },
+ {
+ "$$hashKey": "object:4083",
+ "alias": "Polkadot",
+ "color": "rgb(247, 0, 222)",
+ "stack": false
+ },
+ {
+ "$$hashKey": "object:5303",
+ "alias": "RAM Used",
+ "color": "#FADE2A"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "namedprocess_namegroup_memory_bytes{groupname=\"[[metric_namespace]]\",memtype=\"resident\"}",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "Polkadot",
+ "refId": "F"
+ },
+ {
+ "expr": "node_memory_MemTotal_bytes{instance=\"$node\",job=\"$job\"}",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "RAM Total",
+ "refId": "A",
+ "step": 240
+ },
+ {
+ "expr": "node_memory_MemTotal_bytes{instance=\"$node\",job=\"$job\"} - node_memory_MemFree_bytes{instance=\"$node\",job=\"$job\"} - (node_memory_Cached_bytes{instance=\"$node\",job=\"$job\"} + node_memory_Buffers_bytes{instance=\"$node\",job=\"$job\"})",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "RAM Used",
+ "refId": "B",
+ "step": 240
+ },
+ {
+ "expr": "node_memory_Cached_bytes{instance=\"$node\",job=\"$job\"} + node_memory_Buffers_bytes{instance=\"$node\",job=\"$job\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "RAM Cache + Buffer",
+ "refId": "C",
+ "step": 240
+ },
+ {
+ "expr": "node_memory_MemFree_bytes{instance=\"$node\",job=\"$job\"}",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "RAM Free",
+ "refId": "D",
+ "step": 240
+ },
+ {
+ "expr": "(node_memory_SwapTotal_bytes{instance=\"$node\",job=\"$job\"} - node_memory_SwapFree_bytes{instance=\"$node\",job=\"$job\"})",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "SWAP Used",
+ "refId": "E",
+ "step": 240
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Memory Basic",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:4059",
+ "format": "bytes",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "$$hashKey": "object:4060",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "Recv_bytes_eth2": "#7EB26D",
+ "Recv_bytes_lo": "#0A50A1",
+ "Recv_drop_eth2": "#6ED0E0",
+ "Recv_drop_lo": "#E0F9D7",
+ "Recv_errs_eth2": "#BF1B00",
+ "Recv_errs_lo": "#CCA300",
+ "Trans_bytes_eth2": "#7EB26D",
+ "Trans_bytes_lo": "#0A50A1",
+ "Trans_drop_eth2": "#6ED0E0",
+ "Trans_drop_lo": "#E0F9D7",
+ "Trans_errs_eth2": "#BF1B00",
+ "Trans_errs_lo": "#CCA300",
+ "recv_bytes_lo": "#0A50A1",
+ "recv_drop_eth0": "#99440A",
+ "recv_drop_lo": "#967302",
+ "recv_errs_eth0": "#BF1B00",
+ "recv_errs_lo": "#890F02",
+ "trans_bytes_eth0": "#7EB26D",
+ "trans_bytes_lo": "#0A50A1",
+ "trans_drop_eth0": "#99440A",
+ "trans_drop_lo": "#967302",
+ "trans_errs_eth0": "#BF1B00",
+ "trans_errs_lo": "#890F02"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "Basic network info per interface",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 4,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 6,
+ "x": 12,
+ "y": 8
+ },
+ "hiddenSeries": false,
+ "id": 80,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "$$hashKey": "object:1019",
+ "alias": "/.*trans.*/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_network_receive_bytes_total{instance=\"$node\",job=\"$job\"}[5m])*8",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "recv {{device}}",
+ "refId": "A",
+ "step": 240
+ },
+ {
+ "expr": "irate(node_network_transmit_bytes_total{instance=\"$node\",job=\"$job\"}[5m])*8",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "trans {{device}} ",
+ "refId": "B",
+ "step": 240
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Network Traffic Basic",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:1065",
+ "format": "bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:1066",
+ "format": "pps",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {
+ "io time": "#890F02"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": 3,
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 4,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 6,
+ "x": 18,
+ "y": 8
+ },
+ "hiddenSeries": false,
+ "id": 88,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "max": true,
+ "min": true,
+ "rightSide": false,
+ "show": true,
+ "sort": null,
+ "sortDesc": null,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "maxPerRow": 6,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(node_disk_io_time_seconds_total{instance=\"$node\",job=\"$job\",device=~\"$diskdevices\"} [5m])",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "{{device}} - Time spent doing I/Os",
+ "refId": "A",
+ "step": 240
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "I/O Usage Times",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": false,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:3764",
+ "format": "s",
+ "label": "time",
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "$$hashKey": "object:3765",
+ "format": "s",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": false,
+ "datasource": "Prometheus",
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 15
+ },
+ "id": 52,
+ "panels": [],
+ "title": "Chain",
+ "type": "row"
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "links": [
+ {
+ "title": "",
+ "url": ""
+ }
+ ],
+ "mappings": [
+ {
+ "id": 0,
+ "op": "=",
+ "text": "Stuck",
+ "type": 1,
+ "value": "1492283"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgb(222, 101, 171)",
+ "value": null
+ }
+ ]
+ },
+ "unit": "locale"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 11,
+ "x": 0,
+ "y": 16
+ },
+ "id": 26,
+ "links": [],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "area",
+ "justifyMode": "center",
+ "orientation": "vertical",
+ "reduceOptions": {
+ "calcs": ["last"],
+ "fields": "",
+ "values": false
+ },
+ "text": {},
+ "textMode": "auto"
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "[[metric_namespace]]_block_height{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "interval": "",
+ "legendFormat": "{{status}}",
+ "refId": "B"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Height of the chain",
+ "type": "stat"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 4,
+ "w": 6,
+ "x": 11,
+ "y": 16
+ },
+ "hiddenSeries": false,
+ "id": 14,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "[[metric_namespace]]_block_height{instance=\"[[instance]]\",network=\"[[network]]\",status=~\"finalized|sync_target\"}",
+ "legendFormat": "{{instance}} {{status}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Block Finalized",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:1885",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:1886",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "cacheTimeout": null,
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "mappings": [
+ {
+ "id": 0,
+ "op": "=",
+ "text": "N/A",
+ "type": 1,
+ "value": "null"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "semi-dark-blue",
+ "value": null
+ }
+ ]
+ },
+ "unit": "bytes"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 4,
+ "w": 7,
+ "x": 17,
+ "y": 16
+ },
+ "id": 28,
+ "links": [],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "area",
+ "justifyMode": "auto",
+ "orientation": "vertical",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "text": {},
+ "textMode": "auto"
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "expr": "rate([[metric_namespace]]_sub_libp2p_network_bytes_total{instance=\"[[instance]]\",network=\"[[network]]\"}[5m])",
+ "instant": false,
+ "interval": "",
+ "legendFormat": "{{direction}}",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Networking bytes per second",
+ "type": "stat"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "cacheTimeout": null,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 0,
+ "y": 20
+ },
+ "hiddenSeries": false,
+ "id": 18,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate([[metric_namespace]]_block_height{status=\"finalized\",instance=\"[[instance]]\",network=\"[[network]]\"}[10m])*60",
+ "interval": "",
+ "intervalFactor": 10,
+ "legendFormat": "{{instance}} Blocks / minute",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Block rate",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:1800",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:1801",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 6,
+ "y": 20
+ },
+ "hiddenSeries": false,
+ "id": 20,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "data": "",
+ "expr": "[[metric_namespace]]_ready_transactions_number{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "hide": false,
+ "legendFormat": "{{instance}}",
+ "refId": "A",
+ "target": "txcount",
+ "type": "timeseries"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "TXs Count",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:2138",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:2139",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 12,
+ "y": 20
+ },
+ "hiddenSeries": false,
+ "id": 15,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "[[metric_namespace]]_sub_libp2p_peers_count{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Peers count",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:2223",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:2224",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 18,
+ "y": 20
+ },
+ "hiddenSeries": false,
+ "id": 36,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "[[metric_namespace]]_state_cache_bytes{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "interval": "",
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "State Cache",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:1295",
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:1296",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 0,
+ "y": 26
+ },
+ "hiddenSeries": false,
+ "id": 34,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "[[metric_namespace]]_sync_queued_blocks{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "interval": "",
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Queued Blocks (sync)",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:1970",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:1971",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 6,
+ "y": 26
+ },
+ "hiddenSeries": false,
+ "id": 8,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate([[metric_namespace]]_block_height{status=\"finalized\",instance=\"[[instance]]\",network=\"[[network]]\"}[10m])/rate([[metric_namespace]]_block_height{status=\"finalized\",instance=\"[[instance]]\",network=\"[[network]]\"}[1m])",
+ "intervalFactor": 1,
+ "legendFormat": "rate[10m] / rate[1m]",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Relative Block Production Speed",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:410",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:411",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 12,
+ "y": 26
+ },
+ "hiddenSeries": false,
+ "id": 10,
+ "interval": "",
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "increase([[metric_namespace]]_block_height{instance=\"[[instance]]\",network=\"[[network]]\",status=~\"finalized|sync_target\"}[1m])",
+ "intervalFactor": 5,
+ "legendFormat": "{{status}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Blocks Av per min",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:2972",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:2973",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "cacheTimeout": null,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 18,
+ "y": 26
+ },
+ "hiddenSeries": false,
+ "id": 17,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "scalar([[metric_namespace]]_block_height{status=\"best\",instance=\"[[instance]]\",network=\"[[network]]\"})-scalar([[metric_namespace]]_block_height{status=\"finalized\",instance=\"[[instance]]\",network=\"[[network]]\"})",
+ "intervalFactor": 2,
+ "legendFormat": "[[hostname]]",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Diff -> ( Best Block - Finalized )",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:3057",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:3058",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 0,
+ "y": 32
+ },
+ "hiddenSeries": false,
+ "id": 90,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate([[metric_namespace]]_tasks_polling_started_total{instance=\"[[instance]]\",network=\"[[network]]\"}[1m]) - rate([[metric_namespace]]_tasks_polling_duration_count{instance=\"[[instance]]\",network=\"[[network]]\"}[1m])",
+ "interval": "",
+ "legendFormat": "{{task_name}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Running Tasks",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:956",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:957",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 6,
+ "y": 32
+ },
+ "hiddenSeries": false,
+ "id": 24,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "[[metric_namespace]]_sub_libp2p_connections{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "hide": false,
+ "interval": "",
+ "legendFormat": "{{instance}} connections",
+ "refId": "A"
+ },
+ {
+ "expr": "[[metric_namespace]]_sub_libp2p_is_major_syncing{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "hide": false,
+ "legendFormat": "{{instance}} syncing",
+ "refId": "B"
+ },
+ {
+ "expr": "[[metric_namespace]]_sub_libp2p_kbuckets_num_nodes{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "hide": false,
+ "legendFormat": "{{instance}} num_nodes",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "sub_libp2p",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:3225",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:3226",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 12,
+ "y": 32
+ },
+ "hiddenSeries": false,
+ "id": 22,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "[[metric_namespace]]_sync_extra_justifications{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "interval": "",
+ "legendFormat": "{{instance}} active",
+ "refId": "A"
+ },
+ {
+ "expr": "[[metric_namespace]]_sync_extra_justifications_failed{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "legendFormat": "{{instance}} failed",
+ "refId": "B"
+ },
+ {
+ "expr": "[[metric_namespace]]_sync_extra_justifications_importing{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "legendFormat": "{{instance}} importing",
+ "refId": "C"
+ },
+ {
+ "expr": "[[metric_namespace]]_sync_extra_justifications_pending{instance=\"[[instance]]\",network=\"[[network]]\"}",
+ "legendFormat": "{{instance}} pending",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Sync justifications",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:2339",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:2340",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "links": []
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 18,
+ "y": 32
+ },
+ "hiddenSeries": false,
+ "id": 92,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate([[metric_namespace]]_finality_grandpa_communication_gossip_validator_messages{instance=\"[[instance]]\",network=\"[[network]]\"}[5m])",
+ "interval": "",
+ "legendFormat": "{{message}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "GRANDPA Validator Messages",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:312",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:313",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "datasource": "Prometheus",
+ "fieldConfig": {
+ "defaults": {
+ "mappings": [],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "none"
+ },
+ "overrides": []
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 0,
+ "y": 38
+ },
+ "id": 94,
+ "options": {
+ "displayMode": "lcd",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": ["lastNotNull"],
+ "fields": "",
+ "values": false
+ },
+ "showUnfilled": true,
+ "text": {}
+ },
+ "pluginVersion": "7.5.7",
+ "targets": [
+ {
+ "exemplar": true,
+ "expr": "namedprocess_namegroup_thread_count{groupname=\"[[chain_process]]\"}",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{threadname}}",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Thread count",
+ "type": "bargauge"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "Prometheus",
+ "decimals": null,
+ "fieldConfig": {
+ "defaults": {
+ "unit": "s"
+ },
+ "overrides": []
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 20,
+ "x": 4,
+ "y": 38
+ },
+ "hiddenSeries": false,
+ "id": 96,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": true,
+ "min": true,
+ "rightSide": true,
+ "show": true,
+ "sideWidth": null,
+ "sort": "current",
+ "sortDesc": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 1,
+ "nullPointMode": "null",
+ "options": {
+ "alertThreshold": true
+ },
+ "percentage": false,
+ "pluginVersion": "7.5.7",
+ "pointradius": 2,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "exemplar": true,
+ "expr": "irate(namedprocess_namegroup_thread_cpu_seconds_total{groupname=\"[[chain_process]]\"} [5m])",
+ "format": "time_series",
+ "hide": false,
+ "instant": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{threadname}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeRegions": [],
+ "timeShift": null,
+ "title": "Thread CPU time",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "$$hashKey": "object:3476",
+ "format": "s",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "$$hashKey": "object:3477",
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "refresh": "5s",
+ "schemaVersion": 27,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+ "selected": true,
+ "text": "substrate",
+ "value": "substrate"
+ },
+ "description": null,
+ "error": null,
+ "hide": 0,
+ "includeAll": false,
+ "label": "Chain Metrics",
+ "multi": false,
+ "name": "metric_namespace",
+ "options": [
+ {
+ "selected": true,
+ "text": "substrate",
+ "value": "substrate"
+ },
+ {
+ "selected": false,
+ "text": "polkadot",
+ "value": "polkadot"
+ }
+ ],
+ "query": "substrate, polkadot",
+ "queryValue": "",
+ "skipUrlSync": false,
+ "type": "custom"
+ },
+ {
+ "allValue": null,
+ "current": {
+ "selected": false,
+ "text": "None",
+ "value": ""
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(network)",
+ "description": "Default value: None",
+ "error": null,
+ "hide": 2,
+ "includeAll": false,
+ "label": "Chain Network",
+ "multi": false,
+ "name": "network",
+ "options": [],
+ "query": {
+ "query": "label_values(network)",
+ "refId": "Prometheus-network-Variable-Query"
+ },
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "selected": true,
+ "text": "duniter-rpc:9615",
+ "value": "duniter-rpc:9615"
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(instance)",
+ "description": "Chain instance server host - default value: localhost:9615",
+ "error": null,
+ "hide": 0,
+ "includeAll": false,
+ "label": "Chain Instance Host",
+ "multi": false,
+ "name": "instance",
+ "options": [],
+ "query": {
+ "query": "label_values(instance)",
+ "refId": "Prometheus-instance-Variable-Query"
+ },
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "current": {
+ "selected": false,
+ "text": "Prometheus",
+ "value": "Prometheus"
+ },
+ "description": null,
+ "error": null,
+ "hide": 2,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "DS_PROMETHEUS",
+ "options": [],
+ "query": "prometheus",
+ "queryValue": "",
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": {
+ "selected": false,
+ "text": "None",
+ "value": ""
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(node_uname_info, job)",
+ "description": "Node exporter server job - default value: node_exporter",
+ "error": null,
+ "hide": 2,
+ "includeAll": false,
+ "label": "Server Job",
+ "multi": false,
+ "name": "job",
+ "options": [],
+ "query": {
+ "query": "label_values(node_uname_info, job)",
+ "refId": "Prometheus-job-Variable-Query"
+ },
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "selected": false,
+ "text": "None",
+ "value": ""
+ },
+ "datasource": "Prometheus",
+ "definition": "label_values(node_uname_info{job=\"$job\"}, instance)",
+ "description": "Node exporter server host - default value: localhost:9100",
+ "error": null,
+ "hide": 2,
+ "includeAll": false,
+ "label": "Server Host",
+ "multi": false,
+ "name": "node",
+ "options": [],
+ "query": {
+ "query": "label_values(node_uname_info{job=\"$job\"}, instance)",
+ "refId": "Prometheus-node-Variable-Query"
+ },
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "selected": false,
+ "text": "[a-z]+|nvme[0-9]+n[0-9]+",
+ "value": "[a-z]+|nvme[0-9]+n[0-9]+"
+ },
+ "description": null,
+ "error": null,
+ "hide": 2,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "diskdevices",
+ "options": [
+ {
+ "selected": true,
+ "text": "[a-z]+|nvme[0-9]+n[0-9]+",
+ "value": "[a-z]+|nvme[0-9]+n[0-9]+"
+ }
+ ],
+ "query": "[a-z]+|nvme[0-9]+n[0-9]+",
+ "skipUrlSync": false,
+ "type": "custom"
+ },
+ {
+ "allValue": null,
+ "current": {
+ "selected": true,
+ "text": "duniter",
+ "value": "duniter"
+ },
+ "description": "Name of the process of your node",
+ "error": null,
+ "hide": 0,
+ "includeAll": false,
+ "label": "Chain Process Name",
+ "multi": false,
+ "name": "chain_process",
+ "options": [
+ {
+ "selected": false,
+ "text": "duniter",
+ "value": "duniter"
+ }
+ ],
+ "query": "duniter",
+ "queryValue": "",
+ "skipUrlSync": false,
+ "type": "custom"
+ }
+ ]
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ]
+ },
+ "timezone": "",
+ "title": "Polkadot Essentials 2021",
+ "uid": "bLdSubstrates",
+ "version": 1
+}
diff --git a/docker/compose/validator-prod/process-exporter/config.yml b/docker/compose/validator-prod/process-exporter/config.yml
new file mode 100644
index 0000000000000000000000000000000000000000..b68afb5462020cbdaacdb7eb993267218b2c3b88
--- /dev/null
+++ b/docker/compose/validator-prod/process-exporter/config.yml
@@ -0,0 +1,4 @@
+process_names:
+ - name: "{{.Comm}}"
+ cmdline:
+ - '.+'
diff --git a/docker/compose/validator-prod/prometheus/alert.yml b/docker/compose/validator-prod/prometheus/alert.yml
new file mode 100644
index 0000000000000000000000000000000000000000..851e696eaf051966d93d2daae9d18318b1989bdf
--- /dev/null
+++ b/docker/compose/validator-prod/prometheus/alert.yml
@@ -0,0 +1,20 @@
+groups:
+ - name: alert_rules
+ rules:
+ - alert: InstanceDown
+ expr: up == 0
+ for: 5m
+ labels:
+ severity: critical
+ annotations:
+ summary: "Instance $labels.instance down"
+ description: "[{{ $labels.instance }}] of job [{{ $labels.job }}] has been down for more than 1 minute."
+
+ - alert: HostHighCpuLoad
+ expr: 100 - (avg by(instance)(rate(node_cpu_seconds_total{mode="idle"}[2m])) * 100) > 80
+ for: 0m
+ labels:
+ severity: warning
+ annotations:
+ summary: Host high CPU load (instance bLd Kusama)
+ description: "CPU load is > 80%\n VALUE = {{ $value }}\n LABELS: {{ $labels }}"
diff --git a/docker/compose/validator-prod/prometheus/prometheus.yml b/docker/compose/validator-prod/prometheus/prometheus.yml
new file mode 100644
index 0000000000000000000000000000000000000000..dcd40f77bf6249d0c279cfb48690673e0b776e3c
--- /dev/null
+++ b/docker/compose/validator-prod/prometheus/prometheus.yml
@@ -0,0 +1,26 @@
+rule_files:
+ - alert.yml
+
+alerting:
+ alertmanagers:
+ - static_configs:
+ - targets:
+ - localhost:9093
+
+scrape_configs:
+ - job_name: "prometheus"
+ scrape_interval: 5s
+ static_configs:
+ - targets: ["prometheus:9090"]
+ - job_name: "duniter-validator"
+ scrape_interval: 5s
+ static_configs:
+ - targets: ["duniter-validator:9615"]
+ - job_name: "node_exporter"
+ scrape_interval: 5s
+ static_configs:
+ - targets: ["host.docker.internal:9100"]
+ - job_name: "process-exporter"
+ scrape_interval: 5s
+ static_configs:
+ - targets: ["process-exporter:9256"]