torjus/nixos-servers
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects 1 Releases Wiki Activity

grafana: fix datasource UIDs for VictoriaMetrics migration
Some checks failed
Run nix flake check / flake-check (push) Has been cancelled
Details

Browse Source 
Update all dashboard datasource references from "prometheus" to
"victoriametrics" to match the declared datasource UID. Enable
prune and deleteDatasources to clean up the old Prometheus
(monitoring01) datasource from Grafana's database.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Torjus Håkestad
2026-02-17 22:23:04 +01:00
parent 95a96b2192
commit 65acf13e6f
8 changed files with 90 additions and 86 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
services/grafana/dashboards/apiary.json
View File

@@ -19,7 +19,7 @@
"title": "SSH Connections",
"type": "stat",
"gridPos": {"h": 4, "w": 6, "x": 0, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(oubliette_ssh_connections_total{job=\"apiary\"})",
@@ -51,7 +51,7 @@
"title": "Active Sessions",
"type": "stat",
"gridPos": {"h": 4, "w": 6, "x": 6, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "oubliette_sessions_active{job=\"apiary\"}",
@@ -86,7 +86,7 @@
"title": "Unique IPs",
"type": "stat",
"gridPos": {"h": 4, "w": 6, "x": 12, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "oubliette_storage_unique_ips{job=\"apiary\"}",
@@ -118,7 +118,7 @@
"title": "Total Login Attempts",
"type": "stat",
"gridPos": {"h": 4, "w": 6, "x": 18, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "oubliette_storage_login_attempts_total{job=\"apiary\"}",
@@ -150,7 +150,7 @@
"title": "SSH Connections Over Time",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"interval": "60s",
"targets": [
{
@@ -183,7 +183,7 @@
"title": "Auth Attempts Over Time",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"interval": "60s",
"targets": [
{
@@ -216,7 +216,7 @@
"title": "Sessions by Shell",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 22},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"interval": "60s",
"targets": [
{
@@ -249,7 +249,7 @@
"title": "Attempts by Country",
"type": "geomap",
"gridPos": {"h": 10, "w": 24, "x": 0, "y": 12},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "oubliette_auth_attempts_by_country_total{job=\"apiary\"}",
@@ -318,7 +318,7 @@
"title": "Session Duration Distribution",
"type": "heatmap",
"gridPos": {"h": 8, "w": 24, "x": 0, "y": 30},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"interval": "60s",
"targets": [
{
@@ -359,7 +359,7 @@
"title": "Commands Executed by Shell",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 22},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"interval": "60s",
"targets": [
{

20
services/grafana/dashboards/certificates.json
View File

@@ -16,7 +16,7 @@
"title": "Endpoints Monitored",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(probe_ssl_earliest_cert_expiry{job=\"blackbox_tls\"})",
@@ -48,7 +48,7 @@
"title": "Probe Failures",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(probe_success{job=\"blackbox_tls\"} == 0) or vector(0)",
@@ -82,7 +82,7 @@
"title": "Expiring Soon (< 7d)",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 8, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count((probe_ssl_earliest_cert_expiry{job=\"blackbox_tls\"} - time()) < 86400 * 7) or vector(0)",
@@ -116,7 +116,7 @@
"title": "Expiring Critical (< 24h)",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 12, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count((probe_ssl_earliest_cert_expiry{job=\"blackbox_tls\"} - time()) < 86400) or vector(0)",
@@ -150,7 +150,7 @@
"title": "Minimum Days Remaining",
"type": "gauge",
"gridPos": {"h": 4, "w": 8, "x": 16, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "min((probe_ssl_earliest_cert_expiry{job=\"blackbox_tls\"} - time()) / 86400)",
@@ -187,7 +187,7 @@
"title": "Certificate Expiry by Endpoint",
"type": "table",
"gridPos": {"h": 12, "w": 12, "x": 0, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "(probe_ssl_earliest_cert_expiry{job=\"blackbox_tls\"} - time()) / 86400",
@@ -253,7 +253,7 @@
"title": "Probe Status",
"type": "table",
"gridPos": {"h": 12, "w": 12, "x": 12, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "probe_success{job=\"blackbox_tls\"}",
@@ -340,7 +340,7 @@
"title": "Certificate Expiry Over Time",
"type": "timeseries",
"gridPos": {"h": 8, "w": 24, "x": 0, "y": 16},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "(probe_ssl_earliest_cert_expiry{job=\"blackbox_tls\"} - time()) / 86400",
@@ -378,7 +378,7 @@
"title": "Probe Success Rate",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 24},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "avg(probe_success{job=\"blackbox_tls\"}) * 100",
@@ -418,7 +418,7 @@
"title": "Probe Duration",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 24},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "probe_duration_seconds{job=\"blackbox_tls\"}",

48
services/grafana/dashboards/nixos-fleet.json
View File

@@ -15,7 +15,7 @@
{
"name": "tier",
"type": "query",
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"query": "label_values(nixos_flake_info, tier)",
"refresh": 2,
"includeAll": true,
@@ -30,7 +30,7 @@
"title": "Hosts Behind Remote",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(nixos_flake_revision_behind{tier=~\"$tier\"} == 1)",
@@ -65,7 +65,7 @@
"title": "Hosts Needing Reboot",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(nixos_config_mismatch{tier=~\"$tier\"} == 1)",
@@ -100,7 +100,7 @@
"title": "Total Hosts",
"type": "stat",
"gridPos": {"h": 4, "w": 3, "x": 8, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(nixos_flake_info{tier=~\"$tier\"})",
@@ -128,7 +128,7 @@
"title": "Nixpkgs Age",
"type": "stat",
"gridPos": {"h": 4, "w": 3, "x": 11, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "max(nixos_flake_input_age_seconds{input=\"nixpkgs\", tier=~\"$tier\"})",
@@ -163,7 +163,7 @@
"title": "Hosts Up-to-date",
"type": "stat",
"gridPos": {"h": 4, "w": 3, "x": 14, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(nixos_flake_revision_behind{tier=~\"$tier\"} == 0)",
@@ -192,7 +192,7 @@
"title": "Deployments (24h)",
"type": "stat",
"gridPos": {"h": 4, "w": 3, "x": 17, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(increase(homelab_deploy_deployments_total{status=\"completed\"}[24h]))",
@@ -222,7 +222,7 @@
"title": "Avg Deploy Time",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 20, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(increase(homelab_deploy_deployment_duration_seconds_sum{success=\"true\"}[24h])) / sum(increase(homelab_deploy_deployment_duration_seconds_count{success=\"true\"}[24h]))",
@@ -256,7 +256,7 @@
"title": "Fleet Status",
"type": "table",
"gridPos": {"h": 10, "w": 24, "x": 0, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "nixos_flake_info{tier=~\"$tier\"}",
@@ -430,7 +430,7 @@
"title": "Generation Age by Host",
"type": "bargauge",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 14},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sort_desc(nixos_generation_age_seconds{tier=~\"$tier\"})",
@@ -467,7 +467,7 @@
"title": "Generations per Host",
"type": "bargauge",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 14},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sort_desc(nixos_generation_count{tier=~\"$tier\"})",
@@ -501,7 +501,7 @@
"title": "Deployment Activity (Generation Age Over Time)",
"type": "timeseries",
"gridPos": {"h": 8, "w": 24, "x": 0, "y": 22},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "nixos_generation_age_seconds{tier=~\"$tier\"}",
@@ -534,7 +534,7 @@
"title": "Flake Input Ages",
"type": "table",
"gridPos": {"h": 6, "w": 12, "x": 0, "y": 30},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "max by (input) (nixos_flake_input_age_seconds)",
@@ -577,7 +577,7 @@
"title": "Hosts by Revision",
"type": "piechart",
"gridPos": {"h": 6, "w": 6, "x": 12, "y": 30},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count by (current_rev) (nixos_flake_info{tier=~\"$tier\"})",
@@ -601,7 +601,7 @@
"title": "Hosts by Tier",
"type": "piechart",
"gridPos": {"h": 6, "w": 6, "x": 18, "y": 30},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count by (tier) (nixos_flake_info)",
@@ -641,7 +641,7 @@
"title": "Builds (24h)",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 37},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(increase(homelab_deploy_build_host_total{status=\"success\"}[24h]))",
@@ -671,7 +671,7 @@
"title": "Failed Builds (24h)",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 37},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(increase(homelab_deploy_build_host_total{status=\"failure\"}[24h])) or vector(0)",
@@ -705,7 +705,7 @@
"title": "Last Build",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 8, "y": 37},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "time() - max(homelab_deploy_build_last_timestamp)",
@@ -739,7 +739,7 @@
"title": "Avg Build Time",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 12, "y": 37},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(increase(homelab_deploy_build_duration_seconds_sum[24h])) / sum(increase(homelab_deploy_build_duration_seconds_count[24h]))",
@@ -773,7 +773,7 @@
"title": "Total Hosts Built",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 16, "y": 37},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(homelab_deploy_build_duration_seconds_count)",
@@ -802,7 +802,7 @@
"title": "Build Jobs (24h)",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 20, "y": 37},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(increase(homelab_deploy_builds_total[24h]))",
@@ -832,7 +832,7 @@
"title": "Build Time by Host",
"type": "bargauge",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 41},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sort_desc(homelab_deploy_build_duration_seconds_sum / homelab_deploy_build_duration_seconds_count)",
@@ -869,7 +869,7 @@
"title": "Build Count by Host",
"type": "bargauge",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 41},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sort_desc(sum by (host) (homelab_deploy_build_host_total))",
@@ -903,7 +903,7 @@
"title": "Build Activity",
"type": "timeseries",
"gridPos": {"h": 8, "w": 24, "x": 0, "y": 49},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(increase(homelab_deploy_build_host_total{status=\"success\"}[1h]))",

16
services/grafana/dashboards/node-exporter.json
View File

@@ -11,7 +11,7 @@
{
"name": "instance",
"type": "query",
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"query": "label_values(node_uname_info, instance)",
"refresh": 2,
"includeAll": false,
@@ -26,7 +26,7 @@
"title": "CPU Usage",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "100 - (avg by(instance) (rate(node_cpu_seconds_total{mode=\"idle\", instance=~\"$instance\"}[5m])) * 100)",
@@ -55,7 +55,7 @@
"title": "Memory Usage",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "(1 - (node_memory_MemAvailable_bytes{instance=~\"$instance\"} / node_memory_MemTotal_bytes{instance=~\"$instance\"})) * 100",
@@ -84,7 +84,7 @@
"title": "Disk Usage",
"type": "gauge",
"gridPos": {"h": 8, "w": 8, "x": 0, "y": 8},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "100 - ((node_filesystem_avail_bytes{instance=~\"$instance\",mountpoint=\"/\",fstype!=\"rootfs\"} / node_filesystem_size_bytes{instance=~\"$instance\",mountpoint=\"/\",fstype!=\"rootfs\"}) * 100)",
@@ -113,7 +113,7 @@
"title": "System Load",
"type": "timeseries",
"gridPos": {"h": 8, "w": 8, "x": 8, "y": 8},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "node_load1{instance=~\"$instance\"}",
@@ -142,7 +142,7 @@
"title": "Uptime",
"type": "stat",
"gridPos": {"h": 8, "w": 8, "x": 16, "y": 8},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "time() - node_boot_time_seconds{instance=~\"$instance\"}",
@@ -161,7 +161,7 @@
"title": "Network Traffic",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 16},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "rate(node_network_receive_bytes_total{instance=~\"$instance\",device!~\"lo|veth.*|br.*|docker.*\"}[5m])",
@@ -185,7 +185,7 @@
"title": "Disk I/O",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 16},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "rate(node_disk_read_bytes_total{instance=~\"$instance\",device!~\"dm-.*\"}[5m])",

28
services/grafana/dashboards/proxmox.json
View File

@@ -15,7 +15,7 @@
{
"name": "vm",
"type": "query",
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"query": "label_values(pve_guest_info{template=\"0\"}, name)",
"refresh": 2,
"includeAll": true,
@@ -30,7 +30,7 @@
"title": "VMs Running",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(pve_up{id=~\"qemu/.*\"} * on(id) pve_guest_info{template=\"0\"} == 1)",
@@ -56,7 +56,7 @@
"title": "VMs Stopped",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(pve_up{id=~\"qemu/.*\"} * on(id) pve_guest_info{template=\"0\"} == 0)",
@@ -87,7 +87,7 @@
"title": "Node CPU",
"type": "gauge",
"gridPos": {"h": 4, "w": 4, "x": 8, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_cpu_usage_ratio{id=~\"node/.*\"} * 100",
@@ -120,7 +120,7 @@
"title": "Node Memory",
"type": "gauge",
"gridPos": {"h": 4, "w": 4, "x": 12, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_memory_usage_bytes{id=~\"node/.*\"} / pve_memory_size_bytes{id=~\"node/.*\"} * 100",
@@ -153,7 +153,7 @@
"title": "Node Uptime",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 16, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_uptime_seconds{id=~\"node/.*\"}",
@@ -180,7 +180,7 @@
"title": "Templates",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 20, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(pve_guest_info{template=\"1\"})",
@@ -206,7 +206,7 @@
"title": "VM Status",
"type": "table",
"gridPos": {"h": 10, "w": 24, "x": 0, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_guest_info{template=\"0\", name=~\"$vm\"}",
@@ -362,7 +362,7 @@
"title": "VM CPU Usage",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 14},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_cpu_usage_ratio{id=~\"qemu/.*\"} * on(id) group_left(name) pve_guest_info{template=\"0\", name=~\"$vm\"} * 100",
@@ -391,7 +391,7 @@
"title": "VM Memory Usage",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 14},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_memory_usage_bytes{id=~\"qemu/.*\"} * on(id) group_left(name) pve_guest_info{template=\"0\", name=~\"$vm\"}",
@@ -420,7 +420,7 @@
"title": "VM Network Traffic",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 22},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "rate(pve_network_receive_bytes{id=~\"qemu/.*\"}[5m]) * on(id) group_left(name) pve_guest_info{template=\"0\", name=~\"$vm\"}",
@@ -453,7 +453,7 @@
"title": "VM Disk I/O",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 22},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "rate(pve_disk_read_bytes{id=~\"qemu/.*\"}[5m]) * on(id) group_left(name) pve_guest_info{template=\"0\", name=~\"$vm\"}",
@@ -486,7 +486,7 @@
"title": "Storage Usage",
"type": "bargauge",
"gridPos": {"h": 6, "w": 12, "x": 0, "y": 30},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_disk_usage_bytes{id=~\"storage/.*\"} / pve_disk_size_bytes{id=~\"storage/.*\"} * 100",
@@ -531,7 +531,7 @@
"title": "Storage Capacity",
"type": "table",
"gridPos": {"h": 6, "w": 12, "x": 12, "y": 30},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "pve_disk_size_bytes{id=~\"storage/.*\"}",

26
services/grafana/dashboards/systemd.json
View File

@@ -15,7 +15,7 @@
{
"name": "hostname",
"type": "query",
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"query": "label_values(systemd_unit_state, hostname)",
"refresh": 2,
"includeAll": true,
@@ -30,7 +30,7 @@
"title": "Failed Units",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 0, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(systemd_unit_state{state=\"failed\", hostname=~\"$hostname\"} == 1) or vector(0)",
@@ -60,7 +60,7 @@
"title": "Active Units",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 4, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(systemd_unit_state{state=\"active\", hostname=~\"$hostname\"} == 1)",
@@ -86,7 +86,7 @@
"title": "Hosts Monitored",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 8, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(count by (hostname) (systemd_unit_state{hostname=~\"$hostname\"}))",
@@ -112,7 +112,7 @@
"title": "Total Service Restarts",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 12, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum(systemd_service_restart_total{hostname=~\"$hostname\"})",
@@ -143,7 +143,7 @@
"title": "Inactive Units",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 16, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(systemd_unit_state{state=\"inactive\", hostname=~\"$hostname\"} == 1)",
@@ -169,7 +169,7 @@
"title": "Timers",
"type": "stat",
"gridPos": {"h": 4, "w": 4, "x": 20, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "count(systemd_timer_last_trigger_seconds{hostname=~\"$hostname\"})",
@@ -195,7 +195,7 @@
"title": "Failed Units",
"type": "table",
"gridPos": {"h": 6, "w": 12, "x": 0, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "systemd_unit_state{state=\"failed\", hostname=~\"$hostname\"} == 1",
@@ -251,7 +251,7 @@
"title": "Service Restarts (Top 15)",
"type": "table",
"gridPos": {"h": 6, "w": 12, "x": 12, "y": 4},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "topk(15, systemd_service_restart_total{hostname=~\"$hostname\"} > 0)",
@@ -309,7 +309,7 @@
"title": "Active Units per Host",
"type": "bargauge",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 10},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sort_desc(count by (hostname) (systemd_unit_state{state=\"active\", hostname=~\"$hostname\"} == 1))",
@@ -339,7 +339,7 @@
"title": "NixOS Upgrade Timers",
"type": "table",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 10},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "systemd_timer_last_trigger_seconds{name=\"nixos-upgrade.timer\", hostname=~\"$hostname\"}",
@@ -429,7 +429,7 @@
"title": "Backup Timers",
"type": "table",
"gridPos": {"h": 6, "w": 12, "x": 0, "y": 18},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "systemd_timer_last_trigger_seconds{name=~\"restic.*\", hostname=~\"$hostname\"}",
@@ -524,7 +524,7 @@
"title": "Service Restarts Over Time",
"type": "timeseries",
"gridPos": {"h": 6, "w": 12, "x": 12, "y": 18},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "sum by (hostname) (increase(systemd_service_restart_total{hostname=~\"$hostname\"}[1h]))",

14
services/grafana/dashboards/temperature.json
View File

@@ -19,7 +19,7 @@
"title": "Current Temperatures",
"type": "stat",
"gridPos": {"h": 6, "w": 12, "x": 0, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "hass_sensor_temperature_celsius{entity!~\".*device_temperature\"}",
@@ -71,7 +71,7 @@
"title": "Average Home Temperature",
"type": "gauge",
"gridPos": {"h": 6, "w": 6, "x": 12, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "avg(hass_sensor_temperature_celsius{entity!~\".*device_temperature|.*server.*\"})",
@@ -108,7 +108,7 @@
"title": "Current Humidity",
"type": "stat",
"gridPos": {"h": 6, "w": 6, "x": 18, "y": 0},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "hass_sensor_humidity_percent{entity!~\".*server.*\"}",
@@ -154,7 +154,7 @@
"title": "Temperature History (30 Days)",
"type": "timeseries",
"gridPos": {"h": 10, "w": 24, "x": 0, "y": 6},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "hass_sensor_temperature_celsius{entity!~\".*device_temperature\"}",
@@ -207,7 +207,7 @@
"title": "Temperature Trend (1h rate of change)",
"type": "timeseries",
"gridPos": {"h": 8, "w": 12, "x": 0, "y": 16},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "deriv(hass_sensor_temperature_celsius{entity!~\".*device_temperature\"}[1h]) * 3600",
@@ -268,7 +268,7 @@
"title": "24h Min / Max / Avg",
"type": "table",
"gridPos": {"h": 8, "w": 12, "x": 12, "y": 16},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "min_over_time(hass_sensor_temperature_celsius{entity!~\".*device_temperature\"}[24h])",
@@ -346,7 +346,7 @@
"title": "Humidity History (30 Days)",
"type": "timeseries",
"gridPos": {"h": 8, "w": 24, "x": 0, "y": 24},
"datasource": {"type": "prometheus", "uid": "prometheus"},
"datasource": {"type": "prometheus", "uid": "victoriametrics"},
"targets": [
{
"expr": "hass_sensor_humidity_percent",