Dashboard Standards

Owner: Anchor MSP Operations Lead Last reviewed: 2026-05-24

Purpose

Define standards for Grafana dashboards used to monitor systems under Anchor managed production. Consistent dashboards enable faster incident response, easier onboarding, and reliable operational visibility.

Scope

All Grafana dashboards created for systems managed by Anchor MSP. This includes system dashboards, application dashboards, security dashboards, and overview dashboards.

Policy

Dashboard Naming Convention

All dashboards follow the naming pattern: {client}-{system}-{category}

Examples:

• egi-api-overview
• egi-api-performance
• mast-web-errors
• anchor-infrastructure-nodes

1. client — The client or project identifier (e.g., egi, mast, anchor).
2. system — The specific system or service (e.g., api, web, database, infrastructure).
3. category — The dashboard focus area (e.g., overview, performance, errors, security, nodes).

Dashboard names are lowercase with hyphens. No spaces, no underscores.

Folder Structure

Grafana dashboards are organized into folders by client:

Grafana/
  Anchor/          -- Internal Anchor infrastructure dashboards
  EGI/             -- EGI client dashboards
  Mast/            -- Mast client dashboards
  Shared/          -- Cross-client dashboards (e.g., fleet overview)

1. Each client folder contains only dashboards for that client's systems.
2. The Shared folder contains dashboards that aggregate data across clients (e.g., fleet-wide node health).
3. New folders are created only when a new client is onboarded. Do not create per-system or per-category subfolders.

Required Panels

Every managed system must have at minimum the following dashboard panels. These may be on one or more dashboards.

System Overview Dashboard

Panel	Data Source	Description
System uptime	up metric	Current up/down status with uptime percentage
CPU usage	node_cpu_seconds_total	CPU utilization over time, by mode
Memory usage	node_memory_MemAvailable_bytes	Used vs. available memory
Disk usage	node_filesystem_avail_bytes	Disk utilization per mount point
Network I/O	node_network_*_bytes_total	Inbound and outbound traffic
Active alerts	Alertmanager datasource	Currently firing alerts for this system

Application Health Dashboard

Panel	Data Source	Description
Request rate	http_requests_total	Requests per second over time
Error rate	http_requests_total (filtered by 5xx)	Error percentage over time
Response time	http_request_duration_seconds	P50, P95, P99 latency
Health check	app_up or health endpoint	Current application health status

Error and Log Dashboard

Panel	Data Source	Description
Error log stream	Loki	Live stream of ERROR and FATAL log entries
Error count over time	Loki	Count of error-level logs, bucketed by time
Top errors	Loki	Most frequent error messages

Dashboard Provisioning

1. Dashboards are provisioned as code using Grafana provisioning or the Grafana API. Manual dashboard creation via the Grafana UI is permitted for prototyping but must be exported and committed to the infrastructure repository before it is considered production.
2. Dashboard JSON files are stored in the infrastructure repository under grafana/dashboards/{client}/.
3. Dashboard changes follow the standard Production Change Policy.
4. Variables (template variables) are used for reusable dashboards. Common variables include $host, $service, and $interval.

Alert Integration

1. Every dashboard with alerting panels must link to the relevant Alertmanager silence and alert history.
2. Alert annotations on dashboards (vertical lines marking alert start/end) are enabled for all time-series panels.
3. Dashboard links to related runbooks are included in the dashboard description or as panel links.

Dashboard Access

1. All Anchor operators have Editor access to dashboards in client folders they manage.
2. Client teams (EGI, Mast) receive Viewer access to their dashboards. They can view but not modify.
3. The Shared folder is read-only for all non-Anchor users.
4. Dashboard access is reviewed as part of the quarterly Access Control review.

Dashboard Review

1. Dashboards are reviewed quarterly alongside the access review.
2. Review checks:
   • Are all panels showing data? (No broken queries or "No data" panels.)
   • Are thresholds and alert conditions still appropriate?
   • Are there unused dashboards that should be archived or removed?
3. Dashboard review findings are documented and action items assigned.

Exceptions

Temporary dashboards for incident investigation do not need to follow naming or provisioning standards. They must be deleted or formalized within 7 days of the incident's closure.