Overview

Kupe runs Mimir, Loki, and Grafana for every tenant, so you get metrics, logs, and dashboards without installing or operating any of it yourself.

This section covers what the platform collects automatically, where that data lives, and how to extend it: your own metrics, dashboards, alert rules, and notification routes.

What you get automatically

• tenant-scoped metrics in Mimir
• tenant-scoped workload logs in Loki
• a Grafana org with metrics, logs, and Alertmanager datasources
• baseline dashboards for clusters, namespaces, workloads, and storage
• managed alerting and notification routing infrastructure

Operating model

1. Kupe collects baseline telemetry from every managed cluster.
2. Metrics and logs are stored in shared backends but isolated to your tenant.
3. Grafana exposes that data through dashboards and Explore.
4. Alert rules are evaluated centrally and notifications are routed through Alertmanager.
5. Your team adds app-specific metrics, logs, dashboards, and alert rules as needed.

What your team is responsible for

• exposing application metrics endpoints
• emitting useful structured logs
• creating custom dashboards when the defaults are not enough
• defining PrometheusRule resources for app-specific alerting
• configuring receivers and routing rules for your team

Troubleshooting path

When something looks wrong, work through the data in this order:

1. Start with Grafana Dashboards to see the scope and timing of the problem.
2. Use Metrics to narrow the signal: resource usage, latency, and error rate.
3. Use Logs to find the specific cause in a workload.
4. Check cluster events to confirm what actually changed. Every cluster overview page in console.kupe.cloud has an Events panel: the standard Kubernetes event stream (Type, Reason, Object, Message, Last Seen), sortable and filterable. It covers what metrics and logs do not show directly, such as an evicted pod, a failed scheduling decision, or an image pull error.

From dashboards to alerts

Once you know what “healthy” looks like for a workload, add Alerting and Notifications so the same signals page your team automatically instead of waiting for someone to check a dashboard.

Pages in this section

• Metrics: query and troubleshoot metrics in Grafana
• Grafana Dashboards: use the built-in dashboards or deliver your own
• Logs: search workload logs with the tenant Loki datasource
• Alerting: work with managed rules and custom PrometheusRule resources
• Notifications: send alerts to Slack, PagerDuty, Teams, email, or webhooks