Metrics

This is the section that separates Atlas from the snapshot-based connections. Where kube-state-metrics and RabbitMQ hand back a current-state scrape, Atlas returns real historical time series in a single call — /processes/{id}/measurements with a granularity and a period.

So there is no in-app history store here, and no “collecting history…” state: pick a process and a range (1h / 6h / 24h / 7d) and the charts are simply there, back to where the range starts.

What is charted

The process metric groups cover what you would actually open a monitoring tab for:

• CPU and memory,
• connections,
• operations,
• network,
• query targeting — the ratio that tells you a query is scanning far more than it returns,
• WiredTiger tickets — the saturation signal that explains a stalled cluster.

Each measurement declares its own units, and those units drive the value formatting, so a number is never rendered as the wrong scale.

Scope: process, disk, database

A scope toggle charts the more targeted measurement families too:

• Disk — per-partition space, IOPS, latency and utilisation (/disks/{partition}/measurements);
• Database — per-database storage, objects and collection counts (/databases/{db}/measurements),

each with its own picker for which partition or database.

Add metrics

An “Add metrics” picker charts anything else from an extended catalogue of process measurements, ad-hoc. The curated groups cover the common case; the picker means the uncommon one is not a dead end.