As the hosting operator, use PD AWS DocumentDB’s customer graphs alongside your team’s application checks. CPU, connections, freeable memory and storage used describe different pressures; use the timing and units to decide what needs investigation.

Customer steps: read the service graphs

1. Open the active customer service and confirm its database and AWS region.
2. Choose Usage Graphs where your hosting business makes that section available.
3. Review CPU utilisation, Database connections, Freeable memory and Storage used for the displayed period.
4. Compare the graph timing with the application’s slow requests and recent Activity.
5. Keep the affected period and visible values before making changes.

The charts request DocumentDB CloudWatch data when opened. They are not a promise of continuous live monitoring. CPU is a percentage, connections are a count, and memory and storage have their own displayed scales. Storage used reports the cluster’s stored volume, not free capacity or a fixed disk allowance. Compute metrics are requested at cluster scope and can aggregate instances; they are not measurements for one collection or application. Compare each graph with its normal range rather than treating all four as the same measurement.

Treat missing data separately

A new, stopped or unreachable database may have no usable recent series. A read failure or missing CloudWatch permission can also leave a graph empty. Blank data does not mean zero CPU, unlimited storage or no connections. Your hosting team should compare the intended cluster and provider metric read before changing the plan.

Choose the next investigation

Sustained CPU demand, connection pressure and storage growth need different checks. Review workload demand, application queries and available capacity with the administrator before an upgrade. A larger instance does not by itself establish that the cause of a slowdown is resolved. If the graph reads correctly but the application remains slow, continue workload diagnosis with your database administrator. Keep a failed metric read or display mismatch separate from an accurate high reading.

See the PD AWS DocumentDB Blesta guide for related tasks, or explore PD AWS DocumentDB.