Tair (Redis OSS-compatible) tracks key instance health metrics — CPU utilization, memory usage, average latency, and queries per second (QPS) — in real time. Use the Performance Monitoring page to review trends for any time range within the last month and catch issues before they affect your application.
Monitoring metrics
The following metrics are available on the Performance Monitoring page:
| Metric | Description |
|---|---|
| CPU utilization | Percentage of CPU capacity used by the instance |
| Memory usage | Amount of memory consumed by the instance |
| Average latency | Mean time to process a command |
| QPS (queries per second) | Total number of commands processed per second |
View monitoring data
Log on to the console and go to the Instances page. In the top navigation bar, select the region where your instance resides. Find the instance and click its ID.
In the left-side navigation pane, click Performance Monitoring.
Select a time range. The trend charts update to show data for that period.
The maximum time range per query is three days. Data granularity in trend charts adjusts automatically based on the selected time range.
Interpret the data based on your instance architecture:
Architecture Available tabs How metrics are aggregated Master-replica (default view) Focus on master node metrics Cluster All, Data Node All tab shows aggregated data across all replica nodes or read replicas (CPU utilization = average across nodes; some metrics use sum) Read/write splitting All, Data Node All tab aggregates replica nodes; read replica metrics are not aggregated Any architecture with proxy nodes All, Proxy Node All tab shows overall proxy performance; Proxy Node tab shows per-proxy metrics
FAQ
Why do read replicas have write QPS after I enable read/write splitting?
Write QPS on read replicas does not come from client write requests. It reflects the write operations that occur when the master node synchronizes data to the read replicas.
Why does the connection count show 0 for cluster architecture instances?
In proxy mode, data nodes show 0 connections because those connections are internal — they run between proxy nodes and data nodes, are reused, and typically average between 0 and 2. To see actual client connections, check the proxy node metrics on the Proxy Node tab. You can ignore the connection count on data nodes.
API reference
| API operation | Description |
|---|---|
| DescribeHistoryMonitorValues | Queries the performance monitoring history of an instance |
What's next
If the monitoring data reveals a performance issue, the following topics can help you diagnose and resolve it: