Large keys occur when complex data structures — such as List, Hash, or sorted set (Zset) — grow too large due to poor key design. Because Redis Open-Source Edition uses a single-threaded model, retrieving or deleting a large key blocks all other operations and can degrade your application. On cluster instances, large keys can also cause out of memory (OOM) errors on individual nodes.
Tair provides four methods to identify large keys. Choose based on whether you need real-time data, historical analysis, or programmatic access.
Query large keys from the console (recommended)
The real-time key statistics feature shows large keys currently in memory with minimal performance impact. Use this as your first step when investigating a live performance issue.
Analyze backup data offline
The offline key analysis feature scans historical backup files to identify large keys and hot keys. Because it runs against backup data rather than the live instance, it has no impact on online services. Use this method when you need a comprehensive key analysis without affecting production traffic.
Review audit logs
Audit logs record historical large key access events. This method does not affect the live instance. Use it to investigate past incidents or identify keys that caused performance issues at a specific point in time.
Query large keys using DAS API operations
Database Autonomy Service (DAS) provides API operations for programmatic access to key statistics. Use these operations to integrate large key monitoring into your own tooling or alerting systems.
| API operation | Description |
|---|---|
| DescribeHotBigKeys | Returns hot keys and large keys currently in memory in real time |
| DescribeTopBigKeys | Returns the top 20 large keys over a specified time period |
| DescribeTopHotKeys | Returns the top 20 hot keys over a specified time period |
| DescribeHotKeys | Returns hot keys for an instance |