This is expected behavior. When Tair (Redis OSS-compatible) initializes a new instance, the system automatically generates database metadata. This metadata occupies a small amount of memory that is visible in the Tair (Redis OSS-compatible) console.
What uses the memory
Each Tair instance generates metadata at startup. The amount depends on your instance type, version, and configuration.
| Source | Memory usage |
|---|---|
| Standard instance (base metadata) | 30 MB to 50 MB |
| Tair (Enterprise Edition) additional metadata | Dozens of MB beyond the base amount |
| Per-database metadata (version 24.3.0.0 and later) | Approximately 5 MB to 6 MB per logical database, distributed evenly across all shards |
| Cluster instance metadata | 30 MB to 80 MB per shard; total = sum across all shards |
Standard instances follow the same behavior as Redis—some metadata is always generated when a database is initialized.
Tair (Enterprise Edition) instances generate additional metadata beyond the base amount. This extra metadata supports query acceleration and collects statistics about hotkeys and large keys.
Version 24.3.0.0 and later: Tair DRAM-based instances compatible with Redis 6.0 can roll back within seconds when imperceptible scaling fails. To support this, Tair logs additional metadata for each logical database. If your instance uses many logical databases or frequently runs the SELECT command to switch between databases, this metadata grows accordingly.
Cluster instances generate metadata independently on each shard. The total metadata for a cluster instance equals the sum across all its shards.
Reduce metadata memory usage
If your instance runs version 24.3.0.0 or later and metadata memory usage is a concern, upgrade to version 24.3.2.2 or later. Starting version 24.3.2.2, Tair includes deep optimizations that significantly reduce metadata memory consumption.
To check your current version and upgrade, go to the Tair (Redis OSS-compatible) console.