tencent cloud

TDSQL Boundless

문서TDSQL BoundlessGeneral ReferenceDatabase Parameter Descriptiontdstore_user_cf_level0_slowdown_writes_trigger

tdstore_user_cf_level0_slowdown_writes_trigger

다운로드
포커스 모드
폰트 크기
마지막 업데이트 시간: 2026-05-26 17:00:11

Feature Description

TDStore uses a Log-Structured Merge-tree (LSM-tree) structure to organize and store data. The LSM-tree organizes data in layers, divided into level 0 through level N. The most recently written data is first flushed to disk in the data files at level 0.Periodically, the LSM-tree initiates a compaction operation to reorganize the stored data, gradually moving data from an upper layer (level i) to the next lower layer (level i+1), thereby maintaining a healthy structure for the LSM-tree. During this process, outdated data versions are cleaned up, which frees up redundant disk space to achieve optimal read and write performance.When the database's write pressure becomes too high, the compaction operation may not be able to keep up with reorganizing the newly written data. As a result, the LSM-tree's structure degrades, and read/write performance decreases. After a certain threshold is reached, TDStore will automatically enter a "write slowdown" mode (by rate-limiting write requests) or even a "write stop" state (by rejecting all write requests). At this point, TDStore's write performance has reached its bottleneck.
The tdstore_user_cf_level0_slowdown_writes_trigger parameter defines the threshold for TDStore to enter the write slowdown state. When the number of data files at LSM-tree level 0 exceeds this parameter value, TDStore enters the write slowdown state and adaptively rate-limits all write requests. By properly configuring tdstore_user_cf_level0_slowdown_writes_trigger, you can control the threshold for entering the write slowdown state, thereby optimizing system performance and resource utilization.

Attribute Description

Required
Description
Parameter Type
UINT
Default Value
64
Value Range
1 ~ 2147483647
Effective Scope
GLOBAL
Restart Required
No

도움말 및 지원

문제 해결에 도움이 되었나요?

피드백