Skip to content

Performance regression in covering/duplicate indexes in 20.1 #49658

@keith-mcclellan

Description

@keith-mcclellan

Describe the problem
Tested using MovR, when applying duplicate covering indexes to the promo code table for fast reads in multiple datacenters, the database seems to only use the primary index and not the index closest to the read.

To Reproduce
Step thru MovR in a multi-region config(easier to do in the 19.2 version), create duplicate indexes and pin to multi regions. p99 on 20.1 is ~20ms, on 19.2 it was <5ms

geo-partitioned-replicas-demo.sh in https://github.com/cockroachlabs/scripted-demos

Metadata

Metadata

Assignees

Labels

A-partitioningC-investigationFurther steps needed to qualify. C-label will change.C-performancePerf of queries or internals. Solution not expected to change functional behavior.

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions