We're updating the issue view to help you get more done. 

Locality not good after major compaction

Description

After major compaction, not all regions show high locality ratios (some are 1 but some are still 0).

User already set the following in the cluster:
1. spark.dynamicAllocation.minExecutors = [2 x Number of Region Servers]
2. spark.locality.wait = 60s

Having higher locality ratio will help query performance. The user typically want to performance major compactions on a regular cadence, analyze tables, then query.

Test Plan

[Template: please modify accordingly] Testing considerations and plan: 1. User impact/behavior changes (e.g., error message change) 2. Variation consideration (e.g., Count/Sum/Max/Min/Avg, Parquet/ORC formats) 3. Platform consideration (Cloudera/HDP/MapR, kerberos/ldap, Spark versions) 4. Upgrade/backdown compatibility (e.g., data dictionary changes) 5. Client connectivity impact (JDBC/ODBC) 6. Configuration changes needed

Status

Assignee

Jun Yuan

Reporter

Wei Tang

Fix Build

1825

Components

Sprint

None

Fix versions

Priority

Normal