[HIVE-19501] Fix HyperLogLog to be threadsafe

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Assignee: László Bodor
• Reporter: Zoltan Haindrich
• Created: 11/May/18 12:56
• Updated: 17/May/18 06:02
• Resolved: 17/May/18 06:02

Description

not sure if this is an issue in reality or not; but there are 3 static fields in HyperLogLog which are rewritten during working; if there are multiple threads are calculating HLL in the same JVM, there is a theoretical chance that they might overwrite eachothers value...

static fields:
https://github.com/apache/hive/blob/8028ce8a4cf5a03e2998c33e032a511fae770b47/standalone-metastore/src/main/java/org/apache/hadoop/hive/common/ndv/hll/HyperLogLog.java#L65

usage:
https://github.com/apache/hive/blob/8028ce8a4cf5a03e2998c33e032a511fae770b47/standalone-metastore/src/main/java/org/apache/hadoop/hive/common/ndv/hll/HyperLogLog.java#L216

Issue Links

• is related to: HIVE-18866 Semijoin and analyze: Implement a Long -> Hash64 vector fast-path