[AVRO-1773] Infinite loop caused by race condition

• Type: Sub-task
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.7.7
• Component/s: java
• Assignee: Unassigned
• Reporter: vincent ye
• Created: 23/Dec/15 21:39
• Updated: 07/Feb/20 10:19
• Resolved: 13/Jan/16 18:18

Description

org.apache.avro.LogicalTypes#fromSchemaIgnoreInvalid lookups and update CACHE. CACHE is backed by hashmap without synchronization. In a multithreaded environment, it causes a infinite loop in hashmap lookup. The race condition problem is described in the following blog http://mailinator.blogspot.com/2009/06/beautiful-race-condition.html.

I experience this infinite loop in Spark jobs with high concurrency.

Issue Links

• duplicates: AVRO-1781 Schema.parse is not thread safe