[IMPALA-2273] Make parquet scanner MAX_PAGE_HEADER_SIZE configurable

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Impala 2.1, Impala 2.2
• Fix Version/s: Impala 2.5.0, Impala 2.3.2, Impala 2.2.9
• Target Version: Impala 2.5.0, Impala 2.3.2, Impala 2.2.9
• Assignee: Bharath Vissapragada
• Reporter: Juan Yu
• Created: 31/Aug/15 22:54
• Updated: 23/Feb/16 04:36
• Resolved: 05/Nov/15 06:30

Description

Parquet file page headers can get arbitrarily large because of page statistics. For example, the statistics will store the min/max value of string columns, which might be very long strings.
Currently Impala hard coded MAX_PAGE_HEADER_SIZE to 8M so it cannot access parquet file that has large page headers.
We should make MAX_PAGE_HEADER_SIZE configurable. but we should also be careful and make sure a corrupt file could not cause Impala to try to buffer the entire file in memory.