[QPID-8184] [linearstore] Recovery intermittently produces JERR_EFP_BADEFPDIRNAME error followed by core

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Component/s: C++ Broker
• Assignee: Kim van der Riet
• Reporter: Kim van der Riet
• Created: 07/May/18 20:38
• Updated: 15/May/18 14:06
• Resolved: 15/May/18 14:01

Description

Some users are experiencing difficulty recovering the store, especially when there are a large  number of queues (several thousand). The log files show the following pattern:

JERR_EFP_BADEFPDIRNAME in which some arbitrary number which is not divisible by 4 is being used as the EFP file size (called EFP directory in the log), followed by a segfault:

May 4 18:55:00 somehostname qpidd[6240]: 2018-05-04 18:55:00 [Store] warning Linear Store: EmptyFilePool create failed: jexception 0x0d03 EmptyFilePool::fileSizeKbFromDirName() threw JERR_EFP_BADEFPDIRNAME: Bad Empty File Pool directory name (must be 'NNNk', where NNN is a number which is a multiple of 4) (Partition: 1; EFP directory: '9k')
May 4 18:55:00 somehostname kernel: qpidd[6240]: segfault at 10 ip 00007f4219af8e19 sp 00007ffc227a6350 error 4 in linearstore.so[7f4219ac4000+bd000]

 In the event that the random number is divisible by 4, a randomly sized directory containing no files may appear in the partition EFP.