[BLUR-92] After adding a unit test for killing a shard server in BlurClusterTest, WAL has intermittent issue.

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.2.0
• Fix Version/s: 0.2.0
• Environment: The failure is occurring with the HDFS minicluster.
• Assignee: Unassigned
• Reporter: Aaron McCurry
• Created: 19/May/13 18:53
• Updated: 19/May/13 21:31
• Resolved: 19/May/13 21:31

Description

After killing a shard server in the BlueClusterTest, the recovered Rows from the failed shard in the test is not correct. However if an artificial sleep is put in the test before the kill to allow the commit to occur then the failover works correctly. So there seems to be an issue with the TransactionRecorder and/or playback.