[ACCUMULO-27] issues found during scale random-walk testing

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 1.4.0
• Component/s: client, master, tserver
• Labels: 14_qa_bug

Environment

running 20 random walkers on a medium-sized cluster; also experienced hardware failures that failed to kill the tablet servers' lock, but prevented the server from being responsive (bad disk)

• Assignee: Billie Rinaldi
• Reporter: Eric C. Newton
• Created: 13/Oct/11 11:20
• Updated: 14/Feb/12 22:27
• Resolved: 13/Oct/11 19:56

Description

Experienced several small problems while testing accumulo on cluster using random walk testing:

• zookeeper lock was not removed for a hung tablet server (see ACCUMULO-16)
• spurious message about a merge not in progress
• NPE during a trivial merge
• balancer would not run when a tablet server was unresponsive
• NPE reading permissions from a user being deleted
• Time duration of zero in "list scans" output had html mark-up
• Tables were created without the Versioning Iterator
• The example policy file needs to allow accumulo classes to remove files in /tmp (memory dumps can be put in /tmp for isolation purposes)