[HBASE-7550] Synchronization problem in AssignmentManager

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.94.5
• Hadoop Flags: Reviewed
• Assignee: Lars Hofhansl
• Reporter: Lars Hofhansl
• Created: 14/Jan/13 04:54
• Updated: 16/Feb/13 04:14
• Resolved: 14/Jan/13 05:32

Description

While running TestSplitTransactionOnCluster#testShouldClearRITWhenNodeFoundInSplittingState in a loop I encountered this error condition:

java.util.NoSuchElementException
        at java.util.concurrent.ConcurrentSkipListMap$Iter.advance(ConcurrentSkipListMap.java:2224)
        at java.util.concurrent.ConcurrentSkipListMap$EntryIterator.next(ConcurrentSkipListMap.java:2270)
        at java.util.concurrent.ConcurrentSkipListMap$EntryIterator.next(ConcurrentSkipListMap.java:2266)
        at java.util.TreeMap.buildFromSorted(TreeMap.java:2394)
        at java.util.TreeMap.buildFromSorted(TreeMap.java:2344)
        at java.util.TreeMap.<init>(TreeMap.java:195)
        at org.apache.hadoop.hbase.master.AssignmentManager.getRegionsInTransition(AssignmentManager.java:2843)
        at org.apache.hadoop.hbase.regionserver.TestSplitTransactionOnCluster.testShouldClearRITWhenNodeFoundInSplittingState(TestSplitTransactionOnCluster.java:637)

This can only occur when the map being copied into the TreeMap is modified concurrently.
While looking at the (0.94) code I noticed that synchronization is indeed missing from the nodeDeleted hook.