zhouyingchao created HBASE-13835:
------------------------------------

             Summary: KeyValueHeap.current might be in heap when exception 
happens in pollRealKV
                 Key: HBASE-13835
                 URL: https://issues.apache.org/jira/browse/HBASE-13835
             Project: HBase
          Issue Type: Bug
            Reporter: zhouyingchao
            Assignee: zhouyingchao


In a 0.94 hbase cluster, we found a NPE with following stack:
{code}
Exception in thread "regionserver21600.leaseChecker" 
java.lang.NullPointerException
        at 
org.apache.hadoop.hbase.KeyValue$KVComparator.compare(KeyValue.java:1530)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap$KVScannerComparator.compare(KeyValueHeap.java:225)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap$KVScannerComparator.compare(KeyValueHeap.java:201)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap$KVScannerComparator.compare(KeyValueHeap.java:191)
        at 
java.util.PriorityQueue.siftDownUsingComparator(PriorityQueue.java:641)
        at java.util.PriorityQueue.siftDown(PriorityQueue.java:612)
        at java.util.PriorityQueue.poll(PriorityQueue.java:523)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap.close(KeyValueHeap.java:241)
        at 
org.apache.hadoop.hbase.regionserver.StoreScanner.close(StoreScanner.java:355)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap.close(KeyValueHeap.java:237)
        at 
org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.close(HRegion.java:4302)
        at 
org.apache.hadoop.hbase.regionserver.HRegionServer$ScannerListener.leaseExpired(HRegionServer.java:3033)
        at org.apache.hadoop.hbase.regionserver.Leases.run(Leases.java:119)
        at java.lang.Thread.run(Thread.java:662)
{code}

Before this NPE exception, there is an exception happens in pollRealKV, which 
we think is the culprit of the NPE.
{code}
ERROR org.apache.hadoop.hbase.regionserver.HRegionServer:
java.io.IOException: Could not reseek StoreFileScanner[HFileScanner for reader 
reader=....
        at 
org.apache.hadoop.hbase.regionserver.StoreFileScanner.reseek(StoreFileScanner.java:180)
        at 
org.apache.hadoop.hbase.regionserver.StoreFileScanner.enforceSeek(StoreFileScanner.java:371)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap.pollRealKV(KeyValueHeap.java:366)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap.next(KeyValueHeap.java:116)
        at 
org.apache.hadoop.hbase.regionserver.StoreScanner.next(StoreScanner.java:455)
        at 
org.apache.hadoop.hbase.regionserver.KeyValueHeap.next(KeyValueHeap.java:154)
        at 
org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.populateResult(HRegion.java:4124)
        at 
org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.nextInternal(HRegion.java:4196)
        at 
org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.nextRaw(HRegion.java:4067)
        at 
org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl.nextRaw(HRegion.java:4057)
        at 
org.apache.hadoop.hbase.regionserver.HRegionServer.internalNext(HRegionServer.java:2898)
        at 
org.apache.hadoop.hbase.regionserver.HRegionServer.next(HRegionServer.java:2833)
        at 
org.apache.hadoop.hbase.regionserver.HRegionServer.next(HRegionServer.java:2815)
        at sun.reflect.GeneratedMethodAccessor38.invoke(Unknown Source)
        at 
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
        at java.lang.reflect.Method.invoke(Method.java:597)
        at 
org.apache.hadoop.hbase.ipc.SecureRpcEngine$Server.call(SecureRpcEngine.java:337)
        at 
org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.java:1583)
{code}

Simply put, if there is an exception happens in pollRealKV( ), the 
KeyValueHeap.current might be in heap. Later on, when KeyValueHeap.close( ) is 
called, the current would be closed firstly. However, since it might still be 
in the heap, it would either be closed again or its peek() (which is null after 
it is closed) is called by the heap's poll().  Neither case is expected.

Although it is caught in 0.94, it is still in the trunk from the code. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to