[
https://issues.apache.org/jira/browse/HDFS-7066?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Arpit Agarwal resolved HDFS-7066.
---------------------------------
Resolution: Fixed
Fix Version/s: HDFS-6581
Hadoop Flags: Reviewed
+1 I committed this to the feature branch.
Thanks for fixing this Xiaoyu!
> LazyWriter#evictBlocks misses a null check for replicaState
> -----------------------------------------------------------
>
> Key: HDFS-7066
> URL: https://issues.apache.org/jira/browse/HDFS-7066
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode
> Affects Versions: HDFS-6581
> Reporter: Xiaoyu Yao
> Assignee: Xiaoyu Yao
> Priority: Minor
> Fix For: HDFS-6581
>
> Attachments: HDFS-7066.0.patch
>
>
> LazyWriter#evictBlocks (added for HDFS-6581) misses a null check for
> replicaState. As a result, there are many NPEs in the debug log under certain
> conditions.
> {code}
> 2014-09-15 14:27:10,820 DEBUG impl.FsDatasetImpl
> (FsDatasetImpl.java:evictBlocks(2335)) - Evicting block null
> 2014-09-15 14:27:10,821 WARN impl.FsDatasetImpl
> (FsDatasetImpl.java:run(2409)) - Ignoring exception in LazyWriter:
> java.lang.NullPointerException
> at
> org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.FsDatasetImpl$LazyWriter.evictBlocks(FsDatasetImpl.java:2343)
> at
> org.apache.hadoop.hdfs.server.datanode.fsdataset.impl.FsDatasetImpl$LazyWriter.run(FsDatasetImpl.java:2396)
> at java.lang.Thread.run(Thread.java:745)
> {code}
> The proposed fix is to break if there is no candidate available to evict.
> {code}
> while (iterations++ < MAX_BLOCK_EVICTIONS_PER_ITERATION &&
> transientFreeSpaceBelowThreshold()) {
> LazyWriteReplicaTracker.ReplicaState replicaState =
> lazyWriteReplicaTracker.getNextCandidateForEviction();
> ++++++++++++++++++++
> if (replicaState == null) {
> break;
> }
> ++++++++++++++++++++
> if (LOG.isDebugEnabled()) {
> LOG.debug("Evicting block " + replicaState);
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
