[
https://issues.apache.org/jira/browse/HDFS-15115?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17040507#comment-17040507
]
Kihwal Lee commented on HDFS-15115:
-----------------------------------
In your test,
{code}
repl.chooseRandom(1, "/", excludeNodes, 1024L, 3, results, false, types);
{code}
{{"/"}} is not the correct root. Take a look at {{NodeBase.ROOT}} to find out.
This test case was failing in our testing because we have an internal
optimization that prevents wasting time when an unusable rack is specified.
> Namenode crash caused by NPE in BlockPlacementPolicyDefault when dynamically
> change logger to debug
> ---------------------------------------------------------------------------------------------------
>
> Key: HDFS-15115
> URL: https://issues.apache.org/jira/browse/HDFS-15115
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: wangzhixiang
> Assignee: wangzhixiang
> Priority: Major
> Fix For: 3.3.0, 3.1.4, 3.2.2, 2.10.1
>
> Attachments: HDFS-15115.001.patch, HDFS-15115.003.patch,
> HDFS-15115.004.patch, HDFS-15115.005.patch, HDFS-15115.2.patch
>
>
> To get debug info, we dynamically change the logger of
> BlockPlacementPolicyDefault to debug when namenode is running. However, the
> Namenode crashs. From the log, we find some NPE in
> BlockPlacementPolicyDefault.chooseRandom. Because *StringBuilder builder*
> will be used 4 times in BlockPlacementPolicyDefault.chooseRandom method.
> While the *builder* only initializes in the first time of this method. If we
> change the logger of BlockPlacementPolicyDefault to debug after the part, the
> *builder* in remaining part is *NULL* and cause *NPE*
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
