[ https://issues.apache.org/jira/browse/HDFS-12416?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Suhan Mao updated HDFS-12416: ----------------------------- Attachment: HDFS-12416.patch > BlockPlacementPolicyDefault will cause NN shutdown if log level is changed > -------------------------------------------------------------------------- > > Key: HDFS-12416 > URL: https://issues.apache.org/jira/browse/HDFS-12416 > Project: Hadoop HDFS > Issue Type: Bug > Components: block placement > Affects Versions: 2.7.4, 3.0.0-alpha3 > Reporter: Suhan Mao > Attachments: HDFS-12416.patch > > Original Estimate: 5h > Remaining Estimate: 5h > > In BlockPlacementPolicyDefault.chooseRandom method. > The code are in below structure: > {code:java} > StringBuilder builder = null; > if (LOG.isDebugEnabled()) { > builder = debugLoggingBuilder.get(); > builder.setLength(0); > builder.append("["); > } > while(numOfReplicas > 0){ > chooseDataNode(scope, excludedNodes) > if (LOG.isDebugEnabled()) { > builder.append("\nNode ").append(NodeBase.getPath(chosenNode)) > .append(" ["); > } > } > {code} > There's a possibility that the loglevel is INFO before entering while loop, > but the loglevel is changed to DEBUG inside the loop through web UI. > In that case, builder is not initialized in the beginning and > NullPointerException will throw and this will cause NN exiting. -- This message was sent by Atlassian JIRA (v6.4.14#64029) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org