[ https://issues.apache.org/jira/browse/HDFS-6860?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14114291#comment-14114291 ]
Arpit Agarwal commented on HDFS-6860: ------------------------------------- Thanks for fixing this [~lichangleo]. Can you also wrap each call in a {{isDebugEnabled()}} call. That way we don't incur the overhead of generating the log strings unless debug logging is turned on. e.g. {code} if (NameNode.blockStateChangeLog.isDebugEnabled()) { NameNode.blockStateChangeLog.debug("BLOCK* Removing stale replica " + "from location: " + r.getExpectedStorageLocation()); } {code} > BlockStateChange logs are too noisy > ----------------------------------- > > Key: HDFS-6860 > URL: https://issues.apache.org/jira/browse/HDFS-6860 > Project: Hadoop HDFS > Issue Type: Bug > Components: namenode > Affects Versions: 2.5.0 > Reporter: Arpit Agarwal > Assignee: chang li > Labels: newbie > Attachments: HDFS6860.patch > > > Block State Change logs are too noisy at the default INFO level and affect NN > performance on busy clusters. > Most of these state changes can be logged at debug level instead. -- This message was sent by Atlassian JIRA (v6.2#6252)