Hey, Telles:
You can simply decrease the NM expiry-interval. The default value is 5 mins. You can config nm.liveness-monitor.expiry-interval-ms (in millisecond)in yarn-site.xml Thanks Xuan Gong From: Telles Nobrega <tellesnobr...@gmail.com<mailto:tellesnobr...@gmail.com>> Reply-To: "user@hadoop.apache.org<mailto:user@hadoop.apache.org>" <user@hadoop.apache.org<mailto:user@hadoop.apache.org>> Date: Friday, February 6, 2015 at 10:03 AM To: "user@hadoop.apache.org<mailto:user@hadoop.apache.org>" <user@hadoop.apache.org<mailto:user@hadoop.apache.org>> Subject: Hadoop Node Failure Detection Hi, hadoop takes a while to "notice" that a node is gone right? I'm using hadoop 2.6.0 and I would like to change this detection time for experience porpuse, which configuration should I change and where? Thanks