[ https://issues.apache.org/jira/browse/YARN-658?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13783711#comment-13783711 ]
Vinod Kumar Vavilapalli commented on YARN-658: ---------------------------------------------- @davidyan, sorry missed your logs. The logs show the following for the AM Container {code} 13/07/30 11:32:14 DEBUG launcher.ContainerLaunch: Sending signal to pid 30435 as user david for container container_1375209060900_0002_01_000001 13/07/30 11:32:14 DEBUG launcher.ContainerLaunch: Sent signal to pid 30435 as user david for container container_1375209060900_0002_01_000001, result=failed {code} It seems like NM sent a SIGTERM, but I don't see the following SIGKILL getting sent. Is your custom AM somehow trapping SIGTERM? What is yarn.nodemanager.sleep-delay-before-sigkill.ms set to on your cluster? > Command to kill a YARN application does not work with newer Ubuntu versions > --------------------------------------------------------------------------- > > Key: YARN-658 > URL: https://issues.apache.org/jira/browse/YARN-658 > Project: Hadoop YARN > Issue Type: Bug > Affects Versions: 2.0.3-alpha, 2.0.4-alpha > Reporter: David Yan > Attachments: AppMaster.stderr, > yarn-david-nodemanager-david-ubuntu.out, > yarn-david-resourcemanager-david-ubuntu.out > > > After issuing a KillApplicationRequest, the application keeps running on the > system even though the state is changed to KILLED. It happens on both Ubuntu > 12.10 and 13.04, but works fine on Ubuntu 12.04. -- This message was sent by Atlassian JIRA (v6.1#6144)