[BIGTOP-1274] yarn ResourceManager should be started before starting NodeManagers

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.7.0
• Fix Version/s: 0.8.0
• Component/s: deployment
• Assignee: Evans Ye
• Reporter: Evans Ye
• Created: 16/Apr/14 16:48
• Updated: 17/Apr/14 03:56
• Resolved: 17/Apr/14 03:56

Description

Yarn NodeManager failed to start up while using bigtop-puppet to deploy a hadoop cluster with yarn required:

debug: Service[hadoop-yarn-nodemanager](provider=redhat): Executing '/sbin/service hadoop-yarn-nodemanager start'
err: /Stage[main]/Hadoop_worker_node/Hadoop::Nodemanager[nodemanager]/Service[hadoop-yarn-nodemanager]/ensure: change from stopped to running failed: Could not start Service[hadoop-yarn-nodemanager]: Execution of '/sbin/service hadoop-yarn-nodemanager start' returned 1:  at /tmp/vagrant-puppet/modules-0/hadoop/manifests/init.pp:530

This is caused by starting up NodeMangers before ResourceManager gets ready.

The reproduce steps can completely leverage the vagrant-puppet snippets delivered in BIGTOP-1178:

1.) Add components you would like to have by modifying provision.sh

...
jdk_package_name,java-1.7.0-openjdk-devel.x86_64
components,hadoop,hbase,yarn,mapred-app
EOF
...

2.) Spin up the cluster

$ ./startup.sh --cluster