[
https://issues.apache.org/jira/browse/FLINK-2499?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14680253#comment-14680253
]
Stephan Ewen commented on FLINK-2499:
-------------------------------------
Actually, it is intended that you are able to start multiple taskmanagers on
each machine. The intended use is to put the name of a machine (or "localhost")
multiple times in to the "slaves" file.
It makes testing / debugging scenarios much simpler, as you can easily start a
multi-TaskManager (JVM) cluster on your single machine.
> start-cluster.sh can start multiple TaskManager on the same node
> ----------------------------------------------------------------
>
> Key: FLINK-2499
> URL: https://issues.apache.org/jira/browse/FLINK-2499
> Project: Flink
> Issue Type: Bug
> Affects Versions: 0.8.1
> Reporter: Chen He
>
> 11562 JobHistoryServer
> 3251 Main
> 10596 Jps
> 17934 RunJar
> 6879 Main
> 8837 Main
> 19215 RunJar
> 28902 DataNode
> 6627 TaskManager
> 642 NodeManager
> 10408 RunJar
> 10210 TaskManager
> 5067 TaskManager
> 357 ApplicationHistoryServer
> 3540 RunJar
> 28501 ResourceManager
> 28572 SecondaryNameNode
> 17630 QuorumPeerMain
> 9069 TaskManager
> If we keep execute the start-cluster.sh, it may generate infinite
> TaskManagers in a single system.
> And the "nohup" command in the start-cluster.sh can generate nohup.out file
> that disturb any other nohup processes in the system.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
