[ https://issues.apache.org/jira/browse/YARN-422?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13663910#comment-13663910 ]
Zhijie Shen commented on YARN-422: ---------------------------------- bq. There is no need to start NMCommunicator everytime for stop/getStatus? We can cache the one's from startContainer() ? In the previous discussion, you suggested: {quote} Among the "3 choices" you listed in stopContainer(), we should adopt (1) for now as that is what MR AM's ContainerLauncher. Primarily because of the one-connection-per-container issue. This again can and should be changed post YARN-613. {quote} In ContainerLauncherImpl, the proxy closes every time when a container is started or stopped (getStatus is not used there). bq. All the call-backs are blocking. IMHO, the callbacks can be either blocking or non-blocking. It depends on the implementation of the callback interface. To be non-blocking, the implementation can similarly schedule an event when the callbacks are invoked, and the event will be dispatched to a thread for processing. > Add NM client library > --------------------- > > Key: YARN-422 > URL: https://issues.apache.org/jira/browse/YARN-422 > Project: Hadoop YARN > Issue Type: Sub-task > Reporter: Bikas Saha > Assignee: Zhijie Shen > Attachments: AMNMClient_Defination.txt, > AMNMClient_Definition_Updated_With_Tests.txt, proposal_v1.pdf, > YARN-422.1.patch, YARN-422.2.patch, YARN-422.3.patch, YARN-422.4.patch, > YARN-422.5.patch, YARN-422.6.patch > > > Create a simple wrapper over the ContainerManager protocol to provide hide > the details of the protocol implementation. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira