[ https://issues.apache.org/jira/browse/YARN-2884?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14710089#comment-14710089 ]
Jason Lowe commented on YARN-2884: ---------------------------------- Note that not all applications pick up configs from the nodes, and I don't see how relying on a HADOOP_CONF_DIR modification will address them. For example, our setup runs a MapReduce job as a self-contained application -- it does not reference the jars nor the configs on the cluster nodes. This makes rolling upgrades more reliable, otherwise a config change on the node could break old code in a job or new code in a job could break on an old node config. This happened in practice which is why our jobs no longer rely on confs from the nodes. HADOOP_CONF_DIR does _not_ show up on the classpath for such applications, otherwise they would be relying on server-side configs and lead to the rolling upgrade instabilities. Any ideas on how to address the self-contained application scenario? > Proxying all AM-RM communications > --------------------------------- > > Key: YARN-2884 > URL: https://issues.apache.org/jira/browse/YARN-2884 > Project: Hadoop YARN > Issue Type: Sub-task > Components: nodemanager, resourcemanager > Reporter: Carlo Curino > Assignee: Kishore Chaliparambil > Attachments: YARN-2884-V1.patch, YARN-2884-V2.patch, > YARN-2884-V3.patch, YARN-2884-V4.patch, YARN-2884-V5.patch, > YARN-2884-V6.patch, YARN-2884-V7.patch, YARN-2884-V8.patch, YARN-2884-V9.patch > > > We introduce the notion of an RMProxy, running on each node (or once per > rack). Upon start the AM is forced (via tokens and configuration) to direct > all its requests to a new services running on the NM that provide a proxy to > the central RM. > This give us a place to: > 1) perform distributed scheduling decisions > 2) throttling mis-behaving AMs > 3) mask the access to a federation of RMs -- This message was sent by Atlassian JIRA (v6.3.4#6332)