Zhijie Shen created YARN-2520: --------------------------------- Summary: Scalable and High Available Timeline Server Key: YARN-2520 URL: https://issues.apache.org/jira/browse/YARN-2520 Project: Hadoop YARN Issue Type: Sub-task Components: timelineserver Affects Versions: 2.5.0, 3.0.0 Reporter: Zhijie Shen Assignee: Zhijie Shen
YARN-2032 will provide a scalable and reliable timeline store based on HBase. However a single instance of the timeline server is not scalable enough to handle a large volume of user requests, being the single bottleneck. As the timeline server is the stateless machine, it's not difficult to start multiple timeline server instances and write into the same HBase timeline store. We can make use of Zookeeper to register all the timeline servers, as HA RMs do, and client can randomly pick one server to publish the timeline entities for load balancing. Moreover, since multiple timeline servers are started together, they are actually back up each other, solving the high availability problem as well. -- This message was sent by Atlassian JIRA (v6.3.4#6332)