YARN Forum

Yarn JMX QueueMetrics "q=root" vs "q=root,q1=default"

  • #58533
    Hari Sekhon
    Participant

    In the Resource Manager’s JMX can anyone explain to me the difference between Hadoop:service=ResourceManager,name=QueueMetrics “q0=root,q1=default” and “q0=root”?

    I’ve noticed most of the metrics are the same, but q0=root always shows active users 0, while “q0=root,q1=default” shows a positive integer as expected for active users when queries are running in Hive/Tez.

to create new topics or reply. | New User Registration

  • Author
    Replies
  • #58589
    Hari Sekhon
    Participant

    Ok I went back over the docs, it’s hierarchical of course, but I’m surprised that active users are only counted in leaf queues when I would have expected the top level queue to contain the aggregate summary information for all child queues including the number of active users.

    Can anyone explain this behaviour? Is this something that needs to be fixed so that ‘root’ shows the number of active users in all sub-queues?

    #68728
    Aneesh Karunakaran
    Participant

    in your case, q0 is the ‘root queue’ and q1 is the leaf queue. Jobs are allowed to run only on leaf nodes.

    You can read more details about capacity scheduler here – http://docs.hortonworks.com/HDPDocuments/HDP2/HDP-2.1.5/bk_system-admin-guide/content/setting_up_queues.html

    #68741
    Hari Sekhon
    Participant

    I’ve read the docs on it – more than once (you forget after you don’t tune your queues for a few months…)

    I still think the root level should provide aggregate summary statistics for the leaf levels.

You must be to reply to this topic. | Create Account

Support from the Experts

A HDP Support Subscription connects you experts with deep experience running Apache Hadoop in production, at-scale on the most demanding workloads.

Enterprise Support »

Become HDP Certified

Real world training designed by the core architects of Hadoop. Scenario-based training courses are available in-classroom or online from anywhere in the world

Training »

Hortonworks Data Platform
The Hortonworks Data Platform is a 100% open source distribution of Apache Hadoop that is truly enterprise grade having been built, tested and hardened with enterprise rigor.
Get started with Sandbox
Hortonworks Sandbox is a self-contained virtual machine with Apache Hadoop pre-configured alongside a set of hands-on, step-by-step Hadoop tutorials.
Modern Data Architecture
Tackle the challenges of big data. Hadoop integrates with existing EDW, RDBMS and MPP systems to deliver lower cost, higher capacity infrastructure.