Home Forums YARN How to query queue/cluster node list

This topic contains 1 reply, has 2 voices, and was last updated by  Jim Falgout 5 months, 2 weeks ago.

  • Creator
    Topic
  • #40804

    john lilley
    Member

    How can a YARN client query the list of available nodes? We are doing our own task scheduling outside of MapReduce, and in order to do that we want to get the list of available nodes and compare them to the list of nodes on which data file blocks are located. I can see YarnClient.getClusterMetrics() but that only returns the count.

    We should really be querying the list of nodes for the default queue, but I’d be happy to start simple.

    I see that there is a getActiveTrackerNames() call available to MapReduce, but that’s not part of the YARN API.

Viewing 1 replies (of 1 total)

You must be logged in to reply to this topic.

  • Author
    Replies
  • #42477

    Jim Falgout
    Member

    There is a method on YARNClient called getNodeReports(). You pass it in a list of node states and it returns a list of NodeReport instances. On NodeReport you can invoke getCapability(). It returns a Resource object from which you can obtain the number of vcores and amount of memory for the node.

    I’m looking at the YARNClient for Hadoop 2.2.0 (the HortonWorks 2.0.6 GA release).

    Collapse
Viewing 1 replies (of 1 total)