Home Forums HDP on Linux – Installation Assignment of DataNodes

Tagged: 

This topic contains 3 replies, has 2 voices, and was last updated by  Jeff Sposetti 1 year, 11 months ago.

  • Creator
    Topic
  • #11768

    Laurentiu
    Member

    Hi am trying to install HDP2 on a six node cluster: 1 “master” node running master services (Namenode, RM, Zookeeper, HB Master), 4 worker nodes (data node, HB region server), 1 “edge” node running HIVE, HMC, Ganglia, Nagios and Secondary Name Node. For some an un known reason I end up with a data node running on my “edge” node ? Is there a way to prevent that from occurring (the edge node does not have the necessary disk space to be a data node).

Viewing 3 replies - 1 through 3 (of 3 total)

You must be logged in to reply to this topic.

  • Author
    Replies
  • #11922

    Jeff Sposetti
    Moderator

    Correct. This is a current limitation in HMC, regardless of 1.x or 2.x. We will be adding the feature to select which hosts to run DataNode and TaskTracker in an upcoming release. Since it’s an enhancement, we will not be backporting it.

    In the meantime, you have the option of performing a manual install, where you have the most flexibility to choose which hosts run which services. This link is for performing Manual Install using the HDP 1.1 packages.

    http://docs.hortonworks.com/CURRENT/index.htm#Deploying_Hortonworks_Data_Platform/Installing_Manually/rpminstall/rpm_getting_ready/rpm_getting_ready.htm

    Collapse
    #11907

    Laurentiu
    Member

    Jeff,
    This seems to be an issue even with version 1.x of HMC (I went back to 1.x since I could not get the HMC 2.0 install to work reliably), will the fix be back ported into that version ?

    Collapse
    #11841

    Jeff Sposetti
    Moderator

    Hi,

    When installing HDP 2.0 Alpha using HMC, currently all nodes — except the node running NameNode — will also run as DataNodes. That’s why you are getting the DataNode component running on your “edge” server. In a future release of HMC, we will be making this more configurable so you can designate the node(s) to run as DataNode(s) and avoid this situation.

    Thank you for the feedback.

    Jeff

    Collapse
Viewing 3 replies - 1 through 3 (of 3 total)