Define Service Nodes

This next part shows how to configure a xCAT Hierarchy and provision xCAT service nodes from an existing xCAT cluster.


The document assumes that the compute nodes that are part of your cluster have already been defined into the xCAT database and you have successfully provisioned the compute nodes using xCAT

The following table illustrates the cluster being used in this example:

Operating System rhels7.1
Architecture ppc64le
xCAT Management Node xcat01
Compute Nodes (group=rack1) r1n01 r1n02 r1n03 … r1n10
Compute Nodes (group=rack2) r2n01 r2n02 r2n03 … r2n10
  1. Modify site table attribute to include service group’s postscripts in compute node definition:

    chdef -t site hierarchicalattrs="postscripts"
  2. Select the compute nodes that will become service nodes

    The first node in each rack, r1n01 and r2n01, is selected to become the xCAT service nodes and manage the compute nodes in that rack

  3. Change the attributes for the compute node to make them part of the service group:

    chdef -t node -o r1n01,r2n01 -p groups=service
  4. When copycds was run against the ISO image, several osimages were created into the osimage table. The ones named *-service are provided to easily help provision xCAT service nodes.

    # lsdef -t osimage | grep rhels7.1
      rhels7.1-ppc64le-install-compute  (osimage)
      rhels7.1-ppc64le-install-service  (osimage)   <======
      rhels7.1-ppc64le-netboot-compute  (osimage)
  5. Add some common service node attributes to the service nodegroup:

    chdef -t group -o service setupnfs=1 \
                              setupdhcp=1 \
                              setuptftp=1 \
                              setupnameserver=1 \
    • Even if you do not want xCAT to configure any services, you must define the service nodes in the servicenode table with at least one attribute, set to 0, otherwise xCAT will not recognize the node as a service node
    • See the setup* attributes in the node definition man page for the list of available services: man node
    • For clusters with subnetted management networks, you might want to set setupipforward=1
    • For the setupconserver attribute, if conserver is used, set to 1, if goconserver is used, set to 2
  6. Add additional postscripts for Service Nodes (optional)

    By default, xCAT will execute the servicenode postscript when installed or diskless booted. This postscript will set up the necessary credentials and install the xCAT software on the Service Nodes. If you have additional postscripts that you want to execute on the service nodes, copy to /install/postscripts and run the following:

    chdef -t group -o service -p postscripts=<mypostscript>
  7. Assigning Compute Nodes to their Service Nodes

    The node attributes servicenode and xcatmaster, define which Service node will serve the particular compute node.

    • servicenode - defines which Service Node the Management Node should send commands to (e.g xdsh) and should be set to the hostname or IP address of the service node that the management node can contact it by.
    • xcatmaster - defines which Service Node the Compute Node should boot from and should be set to the hostname or IP address of the service node that the compute node can contact it by.

    You must set both servicenode and xcatmaster regardless of whether or not you are using service node pools. For most scenarios, the value will be identical.

    chdef -t group -o rack1 servicenode=r1n01 xcatmaster=r1n01
    chdef -t group -o rack2 servicenode=r2n01 xcatmaster=r2n01
  8. Set the conserver and monserver attributes

    Set which service node should run the conserver (console) and monserver (monitoring) daemon for the nodes in the group. The most typical setup is to have the service node also act as it’s conserver and monserver.

    chdef -t group -o rack1 conserver=r1n01 monserver=r1n01
    chdef -t group -o rack2 conserver=r2n01 monserver=r2n01
  9. Choose location of /install and /tftpboot directories (optional).

    The site table attributes installloc and sharedtftp control mounting of /install and /tftpboot directories from Management Node to Service Node.

    To mount /install and /tftpboot directories from Management node to each Service Node:

    chdef -t site clustersite sharedtftp=1
    chdef -t site clustersite installloc="/install"

    To make /install and /tftpboot directories local on each Service Node, set site table attributes and “sync” /install and /tftpoot directory contents from Management Node to Service Nodes:

    chdef -t site clustersite sharedtftp=0
    chdef -t site clustersite installloc=
    rsync -auv --exclude 'autoinst' /install r1n01:/
    rsync -auv --exclude 'autoinst' /install r2n01:/
    rsync -auv /tftpboot r1n01:/
    rsync -auv /tftpboot r2n01:/


If /install and /tftpboot directories local to each Service Node are used and mknb command is executed to generate a diskless network boot image with custom changes, it will not be automatically copied to the Service Node. Make sure to run the above rsync commands after executing the mknb. Verify /tftpboot/xcat directory on Service node contains genesis.kernel.<arch> and genesis.fs.<arch>.gz files.