trozet

1.  Create VNFs in Tacker, attach them to the management VNF network on br-int.
2.  Create a chain of those VNFs via Tacker
3.  ODL Rest call made to add unique VTEP ports to each OVS br-int that the VNF attaches to.  VTEP uses a unique port allocated to identify a tenant. Does VNID matter here?
4.  Tacker initiates some type of configuration to the VNF to create matching VTEP in (3)
5.  Tacker somehow initiates an ip route call on each host, adding a loopback route for VNF IP to go back into br-int.
6.  Tacker then initiates REST calls to ODL SFC to create the chain.  Uses regular management VNF IP with unique UDP port as its data-plane-locator.

End result ends up looking like this:
Private Neutron Network= 10.0.0.x
Public External Network= 192.0.0.x

trozet

workaround_puppet_version_lock: false
opm_branch: master
installer:
  name: puppet
  short_name: pupt
  network:
    auto_assign_floating_ip: false
    variant:
      short_name: m2vx
    plugin:

trozet

---
classes:
  quickstack::openstack_common: 
  quickstack::pacemaker::ceilometer:
    ceilometer_metering_secret: octopus
  quickstack::pacemaker::cinder:
    backend_nfs: false
    backend_nfs_name: nfs_backend
    backend_rbd: true
    nfs_shares:

trozet

At the end of this everything works, just missing neutron.

1.  First I made changes to quickstack, clone my latest opendaylight branch por favor
2.  Disable hiera in foreman (I guess its on by default?).  Hiera will auto replace variable names not defined in smart class parameters for Foreman, and it automatically reads astapor defaults which ruins everything.  To disable it:

data_binding_terminus = none

^put that in your puppet.conf file under master on Foreman.

3.  restart system services: foreman, foreman-proxy, httpd, puppet (also needed to pick up the library changes in quickstack)

trozet

[global]
fsid = 904c8491-5c16-4dae-9cc3-6ce633a7f4cc
mon_initial_members = oscontroller1, oscompute11, tempest1
mon_host = 10.4.9.2, 10.4.9.4, 10.4.8.51
auth_cluster_required = cephx
auth_service_required = cephx
auth_client_required = cephx
filestore_xattr_use_omap = true

public_network = 10.4.8.0/21

trozet

[client.images]
      key = AQAfHBdUKLnUFxAAtO7WPKQZ8QfEoGqH0CLd7A==
      caps mon = "allow r"
      caps osd = "allow class-read object_prefix rbd_children, allow rwx pool=images"

trozet

[client.volumes]
      key = AQAfHBdUsFPTHhAAfqVqPq31FFCvyyO7oaOQXw==
      caps mon = "allow r"
      caps osd = "allow class-read object_prefix rbd_children, allow rwx pool=volumes"

trozet

#!/bin/bash
#
# This script is run from a node that doubles as both a ceph installer
# and storage node.  It installs ceph mons on the $monnames given
# existing ceph config files and sets up a osd volume.  Make sure
# ceph.conf includes "osd_pool_default_size = 1".
#
# Typical order of operations:
#  * make sure the 3 ha controller nodes/mons have not run puppet yet.
#     (i.e., they are not yet configured as ha controllers,

trozet

---
classes:
  quickstack::openstack_common: 
  quickstack::pacemaker::ceilometer:
    ceilometer_metering_secret: octopus
  quickstack::pacemaker::cinder:
    backend_nfs: false
    backend_nfs_name: nfs_backend
    backend_rbd: true
    nfs_shares:

trozet

class trystack::compute_ha {
  if ($odl_flag != '') and str2bool($odl_flag) { 
     $ml2_mech_drivers = ['opendaylight']
     $this_agent = 'opendaylight'
  }
  else {
    $ml2_mech_drivers = ['openvswitch','l2population'] 
    $this_agent = 'ovs'
  }