Wiki

• HA
• Fully automated install/Definition of what we plan to deliver for R1
• Detailed description of system end-state produced
• Debugging/fixing puppet modules for proxy environments.
• Tempest test case list (deciding on appropriate ones for R1)
• Tempest puppet module for configuring/installing tempest.
• Khaleesi integration/testing.
• Current state installer testing + verfication (follow steps on wiki, provide feedback on what works what doesn't)
• Jenkins integration with Khaleesi

Sequence of integration points with CI ("touch points where your installer is expected to be kicked by Jenkins - along with what you expect to deliver back in case of success or failure").

• pointer to script to kick off installer
• …

• Tempest is now installed/running on a baremetal node in the Intel testbed. Working on a puppet module to install/configure tempest automatically.
• Was able to use Foreman to launch KVM VM's through libvirt. This confirms we can use this for a demo virtualized deployment for the installer if we want. Also, we should move the Tempest node to be virtualized.
• Pod 2 is almost ready in the Intel lab for use.
• Dan Radez is still looking into HA options for control node in OpenStack.
• Tim Rozet working on the temptest puppet module, then Khaleesi integration.

• Foreman will now install/rebuild OpenStack to Intel lab (Topology is 1 Compute Node and 1 Network+Control Node)
• There are 5 servers currently in Intel lab. 2 have BMC issues, 2 are being used as mentioned above, and the final one will be used to test HA.
• Changes to appropriate modules have been made in the Intel lab (official commits pending) to install OpenDaylight and configure it as a neutron plugin. These are optional parameters set in Foreman to decide to use OpenDaylight or other ML2 driver.
• There are two issues blocking more testing with ODL as the neutron plugin right now:
  • When OpenDaylight comes up, it can take up to 60 seconds for all of the features to install and become active. We need to wait for all the features to be up before we enable openvswitch and point to ovsdb as the manager. This ensures we do not miss the time period where ovsdb will detect the new node and configure br-int on it.
  • There is another issue where sometimes Neutron sends API calls to ODL, where the result is Neutron receives a 200 OK from ODL, while ODL never pushes any flows to openvswitch. This needs to be looked at so we can reliably know what we are configuring and get rid of false positives.
• Tim Rozet is working on a fix for issue 1. While issue 2 is TBD after 1.
• Dan Radez is working on getting HA integration done.