wiki:GENIRacksHome/RacksChecklist

Version 32 (modified by asydney@bbn.com, 11 years ago) (diff)

--

High level steps for GENI rack installation

Each rack site develops their own installation and test plan, which roughly follows these steps. Please coordinate with the GPO and the rack development teams when performing the installation.

Common rack name chosen

  • The GPO selects a common rack name for GENI tools to use for a given rack using a pattern and then agrees on the name locally
  • The GPO proposes the common rack name to the rack team, the GMOC, and the site contact, to ensure that there are no major issues with the proposed name

Work with rack teams to identify site specific information and obtain shipping details

  • Provide shipping information
  • Provide any power requirements
    • Provide voltage and connector requirements
    • If you need a UPS for your rack, please coordinate with the GPO and the rack team
  • Communicate control plane and data plane connectivity requirements
    • Work with the GPO to identify your upstream data plane connection
    • Coordinate with rack team on procuring the needed data plane optics
    • Coordinate with the rack team to understand expectations for control plane connectivity
  • Provide rack team with a list of site admins
    • Coordinate with rack team to set up operations contacts
    • Provide usernames for those who should have administrative access to the rack hosts
  • Coordinate with rack teams on network configuration
    • Work with rack teams to allocate an block of IP space for the rack
    • Work with rack teams to prepare DNS for the rack

Install hardware

  • Put the rack in place and set up any needed power connections
  • Power up the devices in the rack

Establish upstream control plane connectivity

  • Ensure that the management switch's control plane uplink is physically connected to a publicly routable network
    • Ensure that the machine on which FlowVisor is running is reachable over a publicly routed network
    • Ensure that all machines on which aggregate manager software is running are reachable over a publicly routed network
  • Ensure that any other devices that should be publicly reachable (i.e. VPN appliances) are physically connected to a publicly routable network

Coordinate configuration and testing of aggregate manager software

  • Coordinate with the GPO for FOAM configuration and testing
  • Coordinate with the rack teams for compute aggregate manager configuration and testing

Coordinate data plane connection and testing

  • If no GMOC master ticket has been created for the rack installation, then the GPO will ensure a ticket gets created before beginning data plane testing
  • Coordinate with the GPO to set up upstream mesoscale connections
    • Before actually establishing a connection, open a GMOC ticket to track data plane connection and testing
    • Set up VLAN for shared mesoscale VLAN connection
    • If applicable, set up exclusive mesoscale VLAN connections
    • On paths with OF VLANs, disable MAC address learning on non-OF switches
    • Disable STP variants on mesoscale VLANs on all devices (or if that is not possible, ensure that BPDUs don't get sent OF switches)
    • Coordinate data plane connectivity testing with the GPO and the GMOC using a GMOC ticket
  • Work with the rack teams and the GPO to set up a pool of other general purpose experimental VLANs
  • Provision a port for downstream campus resources, and add a test point to that port for data plane connectivity testing
  • Coordinate any other data plane connections with the GPO and other rack teams
    • Before actually establishing an additional remote connection, open a GMOC tracking ticket
    • Coordinate with the GMOC on the date for connection and testing
    • Coordinate data plane connectivity testing with rack teams, the GPO, and the GMOC using a GMOC ticket

Set up monitoring

  • Configure active Aggregate Manager monitoring
    • Coordinate with the GMOC for setting your site up to submit active monitoring data to the GMOC database. Live monitoring data is available on the GMOC pages.
    • Coordinate with the rack team to configure FOAM monitoring (instructions)
    • Coordinate with the rack team to configure compute aggregate manager monitoring
      • ExoGENI Instructions coming soon
      • InstaGENI Instructions coming soon
  • Coordinate with the GPO to set up passive monitoring (simple API checks)
  • Coordinate with the GPO to get data plane monitoring set up
  • Coordinate with the GPO to get control node monitoring set up
  • Set up local site monitoring, if desired

Perform confirmation testing pre-work

  • Review the GENI Aggregate Provider Agreement. Indicate via email to the GPO your intent to abide by the agreement.
  • Complete an Aggregate Information GENI wiki page (see template) for your site, which will be automatically linked here. This lets experimenters know how to request and use your OpenFlow connections in GENI.
  • Coordinate with the rack team to complete an operator-facing page

Coordinate confirmation testing

Plastic Slices (GPO OpenFlow Experiment)

Coordinate with GPO to add your site to the long-running Plastic Slices experiment. This experiment runs most of the time and involves most of the deployed OpenFlow sites, so it is a good way to exercise your site in GENI and to help out with interoperability goals at the same time.

Site performs their own testing

Sites do their own testing as needed for their particular deployments. The site owner decides when they are ready to open the site to experimenters. The site owner must confirm that they are ready to support experimenters before a site can become a production GENI site (a.k.a. moving to production). A site may have multiple GENI resources (for example a rack and several OpenFlow switches) that move to production at different times.

Perform final administrative steps before moving to production

  • Work with GMOC to complete monitoring, mailing list, and emergency contact hand-off procedures for your site. When the procedures are complete, the GMOC begins 24x7 monitoring, tracking and dispatching questions and events that involve your site.
  • Subscribe an operational contact for your rack to the response-team@geni.net mailing list
  • Coordinate with the GPO to connect and integrate downstream campus resources with the GENI rack. In most cases, this will be a campus SDN network supporting multiple hosts.

Move to production

  • GPO coordinates with GENI clearinghouse maintainers to add new rack aggregate to the GENI clearinghouse.
  • GPO coordinates with Utah team to add new rack aggregate to their clearinghouse.
  • GPO coordinates with omni developers to update the aggregate nicknames listed in the reference omni config and scripts.
  • GPO marks aggregate as a production resource for GPO monitoring.
  • Rack is officially announced as production.