Deployment: Current state, plans for next 6 months, issues raised.

• sites that are up except:
  • umich - 4 sites up, 8 not. need to coordinate with local person, waiting now for that. they are having switching problems, dying hardware, (mis-configured?).
  • need to request an IT contact
  • had to buy a new KVM - that turned out not to be the problem, but a mis-configured switch there. extra KVM is at Vandy, so we have a spare

• need a technical contact/site user level agreement for what is required.

• UFL and UMich are only places without an IT contact, and this may be what we have at these places.

• deploying latest depot code on all depots as we speak.
• Oak Ridge - not brought up yet.
  • productive meeting end of March
  • we will be put outside ORNL's perimeter
  • same area that TeraGrid hardware is
  • so we could be part of TeraGrid...
  • Dan has some ideas for tests to run across the TeraGrid network
  • and John will integrate with his activities
  • communication between David Giles (at ORNL) and Bobby to get those boxes online. John is ultimately responsible for anything that happens but Bobby and David have a plan that he accepts.
  • David will have SUDO persmissions account on the depot hardware, Bobby will start this as soon as Michigan is done

• New sites?
  • First one below is solid, second has been offered, the rest...???
  • UTK - Gerald R.
  • LoC - if they want...
  • AmericaView sites? Wisconsin? Alaska?
  • Meeting of the executive committee?
  • Further depot deployments should be driven by applications...
  • OSG?
  • SuraGrid

• deployment/upgrade experience from recent upgrades, plans for future
  • management node, with an implementation of pexec for deploying software in parallel.
  • run p commands to push out software updates - os update, depot updates.
  • operating system is on a USB key on depots, which means that the OS is different than the data drives, which makes M&O easy.
  • need to write up a wiki page - document how things are configured - so that there can be uniformity for others who might want to opt in. (???) describe main issue - that os drive should be separate from data drives, other philosophies in use, etc.
  • procedure for upgrading depot
    • shutdown the depot code
    • copy over the new code
    • start back up
    • can all be done with only a short time, and no loss of data.

• monitoring, user info,...
  • nagios will come back online once umich site back online
  • nagios is not required to run on each of the boxes, but can collect info from many nodes.
  • nagios is designed to work across the wide area.
  • helps catch problems early
  • visualization - use part of what we had from SC07 on main reddnet page to allow users to get status of a particular site, see location of sites, show data movement.
  • "Is the depot up and responding" health test ala OSG?
  • want some sort of weathermap or OSG style site status
  • if site is failing, it is taken "offline"