David, Mike, Kenton, Craig

  • NDSC workshop: April 4-5-6
    • Tutorial April 4
    • Demo April 5
  • Walkthough of current NDSLabs environment including GUI and CLI
    • Evident problems with Clowder extractors, just came up today. Everything was working yesterday. Need to investigate.
  • Discussion of Terratoolserve:
    • Currently, we are running the tool server as a system daemon. 
    • Discussed possibility of using Kubernetes, but this is problematic. Since the toolserve spawns it's own containers, there's no guarantee they'll be available if the service is moved.
  • Discussion of "resource management" in the API Server.  Quotas are not currently enforced – no resource management requirements have been written.
  • Discussion of apictl use of APIs.  apictl is not using go api bindings (ala swagger client), but could
  • Discussion of iRods integration with Dataverse
    • The Odum case is complicated (requires running multiple iRods instances). For the demo, it would be easier to run a single instance with the Odum rules.
    • Kenton is OK with this, but Craig will confirm with Jon@Odum
  • NDSC:
    • Demo has been moved to Weds – Jon@Odum and David L asked to attend
  • Task: Need to make videos in case things go wrong 
  • Task: Installer process for GUI, Apiserver, Kubernetes
  • What David's working on:
    • Building our own Kuberenetes binaries + containers
    • This is in part to support the Nebula Cinder integration but also to support a realistic development environment (emulate full cluster networking including loadbalancing)
  • Discussion of Workshop tutorial (April 4th):
    • Every participant gets a CoreOS VM
      • VM setup:
        • Build on VM, document the process, checkpoint the image
        • Create a single jumphost with username/passwords for participants
        • Each participant is given a separate VM instance IP (internal), which they will access via shared key
        • Need to define system specs and contact Nebula team to ensure we have sufficient resources. NDSC is running around the same time as the UND hackathon, which also requires Nebula resources.
        • Use NDS-hackathon project
          • 35 instances, each with 2-4 CPUs, 8GB RAM, 80GB disk
      • What is the image? CoreOS version TBD, Docker 1.?
      • Docker run -it ndslabs/ndsdev bash
        • Pulls image
        • kube-up
        • ndslabs-up (apiserver, gui, load default specs)
        • login/ndslabctl/kubectl/docker
      • Show how to add a service and run it
        • Write docker file, build image, push the image (need a dockerhub account)
        • Write NDSLAbs service spec, load and test that spec
        • Run the service via apictl or UI, start stop, interact with the endpoints (upload data/etc)
      • Assumption
        • hostPath volumes (no one will have OpenStack credentials)
        • Each participant will have a laptop with SSH capability and browser
        • Will continue to use NodePort
        • VMs: Access will be terminated when – end of workshop (end of week)

  • We will still try to integrate Prometheus – but this is not a must-have
  • Discussion of what it means to have a "complete local development environment" – requirements unclear:
    • What are the requirements:
      • Docker v1.9; adequate RAM and HDD (8GB, as much HD as images you want to run)
      • System requirements: We don't understand the requirements, so we need to identify them
    • Bootup scripts pushed into container or tarfile
  • Follow-up on planning
    • Naildown WBS based on previous NDS meetings (on wiki)
    • After workshop: go through item by item and flesh it out
    • This is for 2 years: NDS Labs, NDS Share, overall mission
    • Has been presented to TAC already
  • No labels