Notes from 12/15/2016 discussion of NDS-686.  When redeploying beta, we need to account for the following:

  • Rebuild Gluster FS:
    • NDS-693 - Getting issue details... STATUS
    • We'll need to redeploy Gluster with the new sharding config
  • Master crashing:
    • NDS-660 - Getting issue details... STATUS
    • In the end, this is a bug in the NDSLabs API Server. Needs to be fixed, but doesn't directly impact redeploy
    • Should we resize master? No. No one is using big masters.
  • Upgrading Kubernetes:
    • NDS-413 - Getting issue details... STATUS ,  NDS-708 - Getting issue details... STATUS
    • This is probably our biggest risk/issue at this point for re-deploy.  We're not sure whether we can do it, but we'll try.
  • Disk space issues:
    • NDS-627 - Getting issue details... STATUS
    • Will likely need to create another volume or change /var/lib/docker to mount /var/ to cover all cases.
    • David's recommendation: 1 volume per node with bind mounts.
  • Import/export accounts:
    • NDS-690 - Getting issue details... STATUS
    • We will need to migrate at least the account information
    • Currently, stacks and data will not be migrated. Stacks require catalog migration too.
  • Node sizing:
    • What is the correct size for the nodes? Do we need to resize LMA, master?
    • LMA:
      • Should map correct data volumes
      • 4-8 cores; 32-64GB RAM
  • Deploy new test cluster ASAP
  • No labels