Auto-tools/Projects/Futurama/2012-11-13

Agreed that it is most important to identify tasks with immediate priority that we can get started on in the near future. We can continue the prioritization of other tasks once we get a handle on these.

These tasks have been identified as things we can do more-or-less right away because they are (a) already done, already underway, or relatively easy, and/or (b) required for other tasks (infrastructure, requirements, information).

• Re-building the gofaster dashboards and using that to drive down build/test times
  • This is important only partly because it will give us information up front (such as longest-running tests). Much more importantly, if we maintain this, we will have a record of how build/test times, capacity, and overall throughput are doing **while we implement longer-term fixes**. Basically, we need this to see if other efforts are paying off.
  • Initial task is to determine what we want to measure and what gofaster already measures (i.e. requirements).
  • Second task is to resurrect what we have and fix/add to it.
• What are the top 10% longest running tests? Can the they be put in their own run? 
  • Should be answered by above.
• Review longer term use cases for TBPL and visualizing the amount of data that will soon be thrown at it 
  • As with gofaster, we need to create concrete requirements for a TBPL replacement.
• Bugzilla dashboards
  • Should be launched soon.

• Running tests in VMs.
  • This is being actively pursued by releng; their goal is to have hooks in place by EOQ to allow us to start rolling out VMs early next year.
  • We will need to identify
    • what tests require real hardware (e.g. WebGL tests)
    • how we will differentiate them from other tests (e.g. separate test suite) and tell buildbot where to run them
• Adding new ix machines for windows/linux testing
  • Releng is already working on this.
• Push scheduling (buildbot configs) into tree
  • Releng is already working on this (bhearsum specifically). We could see if we can help out.

• Determine if and how repos other than hg.m.o (e.g. github) can be supported by AutoLand.