Taskcluster/Meetings/2014-01-14

Status Updates

Job Queue (owner: dminor/ted)

• Have prototype implementation using Python 3.3
  • https://github.com/dminor/taskcluster-jobqueue
  • implemented REST interfaces discussed last week for Q1, but some of the JSON passed in is not fully defined yet
  • Current implementation is in memory, changing to use a database (sqlite for Q1)
• Writing correctness tests
  • will be useful even if we end up throwing this prototype away
• Stress testing
  • Notional goal is to simulate 250000 jobs for 10 minutes
  • See how far we can get with Python

Provisioner (owner: jonas)

• jonas rewriting in node

Round Table

• [jgriffin] Bare-bones project page at Auto-tools/Projects/TaskCluster
• [jgriffin] Additional co-owner needed for the Worker component?
  • ted and jonas to help out here
• [dminor] Maybe an owner for 'systems integration' / whole system testing?
  • ted to own this
• [dminor] Can we agree upon a limited set of tags/parameters/? for matching jobs to workers for Q1?
  • Is maybe just os, cpu sufficient for now?
  • Something simple, maybe just AMI initially (or URL for a docker image), until we can explore the problem space a bit more
• [jmaher] - are all parameters required for the json object coming into the job queue?
  • No, many of them are optional
• [jgriffin] - For prototyping, we can use Pulse for any AMQP needs; we may switch out later to cloudamqp or other service

Action Items

• [dustin] to discuss deployment strategies (heroku vs internal?) with lightsofapollo and other module owners
• [jgriffin] to find a different meeting time for next week