2018-01-11 Engineering Meeting


#1

Date/Time: 2018-01-11 @ 14:00 UTC / 09:00 EST / 22:00 HKT
Meeting link: https://zoom.us/j/799691635

Attendees

@jackzampolin
@jude
@larry
@ryan
@yukan.id
@guylepage3
@aaron

Agenda

Pick recurring timeslot for engineering meetings

  • Proposed range: 9am to 11am EST start time Tuesday, Wednesday or Thursday

Desired outcome: A timeslot & general meeting policy

Azure rolling outage postmortem

  • Everyone owns their own infra, but there is sometimes tragedy of commons with shared
  • Discuss evolution of Blockstack hosted infrastructure
    • What systems do we need to host going forward?
      • Currently running:
        • utxo.blockstack.org
        • bitcoin.blockstack.org
        • core.blockstack.org
        • hub.blockstack.org
        • node.blockstack.org
        • browser.blockstack.org
        • explorer.blockstack.org
        • +proofs.blockstack.org
    • How long do we see ourselves hosting them?
  • Alerting system upgrades/improvements
  • Consolidation

New UTXO provider

Web-based Blockstack Browser

  • Find way to implement protocol handlers for each browser

Drivers for Gaia Hub

STACKS blockchain R&D meetings

  • When’s a good time to schedule this?
  • What are the main questions we have?
  • How do we ask questions? How do we answer them?
  • Input from marketing and bizdev
    • Timeline?
    • Previews? Beta? Testnet?
    • What do we need by the Summit?

Issue Tagging

Projects with good tagging:

  1. Kubernetes - 4,700 open
  2. Their system is to tag issues with area/* and kind/* to allow sorting and then priority/* once they have been prioritized
  3. This gist gives a few examples
  4. The category/instance tagging is used in a number of projects similarly to how Kubernetes uses it. (see Hashicorp, Docker)
  5. Lisk - 100 open
  6. No use of category/.
  7. Each issue tagged with “kind” (standards, bug, test, etc…) and “difficulty” (easy, medium, hard)
  8. They also have a concept of “parent” and “child” issues. Child issues link back to parent and are marked as such
  9. React - 293 open
  10. They have Component: , Type: , Status: , and Difficulty:
  11. Most issues tagged with at least component and type

Meeting Notes

Pick recurring timeslot for engineering meetings

Decision: Thursday @ 14:00 UTC / 09:00 EST / 22:00 HKT

Azure rolling outage postmortem

  • We plan to hire someone to do sysops
  • Jack is going write some system d unit files
  • We need repeatable deployments
  • We need 3 shifts of people on call to handle issues
  • Decision: We need to hire someone to do the creation of orchestration and then have people on duty that can solve issues if they come up

UTXO provider

  • Larry: ask exchanges what they do

Web-based blockstack browser

  • Aaron: we could use a cookie to remember that users had signed up on the hosted browser
  • The right time to address this is when address on-boarding process

Drivers for Gaia Hub

  • Jude: port Blockstack drivers and take notes on the experience

STACKS blockchain R&D meetings

  • Jude: split between product meetings and engineering meetings

Issue tagging

  • Jack has done some research on issue tagging which he will share on a forum post for discussion

#2

Thanks for the update.

Using container platform like Kubernetes (recommend) for your servers will help greatly with these kind of rolling updates. You can leverage the platform for your own blue-green deployments with zero downtime. Container platforms bring in lot other benefits of infra automation and scaling.

Azure meltdown patches impacted many organizations, Azure handled this bad.

Thanks