Sprint 21 retrospective
What did we do right:
- Architectural discussions
What can we do better:
- There was no real sprint goal established
- We're not planning anymore:
- we're writing tickets during the planning meeting
- The backlog is "inadequate"
- Nebula disruption: not sure if it's fixed
- Priorities are not clear and roadmap is changing
What should we keep doing/need to do
- We need to incorporate more planning and flesh out the backlog with full, prioritized tickets before the Thursday meeting.
- We need a roadmap for 3-6 months and a process for changing it
Sprint 22 commit
Goal: Define requirements and design architecture for: modular auth, developer/personal instance, and production clusters (with minimal cluster configuration that supports scaling and fault tolerance). Outputs of this sprint are requirements and design documents and enough fully-specified JIRA tickets with estimates for next Sprint.
- Strategy for modular auth – Cyverse has their own custom system – do some investigation and determine how to move forward (NDS751)
- Developer/personal installation (was kick the tires)
- Use case is Cyverse or TERRA-REF developer, interested in exploring Labs for their project, wants to bring up a local instance – this should be a 1 or 2 command process
- Where? On laptop, shared server, VM, Amazon (define the system requirements for the user)
- What does it mean for someone to install this? Be able to give to a developer to install? Virtual machine image
- Get feedback from an outside user?
- Minimizing resources while maximizing reliability and elasticity
- Ability to scale
- Alternate provider support (Commercial cloud)
- Revisit storage strategy (Gluster)