/
10.10.2023 Release V8.1.0

10.10.2023 Release V8.1.0

Attendees: @James Friel @James Sutherland (Staff) @Ross Teviotdale (Staff) @Susan K

 

What was the goal?

Release RDMP v8.1.0 with improved performance, managed (scheduled and controlled) and timely release .Engage with Stakeholders around release cycle.

(Late addition) smooth transition of ownership and process.

Did you achieve what you set out to do?

  1. Did the project fully solve the problem it was designed to address?

Did we improve performance? yes

Managed release? Ish

  • Delays in releasing

  • Large weakness is lack of performance related testing

  • no benchmarking

  • no release pipeline

  1. What lessons did we learn that we can apply to future projects (next section)?

    1. Should do more interim releases?

    2. Better approach to releases - auto numbered betas? More aggressive patch releases?

    3. more managed releases, makes idenitfiy when a bug is released easier

Lessons learnt for more effective future projects

  1. What went well and what should we keep doing?

    1. Having a team

    2. Involving the Software team

    3. Proper look at the UI - improvements, and long-term look

    4. managed release process was useful

    5. large amount of improvements (is a bit hidden)

  2. What could have been improved or what should we stop doing?

    1. Data team interactions

    2. EPCC deployments - MSI packaging?

    3. frequency - minimum once a quarter

  3. In hindsight, what were the risks we didn’t see coming?

  • 8 months to replace Thomas - 3 rounds of recruitment -not suitable candidates

  • Recruitment in general

  • Small team size (bus factor) - need to distribute the knowledge

  • RDMP being a legacy project - all dependencies are old, may just stop working.

  • insufficient handover(s)

  • unexpected quirks/hardcoding that are sub-optimal (inefficient, should be in mainline)

  • 1 off hacks for specific projects, don’t go back to mop them up

 

  1. How do we improve as a team? 

  • backfill documentation as we go through it

-i.e. cohorts -normal rdmp users & how hic wants it to work, how it actually works in HIC due to Jira not working

  • pipeline for moving one-time hacks to mainline code