During previous times working on data center (DC) switchovers, we have observed a lot of room for improvement. The goal of this task is to provide a better quality service.
Here is the improvement plan:
- Discuss with Erica about having a third person involved in the process. Rationale:
- We want to add quality checkins, with one person doing the task and the other checking up on it.
- Since we only work for 1,5 FTE in total, can have vacations and times off, the idea is to have a third person to be a reliable backup and cover these times off.
- Based on Tech News experience, we prefer not to rely on volunteers to help for now, even if some would be available and skilled.
- Create a Phabricator template, maintained by us, to request banners for server switches
- Reuse the current steps we have
- Add a checkup step for translations (are they done, are links functioning, etc?)
- Add a checking step for each item, written down in the checklist.
- Improve messages sent to communities
- We only have reusable messages, which cover our needs: One for DC switchover
- Change the date format so that translators don't have to deal with dates translations at each usage of the message (reuse Tech News') (see related translators-l@ thread). But leave space for local times for languages that cover multiple timezones.
- Decide if we keep the possibility to translate the time of the event, which includes possible mistakes if 1/ the UTC time changes 2/ the local time observes DST.
- Document publicly the fact that:
- two people work on a task, the assignee and the checker
- the assignee does the job, and the checker checks it
- assignee can change but in this case, the task is reassigned
- we debrief after each task, and we iterate on the process if needed
- provide links to the two messages so that translators can find them
- Document a Q&A process so that another CRS can check a banner configuration.
Until March 16 2022, this task covered regular maintenance read-onlys and databases switchover (DC switchover). This process changed (T303605), and the scope of the current task was redefined to only take care of DC switchover.
Regular maintenance read-only times were discussed in this task. The conclusion of the discussion is that we don't need to announce them anymore as the read-only time is minimal.