Pietro’s week
- Joined in the Canonical Engineering Sprint in Frankfurt
- Joined two workshops on working out an ‘official’ reference for the juju state machine / event hooks; result to be published soon somewhere. Preview here
- Worked out a reference for the Unit/Workload statuses here
- Finished work on this pr
- Continued work on this loki-k8s operator pr, debugged a lot and found out this bug; will now work on finalizing the PR by xfailing the syslog tests.
- attended community workshop for beginner charmers
This week, I …
-
Participated in some excellent conversations around event ordering and best practices.
My core advice is that you should feel free to rely on event ordering when needed, but you should try to fetch state from the Operator Framework’s cache whenever possible, and you should be mindful of the non happy paths where events might occur out of order. (E.g., you defer an event, then defer again on the next hook run, thus removing that event from the expected ordering.)
-
Tested the rollingops library in a live model . Found and fixed some bugs. And started work on a formal integration test against the live model (which can be generalized and applied elsewhere).
-
Various code reviews and follow-ups on issues. If I promised to follow up on something, and don’t get to it by EOD today (Friday, March 4th), please feel free to ping me – it’s possible that it didn’t make it onto my todo list.
- Reviewed several PRs including health checks, test harness send_signal, and relation data tracking.
- Helped discuss design and do review of Pen’s rolling operations lib.
- Fixed a CI dependency issue.
- Finished up recursive Push/Pull spec - waiting for review from above.
- Added Container.exists and Container.isdir.
- Exposed JUJU_DEPARTING_UNIT for RelationDeparted events.
- triaged a few github issues including #547, #708, and #692.