Since Wikibase Cloud went live, there have been recurring issues with the QueryService missing data. The oldest report of this was in November last year: https://phabricator.wikimedia.org/T322753
Since the last few weeks, we've been seeing increased reports:
* https://phabricator.wikimedia.org/T344142 (gives specific examples, helpful for investigation)
* https://phabricator.wikimedia.org/T343034 (most recent)
* https://phabricator.wikimedia.org/T345199 (also contains reports from other Wikibases in the comments)
We're already aware of 2 potential solutions we want to try out:
* Adjusting the order in which we batch updates for the QueryService, so that "last in line" data doesn't appear in the QueryService with a massive delay. https://phabricator.wikimedia.org/T333164
* Don't mark batch updates as "done" without the full check that they are, to make sure no data is missed there https://phabricator.wikimedia.org/T333163
There are multiple places where data injection into the QueryService can be missed.On the 19th of November 2023 in the morning, In addition to the 2 solutions we're already going to dothe Queryservice updater started crashing, we want to do an investigation to find the root cause (s).
AC:
* Investigation is done independently from https://phabricator.wikimedia.org/T333164 and https://phabricator.wikimedia.org/T333163
* User reported issues are investigated and the root cause is found
* Followup ticket(s) from this investigation is/are created
resulting in no data being updated in the Queryservice for any wiki on wikibase.cloud