Page MenuHomePhabricator

[SPIKE] Verify Performance of Manual Topic Subscriptions at Phase 2 Wikis
Closed, ResolvedPublic

Description

This ticket involve the work with verifying with the Performance / Data Persistence Teams that the deployment of Manual Topic Subscriptions as an opt-in Beta Feature at the Phase 2 Wikis (T287801) has NOT caused any performance regressions.

Open questions

  • What – if any – impact has the introduction of Manual Topic Subscriptions as a Beta Feature at all Wikipedias except the English Wikipedia (T287801) had on the site's performance?

Background

T263817#7030301:

  • Expected growth per year (number of rows).
    • We don't know yet. Hypothetically, if every one of ~30,000 users who edited talk pages last month subscribed to 3 new threads per day, we'd end up somewhere around 32M rows per year (on enwiki). This is probably a gross overestimate though.
    • We should be able to get a better idea once topic subscriptions are deployed on smaller wikis or as a beta feature.

T280412:

Once the topic subscription feature is deployed to production (T274188), any time a new comment is posted in a section, that event is stored in a database (T264885).

If no changes are made to how these comments are stored, the performance of the database, and subsequently, the feature could degrade.

Done

  • All ===Open questions are answered and documented in Phabricator

Event Timeline

ppelberg claimed this task.

It's been >6 months since Manual Topic Subscriptions were deployed (T287801#7309606) and we have yet to become aware of any performance issues related to it.

As such, I'm going to assume that Topic Subscriptions have not caused any performance issues and as such, consider this task resolved.

Please re-open this task if you think the judgement I've made above is inaccurate/not considering relevant information.