Bound leaderboard staleness with database-owned cohort cache versions (CMP drift)

Aperta
#524 3 commenti 0 reazioni 0 assegnatari Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

Nessuno ha ancora preso questa issue.

Valutazione

Difficoltà
5/5
Tempo stimato
Più di una settimana
Idoneità per principianti
35/100
Tipo di issue
Bug
Chiarezza
Specificata chiaramente
Stato di attività
Attiva
Stack tecnologico
python

Direzione di ricerca

Read the four authority documents named in the issue, then trace the Cohort model, leaderboard recalculation, HTML row cache, and public export/YAML cache paths. Start by locating their existing tests and migration checks; use synthetic cohorts and learners with separate or stale cache instances to verify durable version reads, atomic rollback, visibility invalidation, and the 60-second configurable TTL. Done means the listed regression and migration checks pass while routes, payloads, privacy behavior, and dashboard behavior remain intact.

Scritto dal modello di indicizzazione a partire dal testo della issue.

Descrizione

bug courses P1

Problem and provenance

Leaderboard invalidation currently increments a process-local cache key. Other Gunicorn processes retain HTML rows for an hour and API output for 24 hours, including a learner who changes leaderboard visibility. CMP dae6e36 bounds TTL and 036ace2 moves the version to the database. Intake 835873e9-5f63-4e5d-b0f9-0773e1d562b6 authorizes small PRs and candidate-only delivery.

Authority

  • _docs/specs/04-courses-and-cohorts.md: cohort-owned leaderboard/visibility preferences; assessment state remains cohort-specific.
  • _docs/specs/07-security-privacy-operations.md: database-owned protected learner state and cache isolation.
  • _docs/architecture/database-only-content.md.
  • _docs/specs/10-verification-strategy.md; _docs/ci/change-selective-ci.md.

Scope

Adapt CMP's database version to the current Cohort model, never reintroduce the historical Course-as-edition model. Add the minimal migration and atomic F increment shared by leaderboard recalculation and visibility invalidation. Both HTML row cache and public leaderboard export/YAML keys read the durable version. Bound data caches to a configurable 60-second default. Keep existing payload shapes, routes, pagination and privacy semantics. Website dashboard builds directly from the database; do not introduce CMP dashboard cache or change edge/browser caching.

Acceptance criteria

  • Existing cohorts migrate with a valid initial leaderboard version; migration drift check and forward/reverse migration coverage pass.
  • HTML and JSON/YAML leaderboard readers key cached data by the cohort's current database version, independently of process-local cache version entries.
  • Recalculation commits totals and a monotonic atomic version advance together; failure rolls both back. Concurrent advances do not lose increments.
  • Changing display_on_leaderboard invalidates both row and export caches through the durable version; tests simulate separate/stale worker-local caches and prove hidden learners disappear on the next fresh read.
  • Data cache TTL defaults to 60 seconds and respects supported configuration; no 1-hour/24-hour internal defaults remain on these leaderboard paths.
  • Cohort isolation, pagination, existing response payloads, learner visibility and private response policies pass focused regression tests. Dashboard behavior is preserved.
  • Exact base/head verification plan, independent tester and PM acceptance, required checks green on PR exact head, merge and candidate live verification at prod.datatalks.club.

Boundaries and verification

No production data export or real-user mail. CMP stays read-only; courses.datatalks.club, current datatalks.club, join.datatalks.club, signup emails and Slack settings unchanged. Test synthetic cohorts/learners and fresh/stale model instances plus separate cache aliases/cleared local caches as appropriate. Internal cache-only work has no intended render change; computed plan determines Playwright and screenshot disposition. No dependency on the CSRF or review-copy ports.

Lingua principale
Python
Stelle
0
Fork
0
Merge medio
37m
PR unite (30g)
36

Preparare l'ambiente

  • Include un Dockerfile o un file Docker Compose
  • Nessun modello di pull request
  • Nessuna guida per i contributori

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di DataTalksClub/website

Tutte le issue di DataTalksClub/website

Issue simili

Altre issue su Python

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.