decentraland / comms-gatekeeper / 34872261654

87%

main: 86%

Build Type

Pull #283

github

Committed by LautaroPetaccio

Commit Message

refactor: replace the assignment-mirror adapter with a dedicated memory cache

The mirror was a thin wrapper over lru-cache with get, set and size, the
same library @dcl/memory-cache-component wraps. It is now an instance of
that component created in components.ts, sized by
CLUSTER_ASSIGNMENT_MIRROR_MAX and CLUSTER_ASSIGNMENT_MIRROR_TTL_MS with the
previous defaults. A separate instance rather than a namespace in the shared
cache, so mirror entries do not compete for slots with room-metadata-sync's
cooldown keys.

The subscriber reads and writes it through the async cache contract. The
un-grouped write is not awaited: the in-memory backend applies it
synchronously, which is what lets a connect queued behind the event resolve
the entry. It passes no per-call TTL, which on this API is in seconds, so
the instance default in milliseconds is the only unit in play.

MirrorEntry moves to the subscriber's types. The mirror settings, which were
missing from .env.default, are added there.

Pull Request Pull Request #283: feat: mint LiveKit island rooms from Pulse cluster feed (Archipelago ⇒ Pulse, iteration 1)

Coverage Stats

1255 of 1566 branches covered (80.14%)

Branch coverage included in aggregate %.

295 of 303 new or added lines in 15 files covered. (97.36%)

4 existing lines in 2 files now uncovered.

3428 of 3825 relevant lines covered (89.62%)

82.22 hits per line