Skip to content

Recording pipeline

How ReplayCore captures, samples, compresses, and uploads replay data.

7 min read

Pipeline overview

The recorder captures gameplay on dedicated background threads, never on the server tick thread. Captured events are written to a local, compressed .replaycore archive on disk and uploaded to the ReplayCore cloud in multipart batches. This design keeps the recorder off the hot path so it does not affect TPS.

Durability during outages

ReplayCore writes and seals authentic local replay archives before upload. During an eligible authenticated outage, uploads queue and retry under the managed Connection policy; when connectivity returns, queued archives are sent in the background.

The managed Storage policy can seal standalone archives by match or by duration and size thresholds. Segment finalisation preserves session continuity and leaves recoverable local evidence instead of treating an interrupted upload as completed.

Compression and upload

Replay data is compressed before multipart upload. Configuration Studio's Connection section controls supported chunking, retry, queue and upload-bandwidth settings. The default shared upload limit is 1 MiB/s. A missing or zero legacy limit uses this safe default. A server operator can set the local-only cloud.upload-unlimited option to true when that server must upload without a limit. The recorder applies only authenticated, validated managed values.