Repository navigation
quest: run redundant importers on distinct epochs - #5963
Conversation
Drop same-epoch importers: delete the epoch-anchor and hop-aligned-import quests, add an m0 quest under Broadcast epochs that removes `moq --epoch` and reverts the shared-epoch-only alignment code, and an m1 guide on redundant ingest over multiple routes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
…instance The maintainer picked switchover, so delete the redundancy-model decision quest and repoint its references at distinct importer epochs. Record the principle in AGENTS.md: one publisher per broadcast instance, and the subscriber switches across instances rather than the relay stitching independent publishers. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1de0d4626d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d5d2c94f8d
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
Summary before merge: redundant importers run on distinct epochs, and each publisher produces its own stream that the end subscriber switches between, combines, or keeps separate. This adds the m0 (Written by Claude Opus 5.5) |
Summary
Re-plans redundant ingest. Two importers fed one stream no longer share an epoch and try to publish identical tracks. Each mints its own epoch, and consumers follow a failover as a marked restart.
quest/m0/broadcast-epoch/distinct-importer-epochs.md[L]: removesmoq --epoch(never shipped, so it is absent fromrelease) and reverts the importer andmoq-muxcode that existed only so same-epoch importers agree (feat(mux): derive same-epoch importers' clock and eras from the epoch #5625's epoch-derived clock and eras, feat(mux): cut passthrough groups and sections on the epoch's grid #5894's grid cuts and windows). It adds theexport tsexits with "frame timestamp is below the live edge" when the relay switches to a same-hop standby quickly #4354 1+1 failover test on distinct epochs and closesexport tsexits with "frame timestamp is below the live edge" when the relay switches to a same-hop standby quickly #4354. It is a child of Broadcast epochs, so it gates the next release.quest/m1/redundant-ingest-guide.md[S]: a generaldoc/setup/redundancy.mdpage on redundant ingest over multiple routes, for any publisher.quest/m1/redundancy-model.md(this PR makes its decision: switchover),quest/m2/epoch-anchor.md, andquest/m2/hop-aligned-import.md, plus every reference to them (m1 and m2 READMEs,feature-review,pes-import,ts-hitless, e2ee, cluster-routing).AGENTS.md(Public API): one bullet saying each publisher produces its own stream and the end subscriber decides how to switch between, combine, or keep them separate, never independent publishers made identical so the relay can stitch them.Why: independently fed importers each see their own loss, start point, and history. Once loss or reordering hits one of them, "same epoch, same content" is false, and a mid-group resume splices two streams together. Each alignment rule added a cost (a 13.25 h wall-mapping bound, half-period failure cases, a 1 s join penalty, empty groups,
GroupFullError), and audio after loss, the catalog tracks, and the timeline still diverged.#5895
Recommendation: close #5895 after this merges. Nothing in it survives: the one-second group grid only exists for shared epochs, and so does the unpushed video-cut fix (
dd52991e, which closes a damaged group as a prefix of the partner's). A comment on #5895 explains this and links here.Decisions
How should redundant TS importers fail over?
--epochis removed, and a failover is a marked restart.$Time$style).When a backup importer starts (newer epoch) while the serving one is up, what should
export ts --stitchdo?How should the work split into quests?
What should happen to #5895?
Which milestone should the quest join?
What bar should the audit use?
Where should the redundant-ingest setup be documented?
cli.md.Where in m0?
Where should the one-publisher principle go?
AGENTS.md, Public API.rs/moq-net/AGENTS.md.Should the AGENTS.md change ride in this PR?
Where should the guide go?
doc/setuppage, requiring the m0 quest.Checks
quest checkpasses (433 documents). Quests and oneAGENTS.mdline, no code.(Written by Claude Opus 5.5)
🤖 Generated with Claude Code