Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 19 additions & 8 deletions .github/workflows/ts.yml
Original file line number Diff line number Diff line change
Expand Up @@ -217,19 +217,30 @@ jobs:
env:
MOQ_TEST_RUNS: ${{ runner.temp }}/moq-test-runs

# 1% random loss and 25 ms each way on the subscriber's leg (moq-dev/moq#4613), through
# moq-shaper: QUIC retransmission must keep the export T-STD compliant with no media lost.
# `--hrd`, since `--headroom` still loses audio just after the join at 1%.
- name: TS lossy path
# The subscriber's leg congested the way real paths are, through moq-shaper: 7 Mb/s under
# the 9 Mb/s video, behind a 2 s queue, 25 ms each way. The export must stay T-STD compliant
# and continuous, shed only whole groups, and keep its latency, so the last group arrives.
- name: TS congested path
if: ${{ !cancelled() && github.event_name != 'pull_request' }}
run: nix develop --command just test ts --hrd --loss 1 --duration 120
run: nix develop --command just test ts --hrd --rate 7000000 --duration 120
env:
MOQ_TEST_RUNS: ${{ runner.temp }}/moq-test-runs

# The passthrough over the same path: retransmission must leave no packet missing or moved.
- name: TS passthrough lossy path
# The same path behind a 250 ms queue, with the relay's loss-based CUBIC: BBRv3 never fills a
# queue this deep, CUBIC overflows it every sawtooth. QUIC must recover what the queue drops
# inside the export's 1 s delay, and the run fails unless the queue overflowed.
- name: TS overflowing path
if: ${{ !cancelled() && github.event_name != 'pull_request' }}
run: nix develop --command just test ts --passthrough --loss 1 --duration 120
run: nix develop --command just test ts --hrd --rate 7000000 --queue 250ms --duration 120
env:
MOQ_TEST_RUNS: ${{ runner.temp }}/moq-test-runs
MOQ_QUIC_CONGESTION_CONTROL: loss

# The passthrough over the congested path, under its 10 Mb/s multiplex: what arrives must be the
# source's packets in place.
- name: TS passthrough congested path
if: ${{ !cancelled() && github.event_name != 'pull_request' }}
run: nix develop --command just test ts --passthrough --rate 7000000 --duration 120
env:
MOQ_TEST_RUNS: ${{ runner.temp }}/moq-test-runs

Expand Down
1 change: 0 additions & 1 deletion quest/m1/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -102,7 +102,6 @@ lands before the cut.
- [Transcode finishes its group](/quest/m1/transcode-open-group.md) - `moq-transcode` never aborts a group a fetch's live loop opened
- [Rust request credit](/quest/m1/rs-request-credit.md) - a moq-net request fails at once without stream credit instead of waiting
- [qmux no-wait opens](/quest/m1/qmux-no-wait.md) - @moq/qmux rejects an over-limit create when waitUntilAvailable is false, like Chrome
- [Bufferbloat harness](/quest/m1/bufferbloat.md) - acceptance and performance checks impair paths with a rate limit and a deep queue, not random loss
- [Random-loss audit](/quest/m1/random-loss-audit.md) - every decision random-loss gates drove is kept, reworked, or reverted on shaped paths
- [Timed-out requests clean up](/quest/m1/legacy-request-timeout-cleanup.md) - on drafts 14-16, Rust and JS send PUBLISH_NAMESPACE_DONE or UNSUBSCRIBE_NAMESPACE for a timed-out request before its local reset
- [TS fidelity and recovery](/quest/m1/ts-reliability/README.md) - both modes have explicit profiles, automated validation, loss tests, and a dedicated guide
Expand Down
53 changes: 0 additions & 53 deletions quest/m1/bufferbloat.md

This file was deleted.

14 changes: 9 additions & 5 deletions quest/m1/random-loss-audit.md
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,15 @@ is not how real paths fail, so anything tuned to survive it as a quality or
performance target may be wrong. Scope is every random-loss-driven
decision, not only the 10% TS work. Small reverts land in the audit's PR;
anything bigger becomes its own quest. Verdicts are measured on the
[bufferbloat harness](/quest/m1/bufferbloat.md), not argued.
bufferbloat harness (`just test ts --rate`, see `test/ts/README.md`), not
argued.

Measured with that harness 2026-10-18: BBRv3 alone keeps about a round trip
queued, so a deep queue never overflows and congestion lands at the relay as
`Old` group evictions instead. A verdict that needs congestion loss, such as
BBR's loss response, needs a queue about a round trip deep or a loss-based
sender sharing the path. The nightly overflowing path (2026-10-19) runs
CUBIC alone behind a 250 ms queue, which it overflows every run.

Leads found 2026-10-17 (no production default or congestion threshold was
changed for random loss; `LOSS_THRESH` is still 2%):
Expand Down Expand Up @@ -46,7 +54,3 @@ Record each verdict (decision, origin PR, keep/rework/revert, evidence) in
the PR description.

Public API: depends on the verdicts. Wire: none expected.

## Required

- [Bufferbloat harness](/quest/m1/bufferbloat.md) - verdicts are measured on shaped paths
2 changes: 1 addition & 1 deletion quest/m1/ts-reliability/grader-determinism.md
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,6 @@ Two sources of run-to-run variance, found by #5810 and #5854:
Decided 2026-10-17 (maintainer): one quest for the grader, with the CI path
filter as a separate m0 quest. The harness
never gates on a random-loss percentage (see
[bufferbloat](/quest/m1/bufferbloat.md)).
the root `AGENTS.md`).

Public API: none. Wire: none.
2 changes: 2 additions & 0 deletions rs/moq-relay/tests/drills.rs
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,8 @@ const TRACK: &str = "drill";
/// hostile one, and the drills' own budgets (a 2s idle timeout, 10s waits) hold
/// under it unchanged. The loss is 5% because a drill only sends a few hundred
/// datagrams, and at 2% that is a drop or two, which barely exercises recovery.
/// Random loss only proves correctness, which is all a drill asserts; nothing
/// here judges quality, latency or throughput.
fn impairment() -> moq_shaper::Profile {
moq_shaper::Profile {
delay: Duration::from_millis(20),
Expand Down
9 changes: 9 additions & 0 deletions test/drill/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -137,6 +137,15 @@ would turn this lane into a second loopback run that passes for free.

There are no retries: a drill that only passes on loopback is a finding.

Random loss is not how real paths fail, so it may prove correctness but never
judge quality, latency or throughput; those run on a congested path (see
`test/ts/README.md`). Every drill here proves correctness: that a resource is
released, a failure surfaces, a route or publisher comes back, or every group is
recovered. The cluster drills' 10s bound on each group is a stall check, a third
of the 30s stalls of #4349, not a latency target. So the impaired lane keeps its
loss, which is what drops the `MAX_STREAMS` and `MAX_DATA` frames the cluster
drills' senders wait on.

The same shaper is a binary, for putting in front of a relay from another
process. It applies one profile both ways, prints the seed at start and the
counters at exit (Ctrl-C or SIGTERM), and exits nonzero if the profile never
Expand Down
Loading
Loading