Skip to content

test_slow_downlink: give the env-limit run enough steps after the window #334

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
michaellchung merged 1 commit into from
Nov 3, 2026
Merged

test_slow_downlink: give the env-limit run enough steps after the window #334

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
14 changes: 11 additions & 3 deletions tests/test_slow_downlink.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,13 +119,13 @@ def steady(tape, syncer_log) -> bool:
return all(merges_with(records, island, joined, time.time() + 1) >= 10 for island in (0, 1))


def start_syncer(tmp_path, port, *extra):
def start_syncer(tmp_path, port, *extra, total_steps=1200):
binary = build_syncer()
tape = tmp_path / "tape.jsonl"
log = (tmp_path / "syncer.log").open("w")
proc = subprocess.Popen(
[str(binary), "--port", str(port), "--learners", "2", "--quorum", "1", "--grace-ms", "50",
"--total-steps", "1200", "--outer-lr", "0.7", "--outer-momentum", "0.0",
"--total-steps", str(total_steps), "--outer-lr", "0.7", "--outer-momentum", "0.0",
"--min-round-interval-ms", "20", "--quorum-timeout-s", "30", "--event-tape", str(tape),
"--island-lease-s", str(SYNCER_LEASE_S), "--syncer-epoch", "0", "--max-base-lag", "2", *extra],
stdout=log, stderr=subprocess.STDOUT, text=True, env={**os.environ, "TOKIO_WORKER_THREADS": "4"})
Expand Down Expand Up @@ -253,6 +253,13 @@ def test_env_downlink_limit_names_one_learner_and_opens_at_its_pull_step(monkeyp


ENV_FROM_STEP = 400
# The window opens at step ENV_FROM_STEP and the run must go on long enough
# after it for island 1 to drain its forwarder backlog (about 100 MB) and be
# merged again. Rounds are at least 20 ms apart, so at most 50 merges/s: by
# the end of the window the syncer is at step <= 400 + 14 * 50 = 1100, and
# 1700 steps leave at least 12 s after it. 1200 steps left 2 s at CI's merge
# rate, and the run could end before island 1 was merged again.
ENV_TOTAL_STEPS = 1700


@pytest.mark.timeout(300)
Expand All @@ -267,7 +274,8 @@ def test_env_downlink_limit_makes_a_running_island_slow(tmp_path, monkeypatch, i

monkeypatch.setenv(DOWNLINK_READ_LIMIT_ENV, f"1:{SLOW_BPS}:{ENV_FROM_STEP}:{SLOW_S}")
port = free_port()
proc, tape, log = start_syncer(tmp_path, port, *(["--inline-broadcast"] if inline else []))
proc, tape, log = start_syncer(tmp_path, port, *(["--inline-broadcast"] if inline else []),
total_steps=ENV_TOTAL_STEPS)
store = tmp_path / "store"
a = CountingIsland(0, port, store, dtype=DTYPE_F32)
b = CountingIsland(1, port, store, dtype=DTYPE_F32)
Expand Down
Loading