Fix work recovery and reduce engine and chart overhead
Docs / docs (push) Successful in 33s
Playwright Tests / test-playwright (1, 2) (push) Failing after 1m37s
Playwright Tests / test-playwright (2, 2) (push) Failing after 27s
pre-commit / pre-commit (push) Failing after 2m37s
Test Backend / test-backend (3.12) (push) Successful in 5m7s
Test Backend / test-backend (3.14) (push) Successful in 5m2s
Compose Smoke Test / test-compose (push) Failing after 28s
Playwright Tests / merge-reports (push) Failing after 2m29s

This commit is contained in:
2026-10-29 14:24:17 +01:00
parent 20a0853f69
commit 75b3a93094
25 changed files with 1316 additions and 301 deletions
+55
View File
@@ -160,6 +160,61 @@ def test_a_flush_the_database_refused_is_written_by_the_next_one(
assert audit.detail == "held back"
def test_batched_rollups_add_and_redelivery_clears_a_finished_run(db: Session) -> None:
collector = MetricsCollector(EventBus())
ts = time.time()
for i in range(2):
collector.handle(
{"type": "cascade_started", "run": f"bulk-{i}", "flow": "bulk", "ts": ts}
)
collector.handle(
{
"type": "node_executed",
"flow": "bulk",
"node": f"bulk.n{i}",
"run": f"bulk-{i}",
"outputs": 1,
"duration_ms": 10.0,
"ts": ts,
}
)
collector.handle({"type": "cascade_finished", "run": "bulk-0", "ts": ts + 1})
# Write the same open/finished batch twice, as a retained run is upserted.
collector._write(collector._buckets, [], list(collector._runs.values()), False)
collector.handle(
{
"type": "cascade_started",
"run": "bulk-0",
"flow": "bulk",
"deliveries": 2,
"ts": ts + 2,
}
)
collector.handle({"type": "cascade_finished", "run": "bulk-1", "ts": ts + 3})
for bucket in collector._buckets.values():
bucket["duration_sum_ms"] = 4.0
bucket["duration_max_ms"] = 4.0
collector._write(collector._buckets, [], list(collector._runs.values()), False)
buckets = db.exec(select(MetricBucket).where(MetricBucket.flow == "bulk")).all()
assert len(buckets) == 2
assert all(
(b.executions, b.messages, b.duration_sum_ms, b.duration_max_ms)
== (2, 2, 14.0, 10.0)
for b in buckets
)
reopened = db.get(FlowRun, "bulk-0")
assert reopened is not None
assert (reopened.status, reopened.finished_at, reopened.deliveries) == (
"running",
None,
2,
)
finished = db.get(FlowRun, "bulk-1")
assert finished is not None and finished.finished_at is not None
assert finished.status == "ok"
def test_a_node_id_wider_than_the_column_still_records(db: Session) -> None:
collector = MetricsCollector(EventBus())
ts = datetime.now(UTC).replace(second=0, microsecond=0).timestamp()