zzstoatzz.io
prefect server in zig
fix scheduler idempotency test race condition
root cause: test was observing partial scheduler tick. each flow run
insertion commits separately, so test could see 25 runs mid-tick and
then 50 when tick completed - appearing as a failure.
fix: wait for count to stabilize before recording baseline, ensuring
we measure after the scheduler tick completes rather than during it.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
+12
-2
+12
-2
scripts/test-api-sequence
+12
-2
scripts/test-api-sequence
···
1018
1018
deployment_id = deployment.get("id")
1019
1019
log(f" deployment: {deployment_id}")
1020
1020
1021
-
# poll for scheduler to create first run (instead of fixed sleep)
1022
-
log("[bold]waiting for scheduler to create runs...[/bold]")
1021
+
# poll for scheduler to create runs and wait for tick to complete
1022
+
# we need to wait for count to stabilize, not just see the first run
1023
+
log("[bold]waiting for scheduler tick to complete...[/bold]")
1023
1024
count_after_first = wait_for_runs(deployment_id, min_count=1)
1025
+
# wait for count to stabilize (scheduler tick may still be inserting)
1026
+
for _ in range(10):
1027
+
time_mod.sleep(0.2)
1028
+
resp = client.post("/flow_runs/filter", json={"limit": 100})
1029
+
if resp.status_code == 200:
1030
+
matching = [r for r in resp.json() if r.get("deployment_id") == deployment_id]
1031
+
if len(matching) == count_after_first:
1032
+
break # count stable
1033
+
count_after_first = len(matching)
1024
1034
log(f" runs after first tick: {count_after_first}")
1025
1035
1026
1036
if count_after_first == 0: