fix(cron): prevent replay after persistence failure

This commit is contained in:
Xubin Ren
2026-08-15 23:34:35 +08:00
parent 8bdf5ed2b2
commit ecef2b055d
2 changed files with 63 additions and 14 deletions
+23 -1
View File
@@ -170,6 +170,7 @@ class CronService:
self._timer_task: asyncio.Task[None] | None = None
self._running = False
self._active_executions = 0
self._store_dirty = False
self.max_sleep_ms = max_sleep_ms
def _should_persist_store(self) -> bool:
@@ -305,6 +306,11 @@ class CronService:
load (during ``start``) can return ``None`` to signal an unrecoverable
state to the caller.
"""
# Never replace state that a previous save failed to persist. Reloading
# the older on-disk snapshot here could make an already executed job due
# again and repeat its side effect.
if self._store_dirty and self._store:
return self._store
if self._active_executions > 0 and self._store and not reload_during_execution:
return self._store
loaded = self._load_jobs()
@@ -347,6 +353,9 @@ class CronService:
if not self._store:
return
# Set this before serialization/write so every exceptional exit keeps
# the in-memory snapshot authoritative until a later save succeeds.
self._store_dirty = True
self.store_path.parent.mkdir(parents=True, exist_ok=True)
data = {
@@ -399,6 +408,7 @@ class CronService:
}
self._atomic_write(self.store_path, json.dumps(data, indent=2, ensure_ascii=False))
self._store_dirty = False
@staticmethod
def _atomic_write(path: Path, content: str) -> None:
@@ -514,10 +524,17 @@ class CronService:
reload_store = self._active_executions == 0
self._active_executions += 1
try:
# A prior tick may have completed external side effects but failed
# to persist their advanced schedule. Persist that exact snapshot
# before reloading or executing anything else; otherwise the older
# disk state can replay the same job.
if self._store_dirty:
self._save_store()
return
store = self._load_store(reload_during_execution=reload_store)
# If a hot reload found a corrupt store on disk, ``self._store``
# may still hold the previous, known-good in-memory snapshot.
# Keep using it rather than crashing the timer or wiping live jobs.
if store is None:
return
@@ -808,6 +825,11 @@ class CronService:
reload_store = self._active_executions == 0
self._active_executions += 1
try:
# A manual run is another side-effecting entrypoint. Do not start
# it while the result of a previous timer execution is still only
# in memory.
if self._store_dirty:
self._save_store()
store = self._require_store(reload_during_execution=reload_store)
for job in store.jobs:
if job.id == job_id:
+40 -13
View File
@@ -1,6 +1,7 @@
import asyncio
import json
import time
from pathlib import Path
import pytest
@@ -960,8 +961,8 @@ def test_stale_instance_remove_preserves_external_add(tmp_path) -> None:
@pytest.mark.asyncio
async def test_save_store_failure_does_not_kill_scheduler(tmp_path, monkeypatch):
"""A persistence failure in _on_timer must not stop future ticks."""
async def test_save_store_failure_retries_without_replaying_job(tmp_path, monkeypatch):
"""A failed post-run save must be retried before jobs can execute again."""
store_path = tmp_path / "cron" / "jobs.json"
calls: list[str] = []
arm_calls: list[str] = []
@@ -990,25 +991,51 @@ async def test_save_store_failure_does_not_kill_scheduler(tmp_path, monkeypatch)
service._save_store()
arm_calls.clear()
# Simulate a disk-write failure on the next tick.
def failing_save() -> None:
raise OSError("disk full")
real_atomic_write = service._atomic_write
save_attempts = 0
writes_fail = True
monkeypatch.setattr(service, "_save_store", failing_save)
def flaky_atomic_write(path: Path, content: str) -> None:
nonlocal save_attempts, writes_fail
save_attempts += 1
if writes_fail:
raise OSError("disk full")
real_atomic_write(path, content)
monkeypatch.setattr(service, "_atomic_write", flaky_atomic_write)
await service._on_timer()
# The scheduler must still be re-armed after the failed save...
# The failed tick stays alive and retains the advanced in-memory state,
# including when a public read would normally reload from disk.
assert arm_calls == ["arm"], "scheduler must re-arm after a failed tick"
assert service._active_executions == 0
# ...the first tick already ran the due job...
assert calls == [job.id]
# ...and a later, healthy tick must still run the job again.
monkeypatch.setattr(service, "_save_store", CronService._save_store.__get__(service))
job.state.next_run_at_ms = max(1, int(time.time() * 1000) - 1_000)
assert service._store_dirty is True
loaded = service.get_job(job.id)
assert loaded is not None
assert loaded.state.last_run_at_ms is not None
# Manual execution is a second side-effecting entrypoint. It must also
# refuse to run until the previous result can be made durable.
with pytest.raises(OSError, match="disk full"):
await service.run_job(job.id, force=True)
assert calls == [job.id]
# The next healthy tick is reserved for persisting the dirty snapshot. It
# must not reload the stale due record or execute the side effect twice.
writes_fail = False
await service._on_timer()
assert calls == [job.id, job.id]
assert arm_calls == ["arm", "arm"]
assert calls == [job.id]
assert arm_calls == ["arm", "arm", "arm"]
assert save_attempts == 3
assert service._store_dirty is False
persisted = CronService(store_path).get_job(job.id)
assert persisted is not None
assert persisted.state.last_run_at_ms is not None
assert persisted.state.next_run_at_ms is not None
assert persisted.state.next_run_at_ms > persisted.state.last_run_at_ms
@pytest.mark.asyncio