fix(mqtt): retire session workers on reconnect

This commit is contained in:
Jason Morcos committed 2026-08-18 11:10:38 -07:00
1 parent cd86424ca0
commit 8c374e17b4
3 files changed
+286 -7

No files matched your search

+1 -1
View File
@@ -166,7 +166,7 @@ def main():
stopping.set()
logger.info("shutting down…")
for b in bridges:
b.stop.set()
b.request_stop()
try: b.set_availability(False)
except Exception: pass
+45 -6
View File
@@ -87,6 +87,7 @@ OCF_STANDARD_SECURE_PORT = 5684
# fixed-source-port reconnect invariant is untouched (see session_once).
_GATE_RETRIES = 1
_GATE_TIMEOUT_S = 4.0
_WORKER_JOIN_TIMEOUT_S = 2.0
class PushBridge:
@@ -123,6 +124,8 @@ class PushBridge:
self.last_cycle_pub = None
self.last_avail_pub: str | None = None
self.stop = threading.Event()
self._session_stop_lock = threading.Lock()
self._session_stop: threading.Event | None = None
self.started_ts = time.time()
self.session_started_ts = None
self.last_change_ts = None
@@ -175,6 +178,13 @@ class PushBridge:
app.topic_prefix, shared.HA_DISCOVERY_PREFIX, app.device_name,
model=descriptor.name.title()))
def request_stop(self) -> None:
"""Stop the bridge and wake workers belonging to its current session."""
self.stop.set()
with self._session_stop_lock:
if self._session_stop is not None:
self._session_stop.set()
# ---- cache plumbing ---------------------------------------------
def _on_cache_change(self, changed: bool, source: str) -> None:
@@ -423,25 +433,54 @@ class PushBridge:
self.keepalive = keepalive
self.observe_refresh = observe_refresh
# These workers belong to this DTLS session, not to the bridge
# process. A reconnect must retire them before the replacement
# session starts or they continue operating on the closed session.
session_stop = threading.Event()
with self._session_stop_lock:
self._session_stop = session_stop
# ``request_stop()`` sets the bridge event before taking this
# lock. Checking it while publishing the handle prevents a lost
# wakeup if shutdown races this session handoff.
if self.stop.is_set():
session_stop.set()
sched_t = threading.Thread(
target=scheduler.run_forever, args=(self.stop,),
target=scheduler.run_forever, args=(session_stop,),
daemon=True, name=f'{self.app.klass}-poll')
ka_t = threading.Thread(
target=keepalive.run_forever, args=(self.stop,),
target=keepalive.run_forever, args=(session_stop,),
daemon=True, name=f'{self.app.klass}-ping')
ref_t = threading.Thread(
target=observe_refresh.run_forever, args=(self.stop,),
target=observe_refresh.run_forever, args=(session_stop,),
daemon=True, name=f'{self.app.klass}-obsref')
sched_t.start()
ka_t.start()
ref_t.start()
workers = (sched_t, ka_t, ref_t)
started_workers = []
try:
for worker in workers:
worker.start()
started_workers.append(worker)
sess.join()
finally:
# A worker already inside a tick can finish after the reader
# exits. Disable old-session reachability callbacks first so it
# cannot change availability after a replacement takes over.
keepalive.on_reachable = None
keepalive.on_unreachable = None
session_stop.set()
join_deadline = time.monotonic() + _WORKER_JOIN_TIMEOUT_S
for worker in started_workers:
worker.join(max(0.0, join_deadline - time.monotonic()))
if worker.is_alive():
self.log.warning(
"session worker did not stop: %s", worker.name)
self.scheduler = None
self.keepalive = None
self.observe_refresh = None
with self._session_stop_lock:
if self._session_stop is session_stop:
self._session_stop = None
def _seed_from_device0(self, sess):
code, pl = sess.get(self.descriptor.seed_path, timeout=15.0)