Add global BLE lock + faster timeouts for multi-device reliability

Root cause: 3 BMS devices fighting for 3 ESPHome proxy connection slots
simultaneously, causing 80% timeout failures and 22s+ poll times.

Fixes:
- Add shared asyncio.Lock so only one BMS polls at a time — eliminates
  proxy slot contention entirely
- Pass ble_device_callback to establish_connection so retry attempts
  get a fresh BLEDevice (handles proxy path changes)
- Reduce command timeout 5s -> 3s, retries 3 -> 2 (BMS responds in
  <200ms when connection is clean)
- Reduce establish_connection max_attempts 3 -> 2 (fail fast, retry
  next cycle instead of blocking 25s)
- Fixed poll timeout to 15s (was poll_interval-5=25s)

Expected: polls complete in 2-5s instead of 22s, ~95%+ success rate.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
2026-04-12 10:05:54 +02:00
parent 1520ed3c0f
commit dcc528b96a
3 changed files with 65 additions and 23 deletions
+32 -13
View File
@@ -22,6 +22,11 @@ _LOGGER = logging.getLogger(__name__)
# the last known data instead so the UI doesn't oscillate.
_FAILURES_BEFORE_UNAVAILABLE = 5
# Hard ceiling on the BLE poll operation (connect + commands + disconnect).
# With the global lock preventing contention, connections should be fast —
# 15 s is generous for 2 commands over a local proxy.
_POLL_TIMEOUT = 15
class BmsCoordinator(DataUpdateCoordinator[dict]):
"""Polls the BMS over BLE and distributes data to all sensor entities.
@@ -36,6 +41,7 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
address: str,
poll_interval: int,
name: str = "Xiaoxiang Smart BMS",
ble_lock: asyncio.Lock | None = None,
) -> None:
super().__init__(
hass,
@@ -45,12 +51,14 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
)
self.address = address
self._device_name = name
self._poll_timeout = max(poll_interval - 5, 20)
self._handler = BmsBluetoothHandler(address)
self.hw_version: str | None = None
self._consecutive_failures = 0
# Kept fresh by the BLE advertisement callback registered in __init__.py
self._ble_device: BLEDevice | None = None
# Shared across all BMS coordinator instances so only one BMS connects
# at a time — prevents ESPHome proxy connection slot exhaustion.
self._ble_lock = ble_lock or asyncio.Lock()
# ------------------------------------------------------------------
# Device info — shared by sensor, binary_sensor, number platforms
@@ -82,7 +90,11 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
raise HomeAssistantError(
f"BMS ({self.address}) not reachable — cannot send MOS command"
)
success = await self._handler.write_mos(device, value)
async with self._ble_lock:
success = await self._handler.write_mos(
device, value,
ble_device_callback=self._get_ble_device,
)
if not success:
raise HomeAssistantError("BMS did not acknowledge the MOS command")
await self.async_request_refresh()
@@ -135,17 +147,24 @@ class BmsCoordinator(DataUpdateCoordinator[dict]):
if self.hw_version is None:
commands.append(CMD_VERSION)
try:
responses = await asyncio.wait_for(
self._handler.poll(device, commands),
timeout=self._poll_timeout,
)
except asyncio.TimeoutError:
return self._handle_failure(
f"BMS poll timed out after {self._poll_timeout}s"
)
except Exception as exc:
return self._handle_failure(f"BMS poll failed: {exc}")
# Only one BMS polls at a time — prevents proxy connection slot contention.
# The timeout wraps only the actual BLE operation, not the lock wait.
async with self._ble_lock:
try:
responses = await asyncio.wait_for(
self._handler.poll(
device,
commands,
ble_device_callback=self._get_ble_device,
),
timeout=_POLL_TIMEOUT,
)
except asyncio.TimeoutError:
return self._handle_failure(
f"BMS poll timed out after {_POLL_TIMEOUT}s"
)
except Exception as exc:
return self._handle_failure(f"BMS poll failed: {exc}")
general_frame, cell_frame = responses[0], responses[1]