Added
- A
telemetry_stalledcheck for an inverter that keeps answering but stopped updating most of its readings (#227). It fails once the voltages and frequency stand still and either none of the solar, grid import and battery discharge counters, which supply the house, has moved for 20 minutes, day or night. A model that reports none of those counters is judged by most readings standing still for 30 minutes instead. Control steps aside, and every measured reading that stopped changing shows unavailable, including one that stands at 0, while those that still move, such as grid and solar power, stay shown. How long each reading has stood still survives a restart. - A repairs issue once
telemetry_stalledhas lasted 30 minutes, asking to restart the inverter. It goes as soon as the readings move again. - A Reading Check diagnostic sensor, on while a check shows readings to be wrong, with the check, since when and the distrusted readings as attributes. Unlike Control Status, it shows them while Modbus Control is off too.
- Coordinator Status shows Readings stalled instead of Success while
telemetry_stalledholds, since a poll that runs fine can still bring readings the inverter no longer updates. - A repairs issue when the inverter has not answered for 5 minutes or more, as after a firmware update that switched Modbus TCP off (#222). The inverter then stays silent instead of reading 0, which the Modbus disabled issue does not catch. The issue says whether the port still takes connections or nothing answers at all, and that in a few cases a firmware update has switched Modbus TCP off. It stays across a reload and clears with the next good read.
- Setup and Reconfigure say when the inverter takes connections but sends no replies, instead of asking to check the address, and that a firmware update has in a few cases switched Modbus TCP off.
- Firmware changes appear in the logbook, from an
ef_powerocean_tcpmodbus_firmware_changedevent.
Changed
- The
frozencheck judges each voltage and the frequency on its own, and fails once most of them, an AC voltage among them, haven't changed for five minutes instead of all of them for two (#227). One reading that still moved hid the others before. Each one that stopped shows unavailable while the check fails, and so does anything calculated from a distrusted reading, such as Battery Remaining Energy from the Battery SOC. - The
frozencheck judges each voltage and the frequency on its own, and fails once most of them haven't changed for five minutes instead of all of them for two (#227). One reading that still moved hid the others before. Each one that stopped shows unavailable while the check fails, and so does anything calculated from a distrusted reading, such as Battery Remaining Energy from the Battery SOC. - A lost connection is retried after 0, 5, 10 and 20 seconds, then every 30 seconds instead of every 2 minutes, so the inverter is read again soon after it is back. On the connection Home Assistant shares, requests time out after 3 seconds, as on an own connection, instead of 10.
- Entities now show unavailable while a poll fails, and available again with the next good one, as the README describes. Before, they kept showing their last value, because a value that had not changed wrote no state (#219).
Fixed
- With Modbus Control off, a reconnect no longer makes every poll write the default control method to the inverter (#227). An inverter that left the write unanswered dropped the connection each time, so polls ran every 25 seconds instead of every 5 and never stopped writing.
- On the connection Home Assistant shares, a request that times out now drops the connection, and the next poll opens a new one. A connection whose packets a firewall dropped stayed open without answers, and kept the inverter unreadable for minutes after it was reachable again.
- Entities whose value had not changed could stay unavailable after a start whose first read failed, until their value next changed (#219).
Full Changelog: v2.10.0-beta.5...v2.10.0-beta.6