Harvester crashes frequently with async io error.

[Furious-George]

I have two practically identical debian harvesters running the same distro. It may take hours or days, but one always crashes with the following error: I've installed chia via apt on both.

2023-03-17T02:07:06.440 harvester harvester_server        : ERROR    Exception:  <class 'asyncio.exceptions.TimeoutError'>, closing connection {'host': '10.5.0.21', 'port': 8447}. Traceback (most recent call last):
  File "chia/server/server.py", line 634, in api_call
  File "asyncio/tasks.py", line 501, in wait_for
asyncio.exceptions.TimeoutError

2023-03-17T02:07:06.440 harvester harvester_server        : DEBUG    Inbound_handler task cancelled
2023-03-17T02:07:06.442 harvester harvester_server        : WARNING  Banning 10.5.0.21 for 10 seconds
2023-03-17T02:07:06.442 harvester harvester_server        : INFO     Connection closed: 10.5.0.21, node id: 201d3f186882a7c571ba5784d07580358e79ac08aafb1ddb4bf045333e916d0a
2023-03-17T02:07:06.443 harvester chia.harvester.harvester: INFO     peer disconnected {'host': '10.5.0.21', 'port': 8447}
2023-03-17T02:07:11.269 daemon chia.daemon.server         : DEBUG    About to ping: chia_harvester
2023-03-17T02:07:41.280 daemon chia.daemon.server         : DEBUG    About to ping: chia_harvester
2023-03-17T02:08:11.305 daemon chia.daemon.server         : DEBUG    About to ping: chia_harvester
2023-03-17T02:08:11.305 daemon chia.daemon.server         : ERROR    Ping error
Traceback (most recent call last):
  File "chia/daemon/server.py", line 283, in ping_task
  File "aiohttp/web_ws.py", line 295, in ping
  File "aiohttp/http_websocket.py", line 676, in ping
  File "aiohttp/http_websocket.py", line 653, in _send_frame
  File "aiohttp/http_websocket.py", line 663, in _write
ConnectionResetError: Cannot write to closing transport
2023-03-17T02:08:11.306 daemon chia.daemon.server         : ERROR    Ping failed, connection closed.

I've created a bug report, which is not getting any attention from the devs, which is disappointing to say the least, considering how much I've invested in their project.

The odd thing is that, once I switched to the apt package, the other harvester became much more reliable. I.e. when I was compiling from source with earlier versions, the harvesters would crash more often. If I could figure this out, then I would for the first time have harvesters that would not require constant monitoring. (The farmer, which is still compiled from source, no longer crashes with newer versions.)

I am not running any vms on the server that crashes, nor am I running any daemons that are not also running on the harvester that doesn't crash. The harvester that crashes has only fairly new hdds, while the server that doesn't crash is running a hodgepodge of new and older disks, some of the latter of which do fail from time to time, but the harvester is rock solid.

I could use some suggestions, as I have not received any in the months since I filed the bug report.

Thanks for reading.