Shouldn't InfluxDBServerError be raised when InfluxDB fail to serve a request (e.g. HTTP status is 5xx) instead of InfluxDBClientError.

I need this distinction, because I've some code retry writing when server fail (not started/starting) but discard malformed input. E.g. :

try:
    client.write_points(...)
except InfluxDBClientError:
    # do not retry this write, it means malformed points
except (InfluxDBServerError, ConnectionError):
   # do retry this write, it means server down/starting

This would also be valuable for InfluxDBClusterClient, who should go to next server if one of the server fail.

One case were InfluxDB return a 500 error code (that is not due to bad input) is during startup of single node, before metastore initialization. In this case I got the error metastore database error: no leader. This is obviously only a temporary error and few seconds later server is available.