i could pull them via nodered from the timescaledb (postgres SQL querry and then create a parquet file from that), this works but is not the greatest solution (nodered is not great at large quantity of data), how did you do it?