[BUG] Remote harvesters fail to connect when main farmer's IP address changes

Describe the bug When using remote harvesters, if the farmer is set to a hostname instead of an IP address in the configuration file, then they will repeatedly fail to reconnect if the farmer changes IP address:

23:04:43.790 harvester harvester : INFO Reconnecting to peer {'host': '(OLD IP ADDRESS)', 'port': 8447}

To reproduce

1. Set up a main farming machine and one or more remote harvesters
2. In the configuration file, set farmer_peer->host to the hostname of the farmer
3. Change the IP address of the farmer
4. Check the error log of the harvester

Expected behavior The remote harvester should not cache the IP address. DNS lookups should be performed on non-caching authoritative servers, either in the first instance or once the remote harvester has failed to connect X number of times.

In the remote harvester wiki page, it does mention to use an IP address. But a hostname should also be acceptable. Firstly, because the configuration parameter is named as host and by default is set to localhost. Secondly, it will be more convenient for many people using DDNS services.

In addition, this error is repeatedly raised every ~14 seconds but is only logged at INFO level. Since it prevents farming, it should be raised at a higher level, either on the first failed connection, or after X number of repeated failed connections.

Environment

• OS: Linux
• OS Version/Flavor: Ubuntu Server 20.04.2

Additional context The problem may also occur with other hostname configuration fields, such as connecting to a remote note or connecting to a remote daemon in “UI only mode”. But I have not tested this and only encountered the issue with remote harvesters.

[jcasc]

I think this is because the hostname is only resolved once and from then on, only the IP is used. A possible fix could be to add this one line

            ...
            if peer_retry: #line 20
                #### DynDNS Fix Attempt
                peer_info = PeerInfo(socket.gethostbyname(peer_info_arg.host), peer_info_arg.port)
                ####
                log.info(f"Reconnecting to peer {peer_info}")
            ...

to chia/server/reconnect_task.py

This would hopefully re-translate the hostname whenever a reconnection is attempted.

[smasharov]

The problem still exist in the 1.2.3