Large numbers of errors in log

[jech]

I've just run Galène against pion/turn instead of coturn. The server was moderately loaded, a few lectures with 20 or so students each. The log is riddled with messages that indicate problems between Galène and the TURN server — I haven't noticed any that indicate issues with communication with clients.

If my understanding of TURN is correct, some of the errors are not actual errors, but simply normal negotiation between the TURN client and server; in that case, the issue is that the default logging is too verbose. Others look worrying, and I'd appreciate a clarifiction.

Typical errors include:

turnc ERROR: 2021/01/19 16:34:56 fail to refresh permissions: CreatePermission error response (error 441: Wrong credentials: the realm value is incorrect

and

turn ERROR: 2021/01/19 16:35:01 error when handling datagram: failed to handle CreatePermission-request from 51.210.14.2:37880: no allocation found 51.210.14.2:37880:51.210.14.2:1194

and

turnc ERROR: 2021/01/19 16:35:09 fail to refresh permissions: all retransmissions failed for yoM0LWftNldeEfRt

and

turn ERROR: 2021/01/19 16:35:51 error when handling datagram: failed to handle Refresh-request from 51.210.14.2:37880: no allocation found 51.210.14.2:37880:51.210.14.2:1194

[jech]

So I've looked at it in some more detail, and I see the following problems:

• function readLoop unconditionally calls s.log.Errorf; it should call Warnf or Logf for errors that are likely to happen in normal behaviour;
• the errors in internal/server are not exported, which means that you cannot easily decide on their type.

I can see two ways of solving the second issue. The simplest one would be to export the errors. The second one would be to add a method IsWarning(error) bool to the internal patckage that returns true if an error should be reported using Warnf.