[dnsdist] Marking downstream Unexplained

[Alejandro Adroher Mellado]

These new ideas sounds really great!! 

-----Original Message-----
From: Remi Gacogne [mailto:remi.gacogne+dnsdist at powerdns.com] 
Sent: martes, 16 de febrero de 2016 16:12
To: Alejandro Adroher Mellado <alejandro.adroher at omniaccess.com>
Cc: dnsdist at mailman.powerdns.com
Subject: Re: [dnsdist] Marking downstream Unexplained

Hi Alejandro,

> Can I enable verbose for loggin?? 

Unfortunately we are not logging more information regarding health check failures, even at verbose level. That's mostly because it might fill up your logs very quickly otherwise.

Based on the messages you posted, only one health check fails and the next ones are fine. So basically, dnsdist sent a UDP query and either did not get any response at all in less than one second, or the response was invalid:
- complete garbage or
- response ID not matching the query ID or
- QR flag not set or
- ServFail or
- NXDomain or Refused (only considered an error if you have set mustResolve to true for this backend).

If you don't have a lot of traffic you might be able to see what happen by using tcpdump.

Perhaps we should add an option to consider a backend as down only after several checks failed in a row. Maybe some counters for health check failures might help too.

--
Remi Gacogne
PowerDNS.COM BV - https://www.powerdns.com/