Open jayvdb opened 4 years ago
The first steps are tests for TCP vs UDP of the dnspython Timeout
exception upstream, and increasing the min ttl as the number of requests are performed, to recover quickly from when the network was really the problem.
Caching of lookup failures has some caching implications which are different to positive lookups, e.g. what does a timeout imply? The probable meaning is different on udp vs tcp based protocol - is that difference correctly understood in the
dnspython
code, and which exception it raises, or do we need to address that here or upstream.Also there are usually other ways to verify the DNS lookup failure, which can be used to get a more reliable/appropriate min ttl for the failure.