Add more debugging messages for network-level events
The network manager code is currently not particularly verbose when it comes to logging debug messages:
$ git grep isc_log_write lib/isc/netmgr/ | wc -l
9
In particular, this applies to "positive" events (non-errors), like establishing a connection, correctly receiving data from a socket, etc. This applies to both non-encrypted transports (like TCP) and encrypted ones.
The problem for me as an administrator/troubleshooter is that I have
very limited visibility into what BIND 9 "sees" on its side of things
when things go south. For example, I recently experimented with getting
systemd-resolved
to talk to named
over DNS-over-TLS; the former
reported, well, errors, and I could not get a grasp of the point at
which things are failing without resorting to Wireshark ("Is it the TCP
connection on port 853 itself? Or maybe the TLS session negotiation?
Or is that part okay and it is something about the data that
system-resolved
sends inside a properly-established TLS session that
makes named
complain?" etc.)
I am opening this issue so that it can serve as a public acknowledgment
of this being a known deficiency. It would be nice to do something
about it in the log run. Obviously there will have to be performance
trade-offs, but I think even hiding certain log messages behind a
build-time switch is fine as long as there is some way of getting
named
to become more talkative logging-wise when it comes to
network-level events.