DNS query does not do any retries when dns server returned error

[olderwei]

I configured three nameserver in /etc/resolv.conf file. When DNS Server is requested, an error is returned similar to ServFail, lua-resty-dns-client and lua-resty-dns will not retry other DNS Server，Is there a specific reason behind this?

[olderwei]

@Tieske

[Tieske]

what's your resolv.conf, and what is your code?

The resolver should drop the configuration in the logs when initialized, can you show those logs?

[Tieske]

Log level must be debug, retries for example: https://github.com/Kong/lua-resty-dns-client/blob/master/src/resty/dns/client.lua#L583

[olderwei]

The log is as follows:
2019/04/30 10:08:05 [debug] 19566#0: [lua] client.lua:571: init(): [dns-client] nameserver 180.76.76.76
2019/04/30 10:08:05 [debug] 19566#0: [lua] client.lua:571: init(): [dns-client] nameserver 219.141.136.10
2019/04/30 10:08:05 [debug] 19566#0: [lua] client.lua:571: init(): [dns-client] nameserver 223.5.5.5
2019/04/30 10:08:05 [debug] 19566#0: [lua] client.lua:576: init(): [dns-client] attempts = 5

The kong's error log is as follows:
2019/04/26 21:16:10 [error] 141470#0: 373038045 [lua] balancer.lua:808: execute(): [dns] dns server error: 2 server failure. Tried: (short)..com:(na) - cache-miss
..com:1 - cache-hit/dns server error: 2 server failure
..com:33 - cache-hit/dns server error: 2 server failure
..com:5 - cache-hit/dns server error: 2 server failure
, client: 172..., server: kong, request: "GET /** HTTP/1.1", host: "..com"

I just saw this pr, which is similar to what I want to express.
openresty/lua-resty-dns#9

I wonder if the following code can be modified.
https://github.com/openresty/lua-resty-dns/blob/bb21982a51cfdab22a8c3bd4a37f92b1c56aa43d/lib/resty/dns/resolver.lua#L885
`
if err and err ~= "id mismatch" then
break
end

if answers.errcode then
break
end

if answers then
return answers, nil, tries
end
`

[olderwei]

@Tieske

[Tieske]

That is a weird log, did you redact the logs to remove the hostname? the name tried is "..com" which doesn't resolve.

Additionally the log snippet you posted above contains - cache-hit/dns server error: 2 server failure which means that you're hitting the cached error. This is not the initial error. Note that errors are also cached, see https://docs.konghq.com/1.1.x/configuration/#dns_error_ttl

What upstream hostname are you trying to resolve? And what is the initial DNS error?

[Tieske]

btw: in the error message 2 server failure actually means that the resolver reached out to the DNS server, and the DNS server responded with an error code 2, which indicates a server failure.

Essentially this means that on Kong side everything works as designed. So my guess is that you mis configured your upstream hostname since it has ..com, which is not a valid hostname.

[Tieske]

@olderwei were you able to resolve this?

Another thing that comes to mind: when the UDP reply gets truncated (answer too big), the resolver will fall back on a TCP query. Now this TCP query does not handle retries. See also openresty/lua-resty-dns#40

[Tieske]

@olderwei any update?

[Tieske]

closing due to no response